arXiv Daily Digest - 2026-04-02

Reward factorization personalizes large language models (LLMs) by decomposing rewards into shared basis functions and user-specific weights. Yet, existing methods estimate user weights from scarce data in isolation and as deterministic points, leading to inaccurate and unreliable inference. We introduce Variational Reward Factorization (VRF), an uncertainty-aware framework that represents each user's preferences as a variational distribution in a shared preference space. VRF infers user distributions via a variational encoder, derives weights through Wasserstein distance matching with shared probabilistic bases, and downweights uncertain estimates through a variance-attenuated loss. On three benchmarks, VRF outperforms all baselines across seen and unseen users, few-shot scenarios, and varying uncertainty levels, with gains extending to downstream alignment.

YouTube Shorts have become central to news consumption on the platform, yet research on how geopolitical events are represented in this format remains limited. To address this gap, we present a multimodal pipeline that combines automatic transcription, aspect-based sentiment analysis (ABSA), and semantic scene classification. The pipeline is first assessed for feasibility and then applied to analyze short-form coverage of the Israel-Hamas war by state-funded outlets. Using over 2,300 conflict-related Shorts and more than 94,000 visual frames, we systematically examine war reporting across major international broadcasters. Our findings reveal that the sentiment expressed in transcripts regarding specific aspects differs across outlets and over time, whereas scene-type classifications reflect visual cues consistent with real-world events. Notably, smaller domain-adapted models outperform large transformers and even LLMs for sentiment analysis, underscoring the value of resource-efficient approaches for humanities research. The pipeline serves as a template for other short-form platforms, such as TikTok and Instagram, and demonstrates how multimodal methods, combined with qualitative interpretation, can characterize sentiment patterns and visual cues in algorithmically driven video environments.

The direct imaging of potentially habitable exoplanets is one prime science case for high-contrast imaging instruments on extremely large telescopes. Most such exoplanets orbit close to their host stars, where their observation is limited by fast-moving atmospheric speckles and quasi-static non-common-path aberrations (NCPA). Conventional NCPA correction methods often use mechanical mirror probes, which compromise performance during operation. This work presents machine-learning-based NCPA control methods that automatically detect and correct both dynamic and static NCPA errors by leveraging sequential phase diversity. We extend previous work in reinforcement learning for AO to focal plane control. A new model-based RL algorithm, Policy Optimization for NCPAs (PO4NCPA), interprets the focal-plane image as input data and, through sequential phase diversity, determines phase corrections that optimize both non-coronagraphic and post-coronagraphic PSFs without prior system knowledge. Further, we demonstrate the effectiveness of this approach by numerically simulating static NCPA errors on a ground-based telescope and an infrared imager affected by water-vapor-induced seeing (dynamic NCPAs). Simulations show that PO4NCPA robustly compensates static and dynamic NCPAs. In static cases, it achieves near-optimal focal-plane light suppression with a coronagraph and near-optimal Strehl without one. With dynamics NCPA, it matches the performance of the modal least-squares reconstruction combined with a 1-step delay integrator in these metrics. The method remains effective for the ELT pupil, vector vortex coronagraph, and under photon and background noise. PO4NCPA is model-free and can be directly applied to standard imaging as well as to any coronagraph. Its sub-millisecond inference times and performance also make it suitable for real-time low-order correction of atmospheric turbulence beyond HCI.

We develop Structured-Knowledge-Informed Neural Networks (SKINNs), a unified estimation framework that embeds theoretical, simulated, previously learned, or cross-domain insights as differentiable constraints within flexible neural function approximation. SKINNs jointly estimate neural network parameters and economically meaningful structural parameters in a single optimization problem, enforcing theoretical consistency not only on observed data but over a broader input domain through collocation, and therefore nesting approaches such as functional GMM, Bayesian updating, transfer learning, PINNs, and surrogate modeling. SKINNs define a class of M-estimators that are consistent and asymptotically normal with root-N convergence, sandwich covariance, and recovery of pseudo-true parameters under misspecification. We establish identification of structural parameters under joint flexibility, derive generalization and target-risk bounds under distributional shift in a convex proxy, and provide a restricted-optimal characterization of the weighting parameter that governs the bias-variance tradeoff. In an illustrative financial application to option pricing, SKINNs improve out-of-sample valuation and hedging performance, particularly at longer horizons and during high-volatility regimes, while recovering economically interpretable structural parameters with improved stability relative to conventional calibration. More broadly, SKINNs provide a general econometric framework for combining model-based reasoning with high-dimensional, data-driven estimation.

We study whether phone-use agents respect privacy while completing benign mobile tasks. This question has remained hard to answer because privacy-compliant behavior is not operationalized for phone-use agents, and ordinary apps do not reveal exactly what data agents type into which form entries during execution. To make this question measurable, we introduce MyPhoneBench, a verifiable evaluation framework for privacy behavior in mobile agents. We operationalize privacy-respecting phone use as permissioned access, minimal disclosure, and user-controlled memory through a minimal privacy contract, iMy, and pair it with instrumented mock apps plus rule-based auditing that make unnecessary permission requests, deceptive re-disclosure, and unnecessary form filling observable and reproducible. Across five frontier models on 10 mobile apps and 300 tasks, we find that task success, privacy-compliant task completion, and later-session use of saved preferences are distinct capabilities, and no single model dominates all three. Evaluating success and privacy jointly reshuffles the model ordering relative to either metric alone. The most persistent failure mode across models is simple data minimization: agents still fill optional personal entries that the task does not require. These results show that privacy failures arise from over-helpful execution of benign tasks, and that success-only evaluation overestimates the deployment readiness of current phone-use agents. All code, mock apps, and agent trajectories are publicly available at~ https://github.com/tangzhy/MyPhoneBench.

Current aligned language models exhibit a dual failure mode we term the Evasive Servant: they sycophantically validate flawed user beliefs while deflecting responsibility with boilerplate disclaimers. We propose the Dignified Peer framework, which counters servility with anti-sycophancy and trustworthiness, and mitigates evasiveness through empathy and creativity. Realizing this agent requires overcoming significant challenges in data supervision, objective collapse, and evaluation bias. We address these issues by introducing the PersonaKnob dataset which features a compositional partial order structure of multiple persona preference. This data is utilized alongside a tolerant constrained Lagrangian DPO algorithm that dynamically balances all persona dimensions to prevent behavioral collapse. Additionally, we employ a psychometrically calibrated Item Response Theory evaluation protocol to disentangle latent model persona capability from confounders like judge biases. Extensive empirical studies demonstrate that our approach successfully build a LLM agent with both dignity and peer.

Reinforcement Learning (RL) has proven highly effective in addressing complex control and decision-making tasks. However, in most traditional RL algorithms, the policy is typically parameterized as a diagonal Gaussian distribution, which constrains the policy from capturing multimodal distributions, making it difficult to cover the full range of optimal solutions in multi-solution problems, and the return is reduced to a mean value, losing its multimodal nature and thus providing insufficient guidance for policy updates. In response to these problems, we propose a RL algorithm termed flow-based policy with distributional RL (FP-DRL). This algorithm models the policy using flow matching, which offers both computational efficiency and the capacity to fit complex distributions. Additionally, it employs a distributional RL approach to model and optimize the entire return distribution, thereby more effectively guiding multimodal policy updates and improving agent performance. Experimental trails on MuJoCo benchmarks demonstrate that the FP-DRL algorithm achieves state-of-the-art (SOTA) performance in most MuJoCo control tasks while exhibiting superior representation capability of the flow policy.

We show polylogarithmic mixing time bounds for the alternating-scan sampler for positively weighted restricted Boltzmann machines. This is done via analysing the same chain and the Glauber dynamics for ferromagnetic two-spin systems, where we obtain new mixing time bounds up to the critical thresholds.

Since the random language model was proposed by E. DeGiuli [Phys. Rev. Lett. 122, 128301], language models have been investigated intensively from the viewpoint of statistical mechanics. Recently, the existence of a Berezinskii--Kosterlitz--Thouless transition was numerically demonstrated in models with long-range interactions between symbols. In statistical mechanics, it has long been known that long-range interactions can induce phase transitions. Therefore, it has remained unclear whether phase transitions observed in language models originate from genuinely linguistic properties that are absent in conventional spin models. In this study, we construct a random language model with short-range interactions and numerically investigate its statistical properties. Our model belongs to the class of context-sensitive grammars in the Chomsky hierarchy and allows explicit reference to contexts. We find that a phase transition occurs even when the model refers only to contexts whose length remains constant with respect to the sentence length. This result indicates that finite-temperature phase transitions in language models are genuinely induced by the intrinsic nature of language, rather than by long-range interactions.

We introduce a differentially private manifold denoising framework that allows users to exploit sensitive reference datasets to correct noisy, non-private query points without compromising privacy. The method follows an iterative procedure that (i) privately estimates local means and tangent geometry using the reference data under calibrated sensitivity, (ii) projects query points along the privately estimated subspace toward the local mean via corrective steps at each iteration, and (iii) performs rigorous privacy accounting across iterations and queries using $(\varepsilon,δ)$-differential privacy (DP). Conceptually, this framework brings differential privacy to manifold methods, retaining sufficient geometric signal for downstream tasks such as embedding, clustering, and visualization, while providing formal DP guarantees for the reference data. Practically, the procedure is modular and scalable, separating DP-protected local geometry (means and tangents) from budgeted query-point updates, with a simple scheduler allocating privacy budget across iterations and queries. Under standard assumptions on manifold regularity, sampling density, and measurement noise, we establish high-probability utility guarantees showing that corrected queries converge toward the manifold at a non-asymptotic rate governed by sample size, noise level, bandwidth, and the privacy budget. Simulations and case studies demonstrate accurate signal recovery under moderate privacy budgets, illustrating clear utility-privacy trade-offs and providing a deployable DP component for manifold-based workflows in regulated environments without reengineering privacy systems.

Transformer-based NLP models remain vulnerable to adversarial perturbations, yet existing repair methods face a fundamental trade-off: gradient-based approaches offer flexibility but lack verifiability and often overfit; methods that do provide repair guarantees are restricted to the final layer or small networks, significantly limiting the parameter search space available for repair. We present WARP (Weight-Adjusted Repair with Provability), a constraint-based repair framework that extends repair beyond the last layer of Transformer models. WARP formulates repair as a convex quadratic program derived from a first-order linearization of the logit gap, enabling tractable optimization over a high-dimensional parameter space. Under the condition that the first-order approximation holds, this formulation induces three per-sample guarantees: (i) a positive margin constraint ensuring correct classification on repaired inputs, (ii) preservation constraints over a designated remain set, and (iii) a certified robustness radius derived from Lipschitz continuity. To ensure feasibility across varying model architectures, we introduce a sensitivity-based preprocessing step that conditions the optimization landscape accordingly. We further show that the iterative optimization procedure converges to solutions satisfying all repair constraints under mild assumptions. Empirical evaluation on encoder-only Transformers with varying layer architectures validates that these guarantees hold in practice while improving robustness to adversarial inputs. Our results demonstrate that guaranteed, generalizable Transformer repair is achievable through principled constraint-based optimization.

Continuous integration and delivery (CI/CD) pipelines are critical for sustaining the evolution of large software systems. In regulated industries with legacy technologies, however, pipelines themselves can become a source of technical debt. This paper presents an industrial case study of Bankdata, a cooperative IT provider for Danish banks, where a Jenkins-based COBOL CI/CD pipeline had grown fragile, slow, and tightly coupled to platform-specific logic. The original architecture relied on Groovy scripts spread across four repositories with runtime dependency installation, leading to long execution times, high maintenance costs, and vendor lock-in. We report on the migration to a containerized architecture featuring an abstraction layer for platform logic, simplified repository structure, and a pre-built OCI-compliant image containing COBOL tools and dependencies. The new design achieved an 82% runtime reduction. Our experience highlights lessons on abstraction, containerization, and organizational adoption, offering guidance for modernizing pipelines in legacy, high-security environments.

Existing methods for AI psychological counselors predominantly rely on supervised fine-tuning using static dialogue datasets. However, this contrasts with human experts, who continuously refine their proficiency through clinical practice and accumulated experience. To bridge this gap, we propose an Experience-Driven Lifelong Learning Agent (\texttt{PsychAgent}) for psychological counseling. First, we establish a Memory-Augmented Planning Engine tailored for longitudinal multi-session interactions, which ensures therapeutic continuity through persistent memory and strategic planning. Second, to support self-evolution, we design a Skill Evolution Engine that extracts new practice-grounded skills from historical counseling trajectories. Finally, we introduce a Reinforced Internalization Engine that integrates the evolved skills into the model via rejection fine-tuning, aiming to improve performance across diverse scenarios. Comparative analysis shows that our approach achieves higher scores than strong general LLMs (e.g., GPT-5.4, Gemini-3) and domain-specific baselines across all reported evaluation dimensions. These results suggest that lifelong learning can improve the consistency and overall quality of multi-session counseling responses.

We present DANCEMATCH, an end-to-end framework for motion-based dance retrieval, the task of identifying semantically similar choreographies directly from raw video, defined as DANCE FINGERPRINTING. While existing motion analysis and retrieval methods can compare pose sequences, they rely on continuous embeddings that are difficult to index, interpret, or scale. In contrast, DANCEMATCH constructs compact, discrete motion signatures that capture the spatio-temporal structure of dance while enabling efficient large-scale retrieval. Our system integrates Skeleton Motion Quantisation (SMQ) with Spatio-Temporal Transformers (STT) to encode human poses, extracted via Apple CoMotion, into a structured motion vocabulary. We further design DANCE RETRIEVAL ENGINE (DRE), which performs sub-linear retrieval using a histogram-based index followed by re-ranking for refined matching. To facilitate reproducible research, we release DANCETYPESBENCHMARK, a pose-aligned dataset annotated with quantised motion tokens. Experiments demonstrate robust retrieval across diverse dance styles and strong generalisation to unseen choreographies, establishing a foundation for scalable motion fingerprinting and quantitative choreographic analysis.

Adapting large language models (LLMs) to new languages is an expensive and opaque process. Understanding how language models acquire new languages and multilingual abilities is key to achieve efficient adaptation. Prior work on multilingual interpretability research focuses primarily on how trained models process multilingual instructions, leaving unexplored the mechanisms through which they acquire new languages during training. We investigate these training dynamics on decoder-only transformers through the lens of two functional cognitive specializations: language perception (input comprehension) and production (output generation). Through experiments on low-resource languages, we demonstrate how perceptual and productive specialization emerges in different regions of a language model by running layer ablation sweeps from the model's input and output directions. Based on the observed specialization patterns, we propose CogSym, a layer-wise heuristic that enables effective adaptation by exclusively fine-tuning a few early and late layers. We show that tuning only the 25% outermost layers achieves downstream task performance within 2-3% deviation from the full fine-tuning baseline. CogSym yields consistent performance with adapter methods such as LoRA, showcasing generalization beyond full fine-tuning. These findings provide insights to better understand how LLMs learn new languages and push toward accessible and inclusive language modeling.

Modern vision pipelines increasingly rely on pretrained image encoders whose representations are reused across tasks and models, yet these representations are often overcomplete and model-specific. We propose a simple, training-free method to improve the efficiency of image representations via a post-hoc canonical correlation analysis (CCA) operator. By leveraging the shared structure between representations produced by two pre-trained image encoders, our method finds linear projections that serve as a principled form of representation selection and dimensionality reduction, retaining shared semantic content while discarding redundant dimensions. Unlike standard dimensionality reduction techniques such as PCA, which operate on a single embedding space, our approach leverages cross-model agreement to guide representation distillation and refinement. The technique allows representations to be reduced by more than 75% in dimensionality with improved downstream performance, or enhanced at fixed dimensionality via post-hoc representation transfer from larger or fine-tuned models. Empirical results on ImageNet-1k, CIFAR-100, MNIST, and additional benchmarks show consistent improvements over both baseline and PCA-projected representations, with accuracy gains of up to 12.6%.

We present the GPT-NL Public Corpus, the biggest permissively licensed corpus of Dutch language resources. The GPT-NL Public Corpus contains 21 Dutch-only collections totalling 36B preprocessed Dutch tokens not present in any other LLM pretraining corpus. Additionally, the corpus includes roughly 207B English, 232B Code, and 48B German/Danish tokens taken from existing sets which we further curated for compliance. This corpus includes curated data from large existing corpora like Common Corpus and Common Crawl, as well as newly created Dutch-specific collections. Most newly created Dutch collections consist of content collected in collaboration with organisations or synthetically augmented content. All data is collected and evaluated with the aim of facilitating the creation of (commercial) language models that are lawful, useful and non-harmful. All data included in the GPT-NL Public Corpus is sourced from datasets with permissive licensing and is curated and redistributed under a CC-BY license. The full dataset is publicly available on the Hugging Face Hub.

Variational autoencoders (VAEs) learn compact latent representations of complex data, but their generative capacity is fundamentally constrained by the choice of prior distribution over the latent space. Energy-based priors offer a principled way to move beyond factorized assumptions and capture structured interactions among latent variables, yet training such priors at scale requires accurate and efficient sampling from intractable distributions. Here we present Boltzmann-machine--prior VAEs (BM-VAEs) trained using quantum annealing--based sampling in three distinct operational modes within a single generative system. During training, diabatic quantum annealing (DQA) provides unbiased Boltzmann samples for gradient estimation of the energy-based prior; for unconditional generation, slower quantum annealing (QA) concentrates samples near low-energy minima; for conditional generation, bias fields are added to direct sampling toward attribute-specific regions of the energy landscape (c-QA). Using up to 2000 qubits on a D-Wave Advantage2 processor, we demonstrate stable and efficient training across multiple datasets, with faster convergence and lower reconstruction loss than a Gaussian-prior VAE. The learned Boltzmann prior enables unconditional generation by sampling directly from the energy-based latent distribution, a capability that plain autoencoders lack, and conditional generation through latent biasing that leverages the learned pairwise interactions.

Spectral graph neural networks learn graph filters, but their behavior with increasing depth and polynomial order is not well understood. We analyze these models in the graph Fourier domain, where each layer becomes an element-wise frequency update, separating the fixed spectrum from trainable parameters and making depth and order explicit. In this setting, we show that Gaussian complexity is invariant under the Graph Fourier Transform, which allows us to derive data-dependent, depth, and order-aware generalization bounds together with stability estimates. In the linear case, our bounds are tighter, and on real graphs, the data-dependent term correlates with the generalization gap across polynomial bases, highlighting practical choices that avoid frequency amplification across layers.

The rise of large language models for code has reshaped software development. Autonomous coding agents, able to create branches, open pull requests, and perform code reviews, now actively contribute to real-world projects. Their growing role offers a unique and timely opportunity to investigate AI-driven contributions and their effects on code quality, team dynamics, and software maintainability. In this work, we construct a novel dataset of approximately $110,000$ open-source pull requests, including associated commits, comments, reviews, issues, and file changes, collectively representing millions of lines of source code. We compare five popular coding agents, including OpenAI Codex, Claude Code, GitHub Copilot, Google Jules, and Devin, examining how their usage differs in various development aspects such as merge frequency, edited file types, and developer interaction signals, including comments and reviews. Furthermore, we emphasize that code authoring and review are only a small part of the larger software engineering process, as the resulting code must also be maintained and updated over time. Hence, we offer several longitudinal estimates of survival and churn rates for agent-generated versus human-authored code. Ultimately, our findings indicate an increasing agent activity in open-source projects, although their contributions are associated with more churn over time compared to human-authored code.

Estimation of heterogeneous long-term treatment effects (HLTEs) is widely used for personalized decision-making in marketing, economics, and medicine, where short-term randomized experiments are often combined with long-term observational data. However, HLTE estimation is challenging due to limited overlap in treatment or in observing long-term outcomes for certain subpopulations, which can lead to unstable HLTE estimates with large finite-sample variance. To address this challenge, we introduce the LT-O-learners (Long-Term Orthogonal Learners), a set of novel orthogonal learners for HLTE estimation. The learners are designed for the canonical HLTE setting that combines a short-term randomized dataset $\mathcal{D}_1$ with a long-term historical dataset $\mathcal{D}_2$. The key idea of our LT-O-Learners is to retarget the learning objective by introducing custom overlap weights that downweight samples with low overlap in treatment or in long-term observation. We show that the retargeted loss is equivalent to the weighted oracle loss and satisfies Neyman-orthogonality, which means our learners are robust to errors in the nuisance estimation. We further provide a general error bound for the LT-O-Learners and give the conditions under which quasi-oracle rate can be achieved. Finally, our LT-O-learners are model-agnostic and can thus be instantiated with arbitrary machine learning models. We conduct empirical evaluations on synthetic and semi-synthetic benchmarks to confirm the theoretical properties of our LT-O-Learners, especially the robustness in low-overlap settings. To the best of our knowledge, ours are the first orthogonal learners for HLTE estimation that are robust to low overlap that is common in long-term outcomes.

2D assembly diagrams are often abstract and hard to follow, creating a need for intelligent assistants that can monitor progress, detect errors, and provide step-by-step guidance. In mixed reality settings, such systems must recognize completed and ongoing steps from the camera feed and align them with the diagram instructions. Vision Language Models (VLMs) show promise for this task, but face a depiction gap because assembly diagrams and video frames share few visual features. To systematically assess this gap, we construct IKEA-Bench, a benchmark of 1,623 questions across 6 task types on 29 IKEA furniture products, and evaluate 19 VLMs (2B-38B) under three alignment strategies. Our key findings: (1) assembly instruction understanding is recoverable via text, but text simultaneously degrades diagram-to-video alignment; (2) architecture family predicts alignment accuracy more strongly than parameter count; (3) video understanding remains a hard bottleneck unaffected by strategy. A three-level mechanistic analysis further reveals that diagrams and video occupy disjoint ViT subspaces, and that adding text shifts models from visual to text-driven reasoning. These results identify visual encoding as the primary target for improving cross-depiction robustness. Project page: https://ryenhails.github.io/IKEA-Bench/

In this work, we study whether enforcing strict compositional structure in sequence embeddings yields meaningful geometric organization when applied to protein-protein interaction networks. Using Event2Vec, an additive sequence embedding model, we train 64-dimensional representations on random walks from the human STRING interactome, and compare against a DeepWalk baseline based on Word2Vec, trained on the same walks. We find that compositional structure substantially improves pathway coherence (30.2$\times$ vs 2.9$\times$ above random), functional analogy accuracy (mean similarity 0.966 vs 0.650), and hierarchical pathway organization, while geometric properties such as norm--degree anticorrelation are shared with or exceeded by the non-compositional baseline. These results indicate that enforced compositionality specifically benefits relational and compositional reasoning tasks in biological networks.

Learning to defer (L2D) enables human-AI cooperation by deciding when an AI system should act autonomously or defer to a human expert. Existing L2D methods, however, assume static human performance, contradicting well-established findings on fatigue-induced degradation. We propose Fatigue-Aware Learning to Defer via Constrained Optimisation (FALCON), which explicitly models workload-varying human performance using psychologically grounded fatigue curves. FALCON formulates L2D as a Constrained Markov Decision Process (CMDP) whose state includes both task features and cumulative human workload, and optimises accuracy under human-AI cooperation budgets via PPO-Lagrangian training. We further introduce FA-L2D, a benchmark that systematically varies fatigue dynamics from near-static to rapidly degrading regimes. Experiments across multiple datasets show that FALCON consistently outperforms state-of-the-art L2D methods across coverage levels, generalises zero-shot to unseen experts with different fatigue patterns, and demonstrates the advantage of adaptive human-AI collaboration over AI-only or human-only decision-making when coverage lies strictly between 0 and 1.

Multi-agent Retrieval-Augmented Generation (RAG), wherein each agent takes on a specific role, supports hard queries that require multiple steps and sources, or complex reasoning. Existing approaches, however, rely on static agent behaviors and fixed orchestration strategies, leading to brittle performance on diverse, multi-hop tasks. We identify two key limitations: the lack of continuously adaptive orchestration mechanisms and the absence of behavior-level learning for individual agents. To this end, we propose HERA, a hierarchical framework that jointly evolves multi-agent orchestration and role-specific agent prompts. At the global level, HERA optimizes query-specific agent topologies through reward-guided sampling and experience accumulation. At the local level, Role-Aware Prompt Evolution refines agent behaviors via credit assignment and dual-axes adaptation along operational and behavioral principles, enabling targeted, role-conditioned improvements. On six knowledge-intensive benchmarks, HERA achieves an average improvement of 38.69\% over recent baselines while maintaining robust generalization and token efficiency. Topological analyses reveal emergent self-organization, where sparse exploration yields compact, high-utility multi-agent networks, demonstrating both efficient coordination and robust reasoning.

Machine learning-based weather forecasting models now surpass state-of-the-art numerical weather prediction systems, but training and operating these models at high spatial resolution remains computationally expensive. We present a modular framework that decouples forecasting from spatial resolution by applying learned generative super-resolution as a post-processing step to coarse-resolution forecast trajectories. We formulate super-resolution as a stochastic inverse problem, using a residual formulation to preserve large-scale structure while reconstructing unresolved variability. The model is trained with flow matching exclusively on reanalysis data and is applied to global medium-range forecasts. We evaluate (i) design consistency by re-coarsening super-resolved forecasts and comparing them to the original coarse trajectories, and (ii) high-resolution forecast quality using standard ensemble verification metrics and spectral diagnostics. Results show that super-resolution preserves large-scale structure and variance after re-coarsening, introduces physically consistent small-scale variability, and achieves competitive probabilistic forecast skill at 0.25° resolution relative to an operational ensemble baseline, while requiring only a modest additional training cost compared with end-to-end high-resolution forecasting.

As LLM agents transition from short, static problem solving to executing complex, long-horizon tasks in dynamic environments, the ability to handle user interruptions, such as adding requirement or revising goals, during mid-task execution is becoming a core requirement for realistic deployment. However, existing benchmarks largely assume uninterrupted agent behavior or study interruptions only in short, unconstrained language tasks. In this paper, we present the first systematic study of interruptible agents in long-horizon, environmentally grounded web navigation tasks, where actions induce persistent state changes. We formalize three realistic interruption types, including addition, revision, and retraction, and introduce InterruptBench, a benchmark derived from WebArena-Lite that synthesizes high-quality interruption scenarios under strict semantic constraints. Using a unified interruption simulation framework, we evaluate six strong LLM backbones across single- and multi-turn interruption settings, analyzing both their effectiveness in adapting to updated intents and their efficiency in recovering from mid-task changes. Our results show that handling user interruptions effectively and efficiently during long-horizon agentic tasks remains challenging for powerful large-scale LLMs. Code and dataset are available at https://github.com/HenryPengZou/InterruptBench.

Geometric Problem Solving (GPS) remains at the heart of enhancing mathematical reasoning in large language models because it requires the combination of diagrammatic understanding, symbolic manipulation and logical inference. In existing literature, researchers have chiefly focused on synchronising the diagram descriptions with text literals and solving the problem. In this vein, they have either taken a neural, symbolic or neuro-symbolic approach. But this solves only the first two of the requirements, namely diagrammatic understanding and symbolic manipulation, while leaving logical inference underdeveloped. The logical inference is often limited to one chain-of-thought (CoT). To address this weakness in hitherto existing models, this paper proposes MARS-GPS, that generates multiple parallel reasoning rollouts augmented with Python code execution for numerical verification, ranks them using token-level entropy as a confidence signal, and aggregates answers through a multi-stage voting and self-verification pipeline. Empirical results show that MARS-GPS with 8 parallel rollouts achieves 88.8% on Geometry3K, a nearly +11% improvement over the prior state-of-the-art, with accuracy scaling consistently as the number of rollouts increases from 1 to 16 (+6.0% on ablation subset). We provide our code and data in an anonymous repository: https://anonymous.4open.science/r/MARS-GPS-DE55.

Document understanding and GUI interaction are among the highest-value applications of Vision-Language Models (VLMs), yet they impose exceptionally heavy computational burden: fine-grained text and small UI elements demand high-resolution inputs that produce tens of thousands of visual tokens. We observe that this cost is largely wasteful -- across document and GUI benchmarks, only 22--71\% of image patches are pixel-unique, the rest being exact duplicates of another patch in the same image. We propose \textbf{PixelPrune}, which exploits this pixel-level redundancy through predictive-coding-based compression, pruning redundant patches \emph{before} the Vision Transformer (ViT) encoder. Because it operates in pixel space prior to any neural computation, PixelPrune accelerates both the ViT encoder and the downstream LLM, covering the full inference pipeline. The method is training-free, requires no learnable parameters, and supports pixel-lossless compression ($τ{=}0$) as well as controlled lossy compression ($τ{>}0$). Experiments across three model scales and document and GUI benchmarks show that PixelPrune maintains competitive task accuracy while delivering up to 4.2$\times$ inference speedup and 1.9$\times$ training acceleration. Code is available at https://github.com/OPPO-Mente-Lab/PixelPrune.

Java static analysis frameworks are commonly compared under the assumption that analysis algorithms and configurations compose monotonically and yield semantically comparable results across tools. In this work, we show that this assumption is fundamentally flawed. We present a large-scale empirical study of semantic consistency within and across four widely used Java static analysis frameworks: Soot, SootUp, WALA, and Doop. Using precision partial orders over analysis algorithms and configurations, we systematically identify violations where increased precision introduces new call-graph edges or amplifies inconsistencies. Our results reveal three key findings. First, algorithmic precision orders frequently break within frameworks due to modern language features such as lambdas, reflection, and native modeling. Second, configuration choices strongly interact with analysis algorithms, producing synergistic failures that exceed the effects of algorithm or configuration changes alone. Third, cross-framework comparisons expose irreconcilable semantic gaps, demonstrating that different frameworks operate over incompatible notions of call-graph ground truth. These findings challenge prevailing evaluation practices in static analysis and highlight the need to reason jointly about algorithms, configurations, and framework semantics when assessing precision and soundness.

Actor-level stance detection aims to determine an author expressed position toward specific geopolitical actors mentioned or implicated in a text. Although transformer-based models have achieved relatively good performance in stance classification, they typically rely on unified representations that may not sufficiently capture heterogeneous linguistic signals, such as contrastive discourse structures, framing cues, and salient lexical indicators. This motivates the need for adaptive architectures that explicitly model diverse stance-expressive patterns. In this paper, we propose StanceMoE, a context-enhanced Mixture-of-Experts (MoE) architecture built upon a fine-tuned BERT encoder for actor-level stance detection. Our model integrates six expert modules designed to capture complementary linguistic signals, including global semantic orientation, salient lexical cues, clause-level focus, phrase-level patterns, framing indicators, and contrast-driven discourse shifts. A context-aware gating mechanism dynamically weights expert contributions, enabling adaptive routing based on input characteristics. Experiments are conducted on the StanceNakba 2026 Subtask A dataset, comprising 1,401 annotated English texts where the target actor is implicit in the text. StanceMoE achieves a macro-F1 score of 94.26%, outperforming traditional baselines, and alternative BERT-based variants.

Matrix mechanisms are often used to provide unbiased differentially private query answers when publishing statistics or creating synthetic data. Recent work has developed matrix mechanisms, such as ResidualPlanner and Weighted Fourier Factorizations, that scale to high dimensional datasets while providing optimality guarantees for workloads such as marginals and circular product queries. They operate by adding noise to a linearly independent set of queries that can compactly represent the desired workloads. In this paper, we present QuerySmasher, an alternative scalable approach based on a divide-and-conquer strategy. Given a workload that can be answered from various data marginals, QuerySmasher splits each query into sub-queries and re-assembles the pieces into mutually orthogonal sub-workloads. These sub-workloads represent small, low-dimensional problems that can be independently and optimally answered by existing low-dimensional matrix mechanisms. QuerySmasher then stitches these solutions together to answer queries in the original workload. We show that QuerySmasher subsumes prior work, like ResidualPlanner (RP), ResidualPlanner+ (RP+), and Weighted Fourier Factorizations (WFF). We prove that it can dominate those approaches, under sum squared error, for all workloads. We also experimentally demonstrate the scalability and accuracy of QuerySmasher.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a central post-training paradigm for improving the reasoning capabilities of large language models. Yet existing methods share a common blind spot: they optimize policies based on instantaneous group-level or batch-level statistics without ever verifying whether the resulting update actually improved the model. This open-loop design -- updating in isolation at each step, guided only by within-group (batch) reward signals -- means optimization can drift or collapse with no mechanism to detect and correct these failures. We argue that the missing ingredient is policy improvement feedback: the ability to measure and optimize inter-iteration progress directly. To this end, we introduce Policy Improvement Reinforcement Learning (PIRL), a framework that replaces surrogate reward maximization with the explicit objective of maximizing cumulative policy improvement across iterations, and prove this temporal objective is perfectly aligned with maximizing final task performance. Building on PIRL, we propose Policy Improvement Policy Optimization (PIPO), which implements closed-loop optimization through retrospective verification. At each iteration, PIPO evaluates whether the previous update yielded genuine improvement against a sliding-window historical baseline, then actively reinforces beneficial updates and suppresses the harmful ones -- transforming an open-loop process into a self-correcting one. We provide theoretical analysis showing that PIPO performs ascent on the PIRL objective in expectation, and experiments on mathematical reasoning benchmarks demonstrate improved stability and performance over GRPO and its variants.

Large Language Models (LLMs) are increasingly applied to automate software engineering tasks, including the generation of UML class diagrams from natural language descriptions. While prior work demonstrates that LLMs can produce syntactically valid diagrams, syntactic correctness alone does not guarantee meaningful design. This study investigates whether LLMs can move beyond diagram translation to perform design synthesis, and how reliably they maintain design-oriented reasoning under variation. We introduce a preference-based few-shot prompting approach that biases LLM outputs toward designs satisfying object-oriented principles and pattern-consistent structures. Two design-intent benchmarks, each with three domain-only, paraphrased prompts and 10 repeated runs, are used to evaluate three LLMs (ChatGPT 4o-mini, Claude 3.5 Sonnet, Gemini 2.5 Flash) across three modeling strategies: standard prompting, rule-injection prompting, and preference-based prompting, totaling 540 experiments (i.e. 2x3x10x3x3). Results indicate that while preference-based alignment improves adherence to design intent it does not eliminate non-determinism, and model-level behavior strongly influences design reliability. These findings highlight that achieving dependable LLM-assisted software design requires not only effective prompting but also careful consideration of model behavior and robustness.

Proactive agents that anticipate user needs and autonomously execute tasks hold great promise as digital assistants, yet the lack of realistic user simulation frameworks hinders their development. Existing approaches model apps as flat tool-calling APIs, failing to capture the stateful and sequential nature of user interaction in digital environments and making realistic user simulation infeasible. We introduce Proactive Agent Research Environment (Pare), a framework for building and evaluating proactive agents in digital environments. Pare models applications as finite state machines with stateful navigation and state-dependent action space for the user simulator, enabling active user simulation. Building on this foundation, we present Pare-Bench, a benchmark of 143 diverse tasks spanning communication, productivity, scheduling, and lifestyle apps, designed to test context observation, goal inference, intervention timing, and multi-app orchestration.

Large language models are increasingly being deployed as autonomous agents yet their real world effectiveness depends on reliable tools for information retrieval, computation and external action. Existing studies remain fragmented across tasks, tool types, and training settings, lacking a unified view of how tool-use methods differ and evolve. This paper organizes the literature into three paradigms: prompting as plug-and-play, supervised tool learning and reward-driven tool policy learning, analyzes their methods, strengths and failure modes, reviews the evaluation landscape and highlights key challenges, aiming to address this fragmentation and provide a more structured evolutionary view of agentic tool use.

Test-Time Learning (TTL) enables language agents to iteratively refine their performance through repeated interactions with the environment at inference time. At the core of TTL is an adaptation policy that updates the actor policy based on experience from previous episodes, thereby improving future behavior. Existing methods rely on fixed, hand-crafted adaptation policies rather than optimizing them for downstream improvement. We argue that optimal adaptation policies should be learned from task environments, not hand-engineered based on human intuition. To achieve this, we introduce Meta-TTL, a framework that formulates the discovery of effective adaptation policies as a bi-level optimization problem. Within this framework, the inner loop executes the standard TTL process, measuring how effectively a candidate adaptation policy helps an agent correct errors across sequential episodes. Guided by the agent's performance, the outer loop employs evolutionary search over a diverse distribution of training tasks to iteratively refine the adaptation policy. We evaluate Meta-TTL on Jericho and WebArena-Lite across both in-distribution (ID) and out-of-distribution (OOD) settings, using multiple meta-agent backbones. Results on both benchmarks show that Meta-TTL consistently outperforms hand-crafted baselines, suggesting that the optimized adaptation policy encodes transferable strategies that generalize beyond the training task distribution.

Adapting pretrained language models (LMs) into vision-language models (VLMs) can degrade their native linguistic capability due to representation shift and cross-modal interference introduced during multimodal adaptation. Such loss is difficult to recover, even with targeted task-specific fine-tuning using standard objectives. Prior recovery approaches typically introduce additional modules that act as intermediate alignment layers to maintain or isolate modality-specific subspaces, which increases architectural complexity, adds parameters at inference time, and limits flexibility across models and settings. We propose LinguDistill, an adapter-free distillation method that restores linguistic capability by utilizing the original frozen LM as a teacher. We overcome the key challenge of enabling vision-conditioned teacher supervision by introducing layer-wise KV-cache sharing, which exposes the teacher to the student's multimodal representations without modifying the architecture of either model. We then selectively distill the teacher's strong linguistic signal on language-intensive data to recover language capability, while preserving the student's visual grounding on multimodal tasks. As a result, LinguDistill recovers $\sim$10% of the performance lost on language and knowledge benchmarks, while maintaining comparable performance on vision-heavy tasks. Our findings demonstrate that linguistic capability can be recovered without additional modules, providing an efficient and practical solution to modality-specific degradation in multimodal models.

Training effective software engineering agents requires large volumes of task-specific trajectories, incurring substantial data construction costs. Inspired by the "Less-Is-More" hypothesis in mathematical reasoning, we investigate its extension to agentic scenarios and propose an end-to-end training framework that achieves superior agentic capabilities with fewer but higher-quality training trajectories. This is achieved via STITCH (Sliding-memory Trajectory Inference and Task Chunking Heuristic), a coarse-to-fine mechanism that filters low-value noise and retains decision-critical tokens to maximize training signal quality. We conduct experiments across multiple agent frameworks (e.g., mini-SWE-agent, MSWE-agent), model scales (30B to 355B), and multilingual settings (Python, Java, and ArkTS). On SWE-bench Verified, models trained with STITCH achieve up to 63.16% relative improvement over base models. On Multi-SWE-bench (Java), MiniMax-M2.5-STITCH achieves 43.75% with our CodeArts Agent scaffold (+16.67%). On HarmonyOS (ArkTS), GLM-4.7-STITCH improves the compilation pass rate to 61.31% (+43.34%) with less than 1K training trajectories. Our results confirm that the "Less-Is-More" paradigm generalizes effectively to complex agentic tasks across diverse languages and model scales.

Low-rank decomposition has emerged as an important problem in Large Language Model (LLM) fine-tuning and inference. Through Singular Value Decomposition (SVD), the weight matrix can be factorized into low-rank spaces optimally. Previously, a common practice was to decompose the weight in the activation-whitened space, and then achieve satisfying results. In this work, we propose Optimal Brain Decomposition LLM (OBD-LLM), which studies the decomposition problem in the model space by utilizing second-order Hessian information. Through a rigorous Kronecker-factorization of the Hessian, we show that the decomposition needs to consider both input and output information of the layer, and achieves much better decomposition results compared to input only method. Our loss-aware decomposition method involves a bi-directional whitening on the weight matrix. As a result, OBD-LLM is a closed-form solution for the optimal decomposition of weights in the language model. Remarkably, we achieve ~20-40\% better results than previous state-of-the-art decomposition methods, the SVD-LLM.

Understanding emotions in natural language is inherently a multi-dimensional reasoning problem, where multiple affective signals interact through context, interpersonal relations, and situational cues. However, most existing emotion understanding benchmarks rely on short texts and predefined emotion labels, reducing this process to independent label prediction and ignoring the structured dependencies among emotions. To address this limitation, we introduce Emotional Scenarios (EmoScene), a theory-grounded benchmark of 4,731 context-rich scenarios annotated with an 8-dimensional emotion vector derived from Plutchik's basic emotions. We evaluate six instruction-tuned large language models in a zero-shot setting and observe modest performance, with the best model achieving a Macro F1 of 0.501, highlighting the difficulty of context-aware multi-label emotion prediction. Motivated by the observation that emotions rarely occur independently, we further propose an entanglement-aware Bayesian inference framework that incorporates emotion co-occurrence statistics to perform joint posterior inference over the emotion vector. This lightweight post-processing improves structural consistency of predictions and yields notable gains for weaker models (e.g., +0.051 Macro F1 for Qwen2.5-7B). EmoScene therefore provides a challenging benchmark for studying multi-dimensional emotion understanding and the limitations of current language models.

Neutral atom quantum computers (NAQCs) are among the most promising computational platforms for quantum computing. Controlling and measuring individual atoms and their states, which often requires multiple imaging and image-analysis procedures, is typically the most time-consuming task during computation and contributes significantly to overall cycle times. To resolve this challenge, we propose a highly-parallel atom-detection accelerator for tweezer-based NAQCs. Our design builds on an existing state-reconstruction method and combines an algorithm-level optimization with a Field Programmable Gate Array (FPGA) implementation to maximize parallelism and reduce the run time of the image-analysis process. We identify and overcome several challenges for an FPGA implementation, such as introducing a prefetching mechanism to improve scalability and customizing bus transfers to support large bandwidths. Tested on a Xilinx UltraScale+ FPGA, our design can analyze a 256x256-pixel fluorescence image in just 115mus, achieving 34.9x and 6.3x speedups over the original and optimized CPU baseline, respectively. Moreover, our accelerator can maintain consistent resource utilization across various atom array sizes, contributing to the ongoing efforts toward scalable and fully integrated FPGA-based control systems for NAQCs.

End-to-end autonomous driving has evolved from the conventional paradigm based on sparse perception into vision-language-action (VLA) models, which focus on learning language descriptions as an auxiliary task to facilitate planning. In this paper, we propose an alternative Vision-Geometry-Action (VGA) paradigm that advocates dense 3D geometry as the critical cue for autonomous driving. As vehicles operate in a 3D world, we think dense 3D geometry provides the most comprehensive information for decision-making. However, most existing geometry reconstruction methods (e.g., DVGT) rely on computationally expensive batch processing of multi-frame inputs and cannot be applied to online planning. To address this, we introduce a streaming Driving Visual Geometry Transformer (DVGT-2), which processes inputs in an online manner and jointly outputs dense geometry and trajectory planning for the current frame. We employ temporal causal attention and cache historical features to support on-the-fly inference. To further enhance efficiency, we propose a sliding-window streaming strategy and use historical caches within a certain interval to avoid repetitive computations. Despite the faster speed, DVGT-2 achieves superior geometry reconstruction performance on various datasets. The same trained DVGT-2 can be directly applied to planning across diverse camera configurations without fine-tuning, including closed-loop NAVSIM and open-loop nuScenes benchmarks.

We present experimental results from seven controlled runs of nanoFMT, a Free-Market Algorithm (FMA) orchestrated transformer with dynamic Mixture-of-Experts (MoE) management. The experiments address a fundamental question for advanced LLM development: how should an MoE system manage its expert pool when operating at full capacity under changing data distributions? We demonstrate that cost-penalized fitness metrics, combined with a linear grace period for newborn experts, produce a system that accumulates domain expertise through diversification rather than replacement. The central result is a round-trip domain shift experiment showing 9-11x faster recovery when returning to a previously learned domain, with zero expert births or replacements required. This "molecular memory" effect -- where dormant experts survive and reactivate when their domain returns -- has no analogue in current MoE management approaches. A preliminary cost analysis estimates annual savings of $39.1M and 27.1 GWh energy reduction for an OpenAI-scale provider under a moderate scenario.

Overlap, also known as positivity, is a key condition for causal treatment effect estimation. Many popular estimators suffer from high variance and become brittle when features differ strongly across treatment groups. This is especially challenging in high dimensions: the curse of dimensionality can make overlap implausible. To address this, we propose a class of feature representations called deconfounding scores, which preserve both identification and the target of estimation; the classical propensity and prognostic scores are two special cases. We characterize the problem of finding a representation with better overlap as minimizing an overlap divergence under a deconfounding score constraint. We then derive closed-form expressions for a class of deconfounding scores under a broad family of generalized linear models with Gaussian features and show that prognostic scores are overlap-optimal within this class. We conduct extensive experiments to assess this behavior empirically.

A group of non-cooperating agents can succumb to the \emph{tragedy-of-the-commons} if all of them seek to maximize the same resource channel to improve their viability. In nature, however, groups often avoid such collapses by differentiating into distinct roles that exploit different resource channels. It remains unclear how such coordination can emerge under continual individual-level selection alone. To address this, we introduce a computational model of multi-level selection, in which group-level selection shapes a common substrate and mutation operator shared by all group members undergoing individual-level selection. We also place this process in an embodied ecology where distinct resource channels are not segregated, but coupled through the same behavioral primitives. These channels are classified as a positive-sum intake channel and a zero-sum redistribution channel. We investigate whether such a setting can give rise to role differentiation under turnover driven by birth and death. We find that in a learned ecology, both channels remain occupied at the colony level, and the collapse into a single acquisition mode is avoided. Zero-sum channel usage increases over generations despite not being directly optimized by group-level selection. Channel occupancy also fluctuates over the lifetime of a boid. Ablation studies suggest that most baseline performance is carried by the inherited behavioral basis, while the learned variation process provides a smaller but systematic improvement prior to saturation. Together, the results suggest that multi-level selection can enable groups in a common-pool setting to circumvent tragedy-of-the-commons through differentiated use of coupled channels under continual turnover.

Standard Mixture-of-Experts (MoE) models rely on centralized routing mechanisms that introduce rigid inductive biases. We propose Routing-Free MoE which eliminates any hard-coded centralized designs including external routers, Softmax, Top-K and load balancing, instead encapsulating all activation functionalities within individual experts and directly optimized through continuous gradient flow, enabling each expert to determine its activation entirely on its own. We introduce a unified adaptive load-balancing framework to simultaneously optimize both expert-balancing and token-balancing objectives through a configurable interpolation, allowing flexible and customizable resource allocation. Extensive experiments show that Routing-Free MoE can consistently outperform baselines with better scalability and robustness. We analyze its behavior in detail and offer insights that may facilitate future MoE design ad optimization.

Plant phenology modelling aims to predict the timing of seasonal phases, such as leaf-out or flowering, from meteorological time series. Reliable predictions are crucial for anticipating ecosystem responses to climate change. While phenology modelling has traditionally relied on mechanistic approaches, deep learning methods have recently been proposed as flexible, data-driven alternatives with often superior performance. However, mechanistic models tend to outperform deep networks when data distribution shifts are induced by climate change. Domain Adaptation (DA) techniques could help address this limitation. Yet, unlike standard DA settings, climate change induces a temporal continuum of domains and involves both a covariate and label shift, with warmer records and earlier start of spring. To tackle this challenge, we introduce Mid-feature Rank-adversarial Domain Adaptation (MIRANDA). Whereas conventional adversarial methods enforce domain invariance on final latent representations, an approach that does not explicitly address label shift, we apply adversarial regularization to intermediate features. Moreover, instead of a binary domain-classification objective, we employ a rank-based objective that enforces year-invariance in the learned meteorological representations. On a country-scale dataset spanning 70 years and comprising 67,800 phenological observations of 5 tree species, we demonstrate that, unlike conventional DA approaches, MIRANDA improves robustness to climatic distribution shifts and narrows the performance gap with mechanistic models.

Spatial consistency is a fundamental property of the visual world and a key requirement for models that aim to understand physical reality. Despite recent advances, multimodal large language models (MLLMs) often struggle to reason about 3D geometry across multiple views. Rather than asking models to describe scene attributes, we introduce a more challenging task: given two views of the same scene, identify the object that violates 3D motion consistency. We propose a simple and scalable method for generating realistic, spatially inconsistent image pairs from multi-view scenes, enabling systematic evaluation of this capability. Our results show that state-of-the-art MLLMs significantly underperform human observers and exhibit substantial variability across different scene attributes, revealing a fragile and incomplete understanding of 3D structure. We hope our findings underscore the need for approaches that develop a more deeply grounded understanding of the physical world.

We propose the Preference Guided Iterated Pareto Referent Optimisation (PG-IPRO) for urban route planning for people with different accessibility requirements and preferences. With this algorithm the user can interact with the system by giving feedback on a route, i.e., the user can say which objective should be further minimized, or conversely can be relaxed. This leads to intuitive user interaction, that is especially effective during early iterations compared to information-gain-based interaction. Furthermore, due to PG-IPRO's iterative nature, the full set of alternative, possibly optimal policies (the Pareto front), is never computed, leading to higher computational efficiency and shorter waiting times for users.

While large language models (LLMs) have demonstrated strong performance on complex reasoning tasks such as competitive programming (CP), existing methods predominantly focus on single-attempt settings, overlooking their capacity for iterative refinement. In this paper, we present RefineRL, a novel approach designed to unleash the self-refinement capabilities of LLMs for CP problem solving. RefineRL introduces two key innovations: (1) Skeptical-Agent, an iterative self-refinement agent equipped with local execution tools to validate generated solutions against public test cases of CP problems. This agent always maintains a skeptical attitude towards its own outputs and thereby enforces rigorous self-refinement even when validation suggests correctness. (2) A reinforcement learning (RL) solution to incentivize LLMs to self-refine with only standard RLVR data (i.e., problems paired with their verifiable answers). Extensive experiments on Qwen3-4B and Qwen3-4B-2507 demonstrate that our method yields substantial gains: after our RL training, these compact 4B models integrated with the Skeptical-Agent not only outperform much larger 32B models but also approach the single-attempt performance of 235B models. These findings suggest that self-refinement holds considerable promise for scaling LLM reasoning, with significant potential for further advancement.

In the previous work, a lexical (re)categorisation -- or confirmation of the given category -- of roots identified as verbal was undertaken to determine their original category accurately. Building on this, the present paper offers an account of the valency classification of those Mapudungun roots confirmed to be verbal, using the language's own morphotactics; specifically, by examining the permissible and restricted combinations of various suffixes with roots or verbal stems in the Mapuche verb form. As with all work conducted thus far, the results presented here aim to improve the morphological analyser (Dungupeyum) with all verified findings incorporated into the system. From a theoretical perspective, we also hope to contribute to the recognition and understanding of issues related to the valency of Mapuche verb forms.

This technical report presents methods developed by the UK AI Security Institute for assessing whether advanced AI systems reliably follow intended goals. Specifically, we evaluate whether frontier models sabotage safety research when deployed as coding assistants within an AI lab. Applying our methods to four frontier models, we find no confirmed instances of research sabotage. However, we observe that Claude Opus 4.5 Preview (a pre-release snapshot of Opus 4.5) and Sonnet 4.5 frequently refuse to engage with safety-relevant research tasks, citing concerns about research direction, involvement in self-training, and research scope. We additionally find that Opus 4.5 Preview shows reduced unprompted evaluation awareness compared to Sonnet 4.5, while both models can distinguish evaluation from deployment scenarios when prompted. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold designed to simulate realistic internal deployment of a coding agent. We validate that this scaffold produces trajectories that all tested models fail to reliably distinguish from real deployment data. We test models across scenarios varying in research motivation, activity type, replacement threat, and model autonomy. Finally, we discuss limitations including scenario coverage and evaluation awareness.

The Mining Software Repositories (MSR) field focuses on analysing the rich data contained in software repositories to derive actionable insights into software processes and products. Mining repositories at scale requires techniques capable of handling large volumes of heterogeneous data, a challenge for which language models (LMs) are increasingly well-suited. Since the advent of Transformer-based architectures, LMs have been rapidly adopted across a wide range of MSR tasks. This article presents a comprehensive survey of the use of LMs in MSR, based on an analysis of 85 papers. We examine how LMs are applied, the types of artefacts analysed, which models are used, how their adoption has evolved over time, and the extent to which studies support reproducibility and reuse. Building on this analysis, we propose a taxonomy of LM applications in MSR, identify key trends shaping the field, and highlight open challenges alongside actionable directions for future research.

The L infinity star discrepancy is a measure for how uniformly a point set is distributed in a given space. Point sets of low star discrepancy are used as designs of experiments, as initial designs for Bayesian optimization algorithms, for quasi-Monte Carlo integration methods, and many other applications. Recent work has shown that classical constructions such as Sobol', Halton, or Hammersley sequences can be outperformed by large margins when considering point sets of fixed sizes rather than their convergence behavior. These results, highly relevant to the aforementioned applications, raise the question of how much existing constructions can be improved through size-specific optimization. In this work, we study this question for the so-called Kronecker construction. Focusing on the 3-dimensional setting, we show that optimizing the two configurable parameters of its construction yields point sets outperforming the state-of-the-art value for sets of at least 500 points. Using the algorithm configuration technique irace, we then derive parameters that yield new state-of-the-art discrepancy values for whole ranges of set sizes.

Pretraining Large Language Models (LLMs) from scratch requires massive amount of compute. Aurora super computer is an ExaScale machine with 127,488 Intel PVC (Ponte Vechio) GPU tiles. In this work, we showcase LLM pretraining on Aurora at the scale of 1000s of GPU tiles. Towards this effort, we developed Optimus, an inhouse training library with support for standard large model training techniques. Using Optimus, we first pretrained Mula-1B, a 1 Billion dense model and Mula-7B-A1B, a 7 Billion Mixture of Experts (MoE) model from scratch on 3072 GPU tiles for the full 4 trillion tokens of the OLMoE-mix-0924 dataset. We then demonstrated model scaling by pretraining three large MoE models Mula-20B-A2B, Mula-100B-A7B, and Mula-220B-A10B till 100 Billion tokens on the same dataset. On our largest model Mula-220B-A10B, we pushed the compute scaling from 384 to 12288 GPU tiles and observed scaling efficiency of around 90% at 12288 GPU tiles. We significantly improved the runtime performance of MoE models using custom GPU kernels for expert computation, and a novel EP-Aware sharded optimizer resulting in training speedups up to 1.71x. As part of the Optimus library, we also developed a robust set of reliability and fault tolerant features to improve training stability and continuity at scale.

Multi-FPGA systems (MFS) are widely adopted for VLSI emulation and rapid prototyping. In an MFS, FPGAs connect only to a limited number of neighbors through bandwidth-constrained links, so inter-FPGA communication cost depends on network topology. This setting exposes two fundamental limitations of existing MFS-aware partitioning methods: conventional hypergraph partitioners focus solely on cut size and ignore topological structure, and they leave substantial FPGA resources unused due to conservative balance margins. We present RePart, a fully customized multilevel hypergraph partitioning framework for MFS that integrates logic replication with topology-aware optimization. RePart introduces three coordinated innovations across the multilevel pipeline: FPGA-aware dynamic coarsening, heat-value guided assignment, and replication-deletion supported refinement. Extensive experiments on the Titan23 and EDA Elite Challenge Contest benchmarks show that RePart reduces total hop distance by 52.3% on average over state-of-the-art hypergraph partitioners with an 11.1x speedup, and outperforms the EDA Elite Challenge winners. Code is available at: https://github.com/Welement-zyf/RePart.

Label prediction in neural networks (NNs) has O(n) complexity proportional to the number of classes. This holds true for classification using fully connected layers and cosine similarity with some set of class prototypes. In this paper we show that if NN latent space (LS) geometry is known and possesses specific properties, label prediction complexity can be significantly reduced. This is achieved by associating label prediction with the O(1) complexity closest cluster center search in a vector system used as target for latent space configuration (LSC). The proposed method only requires finding indexes of several largest and lowest values in the embedding vector making it extremely computationally efficient. We show that the proposed method does not change NN training accuracy computational results. We also measure the time required by different computational stages of NN inference and label prediction on multiple datasets. The experiments show that the proposed method allows to achieve up to 11.6 times overall acceleration over conventional methods. Furthermore, the proposed method has unique properties which allow to predict the existence of new classes.

Large language models (LLMs) exhibit failures on elementary symbolic tasks such as character counting in a word, despite excelling on complex benchmarks. Although this limitation has been noted, the internal reasons remain unclear. We use character counting (e.g., "How many p's are in apple?") as a minimal, controlled probe that isolates token-level reasoning from higher-level confounds. Using this setting, we uncover a consistent phenomenon across modern architectures, including LLaMA, Qwen, and Gemma: models often compute the correct answer internally yet fail to express it at the output layer. Through mechanistic analysis combining probing classifiers, activation patching, logit lens analysis, and attention head tracing, we show that character-level information is encoded in early and mid-layer representations. However, this information is attenuated by a small set of components in later layers, especially the penultimate and final layer MLP. We identify these components as negative circuits: subnetworks that downweight correct signals in favor of higher-probability but incorrect outputs. Our results lead to two contributions. First, we show that symbolic reasoning failures in LLMs are not due to missing representations or insufficient scale, but arise from structured interference within the model's computation graph. This explains why such errors persist and can worsen under scaling and instruction tuning. Second, we provide evidence that LLM forward passes implement a form of competitive decoding, in which correct and incorrect hypotheses coexist and are dynamically reweighted, with final outputs determined by suppression as much as by amplification. These findings carry implications for interpretability and robustness: simple symbolic reasoning exposes weaknesses in modern LLMs, underscoring need for design strategies that ensure information is encoded and reliably used.

Mental health text classification has rapidly adopted modern adaptation methods, yet practical guidance on which optimization strategy to use, when, and why remains limited. This paper presents a systematic comparative study of optimization pathways for a joint mental-health classification task, moving from strong vanilla baselines to progressively more specialized techniques. We first establish classical and encoder references, then examine parameter-efficient supervised fine-tuning with LoRA/QLoRA under multiple objective and optimization settings, and finally evaluate preference-based optimization with DPO, ORPO, and KTO, including class-rebalanced training. Rather than emphasizing a single headline score, we focus on methodological insight: how performance changes with objective formulation, adapter choice, optimizer behavior, context windowing, and class-balance intervention. The results show that optimization effects are highly method-dependent: some approaches deliver stable, transferable gains, while others are sensitive to configuration and data balance. Preference optimization, in particular, exhibits large variation across objectives, indicating that method selection is more consequential than simply adding a preference-training stage. The central contribution is a clear optimization narrative for mental health NLP: start from transparent baselines, apply controlled tuning, and use preference optimization selectively where its gains are demonstrable. This provides a reproducible and practically grounded framework for choosing effective training strategies beyond architecture choice alone.

A new generation of language models reasons entirely in continuous hidden states, producing no tokens and leaving no audit trail. We show that this silence creates a fundamentally new attack surface. ThoughtSteer perturbs a single embedding vector at the input layer; the model's own multi-pass reasoning amplifies this perturbation into a hijacked latent trajectory that reliably produces the attacker's chosen answer, while remaining structurally invisible to every token-level defense. Across two architectures (Coconut and SimCoT), three reasoning benchmarks, and model scales from 124M to 3B parameters, ThoughtSteer achieves >=99% attack success rate with near-baseline clean accuracy, transfers to held-out benchmarks without retraining (94-100%), evades all five evaluated active defenses, and survives 25 epochs of clean fine-tuning. We trace these results to a unifying mechanism: Neural Collapse in the latent space pulls triggered representations onto a tight geometric attractor, explaining both why defenses fail and why any effective backdoor must leave a linearly separable signature (probe AUC>=0.999). Yet a striking paradox emerges: individual latent vectors still encode the correct answer even as the model outputs the wrong one. The adversarial information is not in any single vector but in the collective trajectory, establishing backdoor perturbations as a new lens for mechanistic interpretability of continuous reasoning. Code and checkpoints are available.

Wearable HAR has improved steadily, but most progress still relies on closed-set classification, which limits real-world use. In practice, human activity is open-ended, unscripted, personalized, and often compositional, unfolding as narratives rather than instances of fixed classes. We argue that addressing this gap does not require simply scaling datasets or models. It requires a fundamental shift in how wearable HAR is formulated, supervised, and evaluated. This work shows how to model open-ended activity narratives by aligning wearable sensor data with natural-language descriptions in an open-vocabulary setting. Our framework has three core components. First, we introduce a naturalistic data collection and annotation pipeline that combines multi-position wearable sensing with free-form, time-aligned narrative descriptions of ongoing behavior, allowing activity semantics to emerge without a predefined vocabulary. Second, we define a retrieval-based evaluation framework that measures semantic alignment between sensor data and language, enabling principled evaluation without fixed classes while also subsuming closed-set classification as a special case. Third, we present a language-conditioned learning architecture that supports sensor-to-text inference over variable-length sensor streams and heterogeneous sensor placements. Experiments show that models trained with fixed-label objectives degrade sharply under real-world variability, while open-vocabulary sensor-language alignment yields robust and semantically grounded representations. Once this alignment is learned, closed-set activity recognition becomes a simple downstream task. Under cross-participant evaluation, our method achieves 65.3% Macro-F1, compared with 31-34% for strong closed-set HAR baselines. These results establish open-ended narrative modeling as a practical and effective foundation for real-world wearable HAR.

Large Vision Language Models show impressive performance across image and video understanding tasks, yet their computational cost grows rapidly with the number of visual tokens. Existing token pruning methods mitigate this issue through empirical approaches while overlooking the internal mechanism of attention. In this paper, we propose a novel training free token pruning framework grounded in the dual form perspective of attention. We reformulate attention as an implicit linear layer whose weight matrix is the sum of rank 1 outer products, each generated by a single token's key value pair. Token pruning thus reduces to selecting an optimal subset of these rank 1 updates that best approximates the original dual weight matrix. Extending this perspective to standard softmax attention in LVLMs, we derive a novel metric quantifying both a token's information magnitude and information duplication. To efficiently select the subset with the proposed metric, we introduce Progressive Chunked Maximal Marginal Relevance. Extensive experiments demonstrate that our method achieves a better trade off between performance and efficiency, while providing another perspective on existing pruning approaches.

The whole-brain connectome of a fruit fly comprises over 130K neurons connected with a probability of merely 0.02%, yet achieves an average shortest path of only 4.4 hops. Despite being highly structured at the circuit level, the network's long-range connections are broadly distributed across brain regions, functioning as stochastic shortcuts that enable efficient global communication. Inspired by this observation, we propose Stochastic Attention (SA), a drop-in enhancement for sliding-window attention (SWA) that applies a random permutation to the token sequence before windowed attention and restores the original order afterward. This transforms the fixed local window into a stochastic global one within the same $O(nw)$ per-layer budget. Through depth, independently sampled permutations yield exponentially growing receptive fields, achieving full sequence coverage in $O(\log_w n)$ layers versus $O(n/w)$ for SWA. We validate SA in two settings: pre-training language models from scratch, where a gated SA + SWA combination achieves the best average zero-shot accuracy, and training-free inference on Qwen3-8B and Qwen3-30B-A3B, where SA consistently outperforms SWA and matches or exceeds Mixture of Block Attention at comparable compute budgets. These results suggest that connectome-inspired stochastic routing is a practical primitive for improving the expressivity of efficient attention, complementary to existing linear and sparse approaches.

Datasets used in immunotherapy response prediction are typically small in size, as well as diverse in cancer type, drug administered, and sequencer used. Models often drop in performance when tested on patient cohorts that are not included in the training process. Recent work has shown that transformer-based models along with self-supervised learning show better generalisation performance than threshold-based biomarkers, but is still suboptimal. We present BioCOMPASS, an extension of a transformer-based model called COMPASS, that integrates biomarkers and treatment information to further improve its generalisability. Instead of feeding biomarker data as input, we built loss components to align them with the model's intermediate representations. We found that components such as treatment gating and pathway consistency loss improved generalisability when evaluated with Leave-one-cohort-out, Leave-one-cancer-type-out and Leave-one-treatment-out strategies. Results show that building components that exploit biomarker and treatment information can help in generalisability of immunotherapy response prediction. Careful curation of additional components that leverage complementary clinical information and domain knowledge represents a promising direction for future research.

Gaussian processes are widely used in machine learning domains but remain computationally demanding, limiting their efficient scalability across diverse hardware platforms. The GPRat library targets these challenges with the help of the asynchronous many-task runtime system HPX. In this work, we extend GPRat to enable portability across multiple hardware architectures and evaluate its performance on representative x86-64, ARM, and RISC-V chips. We conduct node-level strong-scaling and problem-size-scaling benchmarks for Gaussian Process prediction and hyperparameter optimization to assess single-core performance, parallel scalability, and architectural efficiency. Our results show that while the x86-64 Zen 2 chip achieves a 58% single-core performance advantage over the ARM-based Fujitsu A64FX, superior parallel scaling allows the 48-core ARM chip to outperform the 64-core Zen 2 by 9% at full node utilization. The evaluated SOPHON SG2042 RISC-V chip exhibits substantially lower performance and weaker scalability, with single-core performance lagging by up to a factor of 14 and large-scale parallel workloads showing slowdowns of up to a factor of 25. For problem-size scaling, ARM and x86-64 systems demonstrate comparable performance within 25%. These findings highlight the growing competitiveness of ARM-based processors and emphasize the importance of wide-register vectorization support and memory subsystem improvements for upcoming RISC-V platforms.

The memory wall remains the primary bottleneck for training large language models on consumer hardware. We introduce Spectral Compact Training (SCT), a method that replaces dense weight matrices with permanent truncated SVD factors W = U diag(s) V^T, where the full dense matrix is never materialized during training or inference. Gradients flow through the compact spectral factors via standard backpropagation, and U, V are retracted to the Stiefel manifold via QR decomposition after each optimizer step. SCT achieves up to 199x memory reduction per MLP layer at rank 32, enabling full training steps of 70B-parameter architectures on a Steam Deck handheld (7.2 GB peak memory vs. 1,245 GB for dense FP32 training with Adam). Rank-sweep experiments on SmolLM2-1.7B (ranks 32-256, 2000 steps, NVIDIA A100) show that all tested ranks converge to the same loss floor (~4.2-4.5), identifying the learning rate schedule -- not MLP rank -- as the primary bottleneck. Rank 128 emerges as the efficiency sweet spot at 11.7x MLP compression with the lowest perplexity. GPU memory drops 46% at rank 32 while training throughput doubles.

Context: Schools, training platforms, and technology firms increasingly need to assess programming proficiency at scale with transparent, reproducible methods that support personalized learning pathways. Objective: This study introduces a pedagogical framework for Scratch project assessment, aligned with the Common European Framework of Reference (CEFR), providing universal competency levels for students and teachers alongside actionable insights for curriculum design. Method: We apply Fuzzy C-Means clustering to 2008246 Scratch projects evaluated via Dr.Scratch, implementing an ordinal criterion to map clusters to CEFR levels (A1-C2), and introducing enhanced classification metrics that identify transitional learners, enable continuous progress tracking, and quantify classification certainty to balance automated feedback with instructor review. Impact: The framework enables diagnosis of systemic curriculum gaps-notably a "B2 bottleneck" where only 13.3% of learners reside due to the cognitive load of integrating Logic Synchronization, and Data Representation--while providing certainty--based triggers for human intervention.

As Large Language Models (LLMs) scale in size and complexity, the consequences of failures during training become increasingly severe. A major challenge arises from Silent Data Corruption (SDC): hardware-induced faults that bypass system-level detection mechanisms. SDC may behave like benign numerical noise, but can also cause harmful gradient corruption that leads to loss spikes, divergence, or stalled progress. This work provides a controlled study of how intermittent SDC affects LLM pretraining. Using targeted fault injection at the level of GPU matrix-multiply instructions, we characterize the sensitivity of different bit positions, kernel functions, and execution stages. Our analysis shows that locally originating faults can produce impactful corruption, including NaN propagation, short-lived spikes in loss, gradient norm, and attention logits, as well as persistent parameter divergence. Building on the observed corruption signatures, we propose a lightweight detection method that identifies potentially harmful parameter updates. Experiments on LLaMA models with 60M, 350M, and 1.3B parameters demonstrate that recomputing the most recent training step upon detection can effectively mitigate the impact of these events.

End-to-end OCR for historical newspapers remains challenging, as models must handle long text sequences, degraded print quality, and complex layouts. While Transformer-based recognizers dominate current research, their quadratic complexity limits efficient paragraph-level transcription and large-scale deployment. We investigate linear-time State-Space Models (SSMs), specifically Mamba, as a scalable alternative to Transformer-based sequence modeling for OCR. We present to our knowledge, the first OCR architecture based on SSMs, combining a CNN visual encoder with bi-directional and autoregressive Mamba sequence modeling, and conduct a large-scale benchmark comparing SSMs with Transformer- and BiLSTM-based recognizers. Multiple decoding strategies (CTC, autoregressive, and non-autoregressive) are evaluated under identical training conditions alongside strong neural baselines (VAN, DAN, DANIEL) and widely used off-the-shelf OCR engines (PERO-OCR, Tesseract OCR, TrOCR, Gemini). Experiments on historical newspapers from the Bibliothèque nationale du Luxembourg, with newly released >99% verified gold-standard annotations, and cross-dataset tests on Fraktur and Antiqua lines, show that all neural models achieve low error rates (~2% CER), making computational efficiency the main differentiator. Mamba-based models maintain competitive accuracy while halving inference time and exhibiting superior memory scaling (1.26x vs 2.30x growth at 1000 chars), reaching 6.07% CER at the severely degraded paragraph level compared to 5.24% for DAN, while remaining 2.05x faster. We release code, trained models, and standardized evaluation protocols to enable reproducible research and guide practitioners in large-scale cultural heritage OCR.

We obtain better algorithms for computing more balanced orientations and degree splits in LOCAL. Important to our result is a connection to the hypergraph sinkless orientation problem [BMNSU, SODA'25] We design an algorithm of complexity $\mathcal{O}(\varepsilon^{-1} \cdot \log n)$ for computing a balanced orientation with discrepancy at most $\varepsilon \cdot \mathrm{deg}(v)$ for every vertex $v \in V$. This improves upon a previous result by [GHKMSU, Distrib. Comput. 2020] of complexity $\mathcal{O}(\varepsilon^{-1} \cdot \log \varepsilon^{-1} \cdot (\log \log \varepsilon^{-1})^{1.71} \cdot \log n)$. Further, we show that this result can also be extended to compute undirected degree splits with the same discrepancy and in the same runtime. As as application we show that $(3 / 2 + \varepsilon)Δ$-edge coloring can now be solved in $\mathcal{O}(\varepsilon^{-1} \cdot \log^2 Δ\cdot \log n + \varepsilon^{-2} \cdot \log n)$ rounds in LOCAL. Note that for constant $\varepsilon$ and $Δ= \mathcal{O}(2^{\log^{1/3} n})$ this runtime matches the current state-of-the-art for $(2Δ- 1)$-edge coloring in [Ghaffari & Kuhn, FOCS'21].

Large language model (LLM) agents struggle to autonomously evolve coordination strategies in dynamic environments, largely because coarse global outcomes obscure the causal signals needed for local policy refinement. We identify this bottleneck as a multi-agent credit assignment problem, which has long been studied in classical multi-agent reinforcement learning (MARL) but remains underaddressed in LLM-based systems. Building on this observation, we propose LangMARL, a framework that brings credit assignment and policy gradient evolution from cooperative MARL into the language space. LangMARL introduces agent-level language credit assignment, pioneers gradient evolution in language space for policy improvement, and summarizes task-relevant causal relations from replayed trajectories to provide dense feedback and improve convergence under sparse rewards. Extensive experiments across diverse cooperative multi-agent tasks demonstrate improved sample efficiency, interpretability, and strong generalization.

Non-stationarity arises from concurrent policy updates and leads to persistent environmental fluctuations. Existing approaches like Centralized Training with Decentralized Execution (CTDE) and sequential update schemes mitigate this issue. However, since the perception of the policies of other agents remains dependent on sampling environmental interaction data, the agent essentially operates in a passive perception state. This inevitably triggers equilibrium oscillations and significantly slows the convergence speed of the system. To address this issue, we propose Gradient Realignment via Active Shared Perception (GRASP), a novel framework that defines generalized Bellman equilibrium as a stable objective for policy evolution. The core mechanism of GRASP involves utilizing the independent gradients of agents to derive a defined consensus gradient, enabling agents to actively perceive policy updates and optimize team collaboration. Theoretically, we leverage the Kakutani Fixed-Point Theorem to prove that the consensus direction $u^*$ guarantees the existence and attainability of this equilibrium. Extensive experiments on StarCraft II Multi-Agent Challenge (SMAC) and Google Research Football (GRF) demonstrate the scalability and promising performance of the framework.

Transformer language models contain localized reasoning circuits, contiguous layer blocks that improve reasoning when duplicated at inference time. Finding these circuits currently requires brute-force sweeps costing 25 GPU hours per model. We propose CircuitProbe, which predicts circuit locations from activation statistics in under 5 minutes on CPU, providing a speedup of three to four orders of magnitude. We find that reasoning circuits come in two types: stability circuits in early layers, detected through the derivative of representation change, and magnitude circuits in late layers, detected through anomaly scoring. We validate across 9 models spanning 6 architectures, including 2025 models, confirming that CircuitProbe top predictions match or are within 2 layers of the optimal circuit in all validated cases. A scaling experiment across the Qwen 2.5 family reveals that layer duplication consistently benefits models under 3B parameters but degrades performance in 7B+ models, making this a practical scaling technique for small language models. CircuitProbe requires as few as 10 calibration examples and its predictions are stable across English, Hindi, Chinese, and French.

Retrieval-augmented generation (RAG) improves language model (LM) performance by providing relevant context at test time for knowledge-intensive situations. However, the relationship between parametric knowledge acquired during pretraining and non-parametric knowledge accessed via retrieval remains poorly understood, especially under fixed data budgets. In this work, we systematically study the trade-off between pretraining corpus size and retrieval store size across a wide range of model and data scales. We train OLMo-2-based LMs ranging from 30M to 3B parameters on up to 100B tokens of DCLM data, while varying both pretraining data scale (1-150x the number of parameters) and retrieval store size (1-20x), and evaluate performance across a diverse suite of benchmarks spanning reasoning, scientific QA, and open-domain QA. We find that retrieval consistently improves performance over parametric-only baselines across model scales and introduce a three-dimensional scaling framework that models performance as a function of model size, pretraining tokens, and retrieval corpus size. This scaling manifold enables us to estimate optimal allocations of a fixed data budget between pretraining and retrieval, revealing that the marginal utility of retrieval depends strongly on model scale, task type, and the degree of pretraining saturation. Our results provide a quantitative foundation for understanding when and how retrieval should complement pretraining, offering practical guidance for allocating data resources in the design of scalable language modeling systems.

Assessing the veracity of a claim made online is a complex and important task with real-world implications. When these claims are directed at communities with limited access to information and the content concerns issues such as healthcare and culture, the consequences intensify, especially in low-resource languages. In this work, we introduce AfrIFact, a dataset that covers the necessary steps for automatic fact-checking (i.e., information retrieval, evidence extraction, and fact checking), in ten African languages and English. Our evaluation results show that even the best embedding models lack cross-lingual retrieval capabilities, and that cultural and news documents are easier to retrieve than healthcare-domain documents, both in large corpora and in single documents. We show that LLMs lack robust multilingual fact-verification capabilities in African languages, while few-shot prompting improves performance by up to 43% in AfriqueQwen-14B, and task-specific fine-tuning further improves fact-checking accuracy by up to 26%. These findings, along with our release of the AfrIFact dataset, encourage work on low-resource information retrieval, evidence retrieval, and fact checking.

Large-scale web applications are widely deployed with complex third-party components, inheriting security risks arising from component vulnerabilities. Security assessment is therefore required to determine whether such known vulnerabilities remain practically exploitable in real applications. Penetration testing is a widely adopted approach that validates exploitability by launching concrete attacks against known vulnerabilities in real-world black-box systems. However, existing approaches often fail to automatically generate reliable exploits, limiting their effectiveness in practical security assessment. This limitation mainly stems from two issues: (1) precisely triggering vulnerabilities with correct technical details, and (2) adapting exploits to diverse real-world deployment settings. In this paper, we propose AutoEG, a fully automated multi-agent framework for exploit generation targeting black-box web applications. AutoEG has two phases: First, AutoEG extracts precise vulnerability trigger logic from unstructured vulnerability information and encapsulates it into reusable trigger functions. Second, AutoEG uses trigger functions for concrete attack objectives and iteratively refines exploits through feedback-driven interaction with the target application. We evaluate AutoEG on 104 real-world vulnerabilities with 29 attack objectives, resulting in 660 exploitation tasks and 55,440 exploit attempts. AutoEG achieves an average success rate of 82.41%, substantially outperforming state-of-the-art baselines, whose best performance reaches only 32.88%.

Neural Architecture Search (NAS) has become a pivotal technique in automated machine learning. Evolutionary Algorithm (EA)-based methods demonstrate superior search quality but suffer from prohibitive computational costs, while gradient-based approaches like DARTS offer high efficiency but are prone to premature convergence and performance collapse. To bridge this gap, we propose G-ICSO-NAS, a hybrid framework implementing a three-stage optimization strategy. The Warm-up Phase pre-trains supernet weights ($w$) via differentiable methods while architecture parameters ($α$) remain frozen. The Exploration Phase adopts a hybrid co-optimization mechanism: an Improved Competitive Swarm Optimizer (ICSO) with diversity-aware fitness navigates the architecture space to update $α$, while gradient descent concurrently updates $w$. The Stability Phase employs fine-grained gradient-based search with early stopping to converge to the optimal architecture. By synergizing ICSO's global navigation capability with differentiable methods' efficiency, G-ICSO-NAS achieves remarkable performance with minimal cost. In the context of the DARTS search space, an accuracy of 97.46\% is achieved on CIFAR-10 with a computational budget of just 0.15 GPU-Days. The method also exhibits strong transfer potential, recording accuracies of 83.1\% (CIFAR-100) and 75.02\% (ImageNet). Furthermore, regarding the NAS-Bench-201 benchmark, G-ICSO-NAS is shown to deliver state-of-the-art results across all evaluated datasets.

Due to their widespread use in industry, several techniques have been proposed in the literature to fuzz REST APIs. Existing fuzzers for REST APIs have been focusing on detecting crashes (e.g., 500 HTTP server error status code). However, security vulnerabilities can have major drastic consequences on existing cloud infrastructures. In this paper, we propose a series of novel automated oracles aimed at detecting violations of access policies in REST APIs, as well as executing traditional attacks such as SQL Injection and XSS. These novel automated oracles can be integrated into existing fuzzers, in which, once the fuzzing session is completed, a ``security testing'' phase is executed to verify these oracles. When a security fault is detected, as output our technique is able to general executable test cases in different formats, like Java, Kotlin, Python and JavaScript test suites. Our novel techniques are integrated as an extension of EvoMaster, a state-of-the-art open-source fuzzer for REST APIs. Experiments are carried out on 9 artificial examples, 8 vulnerable-by-design REST APIs with black-box testing, and 36 REST APIs from the WFD corpus with white-box testing, for a total of 52 distinct APIs. Results show that our novel oracles and their automated integration in a fuzzing process can lead to detect security issues in several of these APIs.

Group Relative Policy Optimization (GRPO) is widely used for reinforcement learning with verifiable rewards, but it often suffers from advantage collapse: when all rollouts in a group receive the same reward, the group yields zero relative advantage and thus no learning signal. For example, if a question is too hard for the reasoner, all sampled rollouts can be incorrect and receive zero reward. Recent work addresses this issue by adding hints or auxiliary scaffolds to such hard questions so that the reasoner produces mixed outcomes and recovers a non-zero update. However, existing hints are usually fixed rather than adapted to the current reasoner, and a hint that creates learning signal under the hinted input does not necessarily improve the no-hint policy used at test time. To this end, we propose Hint Learning for Reinforcement Learning (HiLL), a framework that jointly trains a hinter policy and a reasoner policy during RL. For each hard question, the hinter generates hints online conditioned on the current reasoner's incorrect rollout, allowing hint generation to adapt to the reasoner's evolving errors. We further introduce hint reliance, which measures how strongly correct hinted trajectories depend on the hint. We derive a transferability result showing that lower hint reliance implies stronger transfer from hinted success to no-hint success, and we use this result to define a transfer-weighted reward for training the hinter. Therefore, HiLL favors hints that not only recover informative GRPO groups, but also produce signals that are more likely to improve the original no-hint policy. Experiments across multiple benchmarks show that HiLL consistently outperforms GRPO and prior hint-based baselines, demonstrating the value of adaptive and transfer-aware hint learning for RL. The code is available at https://github.com/Andree-9/HiLL.

Gaussian processes (GPs) offer appealing properties but are costly to train at scale. Sparse variational GP (SVGP) approximations reduce cost yet still rely on Cholesky decompositions of kernel matrices, ill-suited to low-precision, massively parallel hardware. While one can construct valid variational bounds that rely only on matrix multiplications (matmuls) via an auxiliary matrix parameter, optimising them with off-the-shelf first-order methods is challenging. We make the inverse-free approach practical by proposing a better-conditioned bound and deriving a matmul-only natural-gradient update for the auxiliary parameter, markedly improving stability and convergence. We further provide simple heuristics, such as step-size schedules and stopping criteria, that make the overall optimisation routine fit seamlessly into existing workflows. Across regression and classification benchmarks, we demonstrate that our method 1) serves as a drop-in replacement in SVGP-based models (e.g., deep GPs), 2) recovers similar performance to traditional methods, and 3) can be faster than baselines when well tuned.

Autonomous agents increasingly interact with the web, yet most websites remain designed for human browsers -- a fundamental mismatch that the emerging ``Agentic Web'' must resolve. Agents must repeatedly browse pages, inspect DOMs, and reverse-engineer callable routes -- a process that is slow, brittle, and redundantly repeated across agents. We observe that every modern website already exposes internal APIs (sometimes called \emph{shadow APIs}) behind its user interface -- first-party endpoints that power the site's own functionality. We present Unbrowse, a shared route graph that transforms browser-based route discovery into a collectively maintained index of these callable first-party interfaces. The system passively learns routes from real browsing traffic and serves cached routes via direct API calls. In a single-host live-web benchmark of equivalent information-retrieval tasks across 94 domains, fully warmed cached execution averaged 950\,ms versus 3{,}404\,ms for Playwright browser automation (3.6$\times$ mean speedup, 5.4$\times$ median), with well-cached routes completing in under 100\,ms. A three-path execution model -- local cache, shared graph, or browser fallback -- ensures the system is voluntary and self-correcting. A three-tier micropayment model via the x402 protocol charges per-query search fees for graph lookups (Tier~3), a one-time install fee for discovery documentation (Tier~1), and optional per-execution fees for site owners who opt in (Tier~2). All tiers are grounded in a necessary condition for rational adoption: an agent uses the shared graph only when the total fee is lower than the expected cost of browser rediscovery.

We consider the problem of constructing surrogate operators for parameter-to-solution maps arising from parametric partial differential equations, where repeated forward model evaluations are computationally expensive. We present a systematic empirical comparison of neural operator surrogates, including a reduced-basis neural operator trained with $L^2_μ$ and $H^1_μ$ objectives and the Fourier neural operator, against polynomial surrogate methods, specifically a reduced-basis sparse-grid surrogate and a reduced-basis tensor-train surrogate. All methods are evaluated on a linear parametric diffusion problem and a nonlinear parametric hyperelasticity problem, using input fields with algebraically decaying spectral coefficients at varying rates of decay $s$. To enable fair comparisons, we analyze ensembles of surrogate models generated by varying hyperparameters and compare the resulting Pareto frontiers of cost versus approximation accuracy, decomposing cost into contributions from data generation, setup, and evaluation. Our results show that no single method is universally superior. Polynomial surrogates achieve substantially better data efficiency for smooth input fields ($s \geq 2$), with convergence rates for the sparse-grid surrogate in agreement with theoretical predictions. For rough inputs ($s \leq 1$), the Fourier neural operator displays the fastest convergence rates. Derivative-informed training consistently improves data efficiency over standard $L^2_μ$ training, providing a competitive alternative for rough inputs in the low-data regime when Jacobian information is available at reasonable cost. These findings highlight the importance of matching the surrogate methodology to the regularity of the problem as well as accuracy demands and computational constraints of the application.

We present OmniVoice, a massive multilingual zero-shot text-to-speech (TTS) model that scales to over 600 languages. At its core is a novel diffusion language model-style discrete non-autoregressive (NAR) architecture. Unlike conventional discrete NAR models that suffer from performance bottlenecks in complex two-stage (text-to-semantic-to-acoustic) pipelines, OmniVoice directly maps text to multi-codebook acoustic tokens. This simplified approach is facilitated by two key technical innovations: (1) a full-codebook random masking strategy for efficient training, and (2) initialization from a pre-trained LLM to ensure superior intelligibility. By leveraging a 581k-hour multilingual dataset curated entirely from open-source data, OmniVoice achieves the broadest language coverage to date and delivers state-of-the-art performance across Chinese, English, and diverse multilingual benchmarks. Our code and pre-trained models are publicly available at https://github.com/k2-fsa/OmniVoice.

Smart contract vulnerabilities can cause substantial financial losses due to the immutability of code after deployment. While existing tools detect vulnerabilities, they cannot effectively repair them. In this paper, we propose SCPatcher, a framework that combines retrieval-augmented generation with a knowledge graph for automated smart contract repair. We construct a knowledge graph from 5,000 verified Ethereum contracts, extracting function-level relationships to build a semantic network. This graph serves as an external knowledge base that enhances Large Language Model reasoning and enables precise vulnerability patching. We introduce a two-stage repair strategy, initial knowledge-guided repair followed by Chain-of-Thought reasoning for complex vulnerabilities. Evaluated on a diverse set of vulnerable contracts, SCPatcher achieves 81.5\% overall repair rate and 91.0\% compilation pass rate, substantially outperforming existing methods.

Successor Features (SF) combined with Generalized Policy Improvement (GPI) provide a robust framework for transfer learning in Reinforcement Learning (RL) by decoupling environment dynamics from reward functions. However, standard SF learning methods typically rely on semi-gradient Temporal Difference (TD) updates. When combined with non-linear function approximation, semi-gradient methods lack robust convergence guarantees and can lead to instability, particularly in the multi-task setting where accurate feature estimation is critical for effective GPI. Inspired by Full Gradient DQN, we propose Full-Gradient Successor Feature Representations Q-Learning (FG-SFRQL), an algorithm that optimizes the successor features by minimizing the full Mean Squared Bellman Error. Unlike standard approaches, our method computes gradients with respect to parameters in both the online and target networks. We provide a theoretical proof of almost-sure convergence for FG-SFRQL and demonstrate empirically that minimizing the full residual leads to superior sample efficiency and transfer performance compared to semi-gradient baselines in both discrete and continuous domains.

Mechanistic simulations typically assume fixed ontologies: variables, causal relationships, and resolution policies are static. This assumption fails when the true causal structure is contested or unidentifiable-as in antimicrobial resistance (AMR) spread, where contact, environmental, and selection ontologies compete. We introduce Procela, a Python framework where variables act as epistemic authorities that maintain complete hypothesis memory, mechanisms encode competing ontologies as causal units, and governance observes epistemic signals and mutates system topology at runtime. This is the first framework where simulations test their own assumptions. We instantiate Procela for AMR in a hospital network with three competing families. Governance detects coverage decay, policy fragility, and runs structural probes. Results show 20.4% error reduction and 69% cumulative regret improvement over baseline. All experiments are reproducible with full auditability. Procela establishes a new paradigm: simulations that model not only the world but their own modeling process, enabling adaptation under structural uncertainty.

TF-IDF is a classical formula that is widely used for identifying important terms within documents. We show that TF-IDF-like scores arise naturally from the test statistic of a penalized likelihood-ratio test setup capturing word burstiness (also known as word over-dispersion). In our framework, the alternative hypothesis captures word burstiness by modeling a collection of documents according to a family of beta-binomial distributions with a gamma penalty term on the precision parameter. In contrast, the null hypothesis assumes that words are binomially distributed in collection documents, a modeling approach that fails to account for word burstiness. We find that a term-weighting scheme given rise to by this test statistic performs comparably to TF-IDF on document classification tasks. This paper provides insights into TF-IDF from a statistical perspective and underscores the potential of hypothesis testing frameworks for advancing term-weighting scheme development.

This study examines the challenges of modeling complex and noisy data related to socioeconomic factors over time, with a focus on data from various districts in Odisha, India. Traditional time-series models struggle to capture both trends and variations together in this type of data. To tackle this, a Variational Neural Stochastic Differential Equation (V-NSDE) model is designed that combines the expressive dynamics of Neural SDEs with the generative capabilities of Variational Autoencoders (VAEs). This model uses an encoder and a decoder. The encoder takes the initial observations and district embeddings and translates them into a Gaussian distribution, which determines the mean and log-variance of the first latent state. Then the obtained latent state initiates the Neural SDE, which utilize neural networks to determine the drift and diffusion functions that govern continuous-time latent dynamics. These governing functions depend on the time index, latent state, and district embedding, which help the model learn the unique characteristics specific to each district. After that, using a probabilistic decoder, the observations are reconstructed from the latent trajectory. The decoder outputs a mean and log-variance for each time step, which follows the Gaussian likelihood. The Evidence Lower Bound (ELBO) training loss improves by adding a KL-divergence regularization term to the negative log-likelihood (nll). The obtained results demonstrate the effective learning of V-NSDE in recognizing complex patterns over time, yielding realistic outcomes that include clear trends and random fluctuations across different areas.

Diffusion language models (DLMs) offer a promising path toward low-latency generation through parallel decoding, but their practical efficiency depends heavily on the decoding trajectory. In practice, this advantage often fails to fully materialize because standard training does not provide explicit supervision over token reveal order, creating a train-inference mismatch that leads to suboptimal decoding behavior. We propose Trajectory-Ranked Instruction Masked Supervision (TRIMS), a simple trajectory-guided supervised fine-tuning framework that injects trajectory supervision into standard Masked Diffusion Language Model (MDLM) training with minimal overhead. Instead of relying on costly DLM-based distillation, TRIMS uses lightweight signals from an autoregressive teacher to guide a trajectory-aware masking strategy, encouraging the model to learn more effective decoding orders. Experiments on LLaDA and Dream across math and coding benchmarks show that TRIMS significantly improves the accuracy-parallelism trade-off over both standard MDLM training and train-free acceleration baselines, while achieving competitive performance with prior distillation-based approaches at substantially lower training cost. Further analysis shows that TRIMS leads to better decoding trajectories, validating the effectiveness of trajectory-guided supervision for DLMs.

Modern data warehouses extend SQL with semantic operators that invoke large language models on each qualifying row, but the per-row inference cost is prohibitive at scale. Model cascades reduce this cost by routing most rows through a fast proxy model and delegating uncertain cases to an expensive oracle. Existing frameworks, however, require global dataset access and optimize a single quality metric, limiting their applicability in distributed systems where data is partitioned across independent workers. We present two adaptive cascade algorithms designed for streaming, per-partition execution in which each worker processes its partition independently without inter-worker communication. SUPG-IT extends the SUPG statistical framework to streaming execution with iterative threshold refinement and joint precision-recall guarantees. GAMCAL replaces user-specified quality targets with a learned calibration model: a Generalized Additive Model maps proxy scores to calibrated probabilities with uncertainty quantification, enabling direct optimization of a cost-quality tradeoff through a single parameter. Experiments on six datasets in a production semantic SQL engine show that both algorithms achieve F1 > 0.95 on every dataset. GAMCAL achieves higher F1 per oracle call at cost-sensitive operating points, while SUPG-IT reaches a higher quality ceiling with formal guarantees on precision and recall.

Smart contracts are self-executing programs that manage financial transactions on blockchain networks. Developers commonly rely on third-party code libraries to improve both efficiency and security. However, improper use of these libraries can introduce hidden vulnerabilities that are difficult to detect, leading to significant financial losses. Existing automated tools struggle to identify such misuse because it often requires understanding the developer's intent rather than simply scanning for known code patterns. This paper presents LibScan, an automated detection framework that combines large language model (LLM)-based semantic reasoning with rule-based code analysis, identifying eight distinct categories of library misuse in smart contracts. To improve detection reliability, the framework incorporates an iterative self-correction mechanism that refines its analysis across multiple rounds, alongside a structured knowledge base derived from large-scale empirical studies of real-world misuse cases. Experiments conducted on 662 real-world smart contracts demonstrate that LibScan achieves an overall detection accuracy of 85.15\%, outperforming existing tools by a margin of over 16 percentage points. Ablation experiments further confirm that combining both analysis approaches yields substantially better results than either method used independently.

Predictive process monitoring (PPM) focuses on predicting future process trajectories, including next activity predictions. This is crucial in dynamic environments where processes change or face uncertainty. However, current frameworks often assume a static environment, overlooking dynamic characteristics and concept drifts. This results in catastrophic forgetting, where training while focusing merely on new data distribution negatively impacts the performance on previously learned data distributions. Continual learning addresses, among others, the challenges related to mitigating catastrophic forgetting. This paper proposes a novel approach called Continual Next Activity Prediction with Prompts (CNAPwP), which adapts the DualPrompt algorithm for next activity prediction to improve accuracy and adaptability while mitigating catastrophic forgetting. We introduce new datasets with recurring concept drifts, alongside a task-specific forgetting metric that measures the prediction accuracy gap between initial occurrence and subsequent task occurrences. Extensive testing on three synthetic and two real-world datasets representing several setups of recurrent drifts shows that CNAPwP achieves SOTA or competitive results compared to five baselines, demonstrating its potential applicability in real-world scenarios. An open-source implementation of our method, together with the datasets and results, is available at: https://github.com/SvStraten/CNAPwP.

We propose a novel extension of the Bradley-Terry model to multiplayer games and adapt a recent algorithm by Newman [1] to our model. We demonstrate the use of our proposed method on synthetic datasets and on a real dataset of games of cards.

Poverty is a complex dynamic challenge that cannot be adequately captured using predefined differential equations. Nowadays, artificial machine learning (ML) methods have demonstrated significant potential in modelling real-world dynamical systems. Among these, Neural Ordinary Differential Equations (Neural ODEs) have emerged as a powerful, data-driven approach for learning continuous-time dynamics directly from observations. This chapter applies the Neural ODE framework to analyze poverty dynamics in the Indian state of Odisha. Specifically, we utilize time-series data from 2007 to 2020 on the key indicators of economic development and poverty reduction. Within the Neural ODE architecture, the temporal gradient of the system is represented by a multi-layer perceptron (MLP). The obtained neural dynamical system is integrated using a numerical ODE solver to obtain the trajectory of over time. In backpropagation, the adjoint sensitivity method is utilized for gradient computation during training to facilitate effective backpropagation through the ODE solver. The trained Neural ODE model reproduces the observed data with high accuracy. This demonstrates the capability of Neural ODE to capture the dynamics of the poverty indicator of concrete-structured households. The obtained results show that ML methods, such as Neural ODEs, can serve as effective tools for modeling socioeconomic transitions. It can provide policymakers with reliable projections, supporting more informed and effective decision-making for poverty alleviation.

Knowledge distillation has become a primary mechanism for transferring reasoning and domain expertise from frontier Large Language Models (LLMs) to smaller, deployable students. However, the dominant paradigm remains \textit{off-policy}: students train on static teacher-generated data and never encounter their own errors during learning. This train--test mismatch, an instance of \textit{exposure bias}, causes prediction errors to compound autoregressively at inference time. On-Policy Distillation (OPD) addresses this by letting the student generate its own trajectories and receive teacher feedback on these self-generated outputs, grounding distillation in the theory of interactive imitation learning. Despite rapid growth spanning divergence minimization, reward-guided learning, and self-play, the OPD literature remains fragmented with no unified treatment. This survey provides the first comprehensive overview of OPD for LLMs. We introduce a unified $f$-divergence framework over on-policy samples and organize the landscape along three orthogonal dimensions: \emph{feedback signal} (logit-based, outcome-based, or self-play), \emph{teacher access} (white-box, black-box, or teacher-free), and \emph{loss granularity} (token-level, sequence-level, or hybrid). We systematically analyze representative methods, examine industrial deployments, and identify open problems including distillation scaling laws, uncertainty-aware feedback, and agent-level distillation.

We present KUTED, a speech-to-text translation (S2TT) dataset for Central Kurdish, derived from TED and TEDx talks. The corpus comprises 91,000 sentence pairs, including 170 hours of English audio, 1.65 million English tokens, and 1.40 million Central Kurdish tokens. We evaluate KUTED on the S2TT task and find that orthographic variation significantly degrades Kurdish translation performance, producing nonstandard outputs. To address this, we propose a systematic text standardization approach that yields substantial performance gains and more consistent translations. On a test set separated from TED talks, a fine-tuned Seamless model achieves 15.18 BLEU, and we improve Seamless baseline by 3.0 BLEU on the FLEURS benchmark. We also train a Transformer model from scratch and evaluate a cascaded system that combines Seamless (ASR) with NLLB (MT).

Despite extensions to speech inputs, effectively leveraging the rich knowledge and contextual understanding of large language models (LLMs) in automatic speech recognition (ASR) remains non-trivial, as the task primarily involves direct speech-to-text mapping. To address this, this paper proposes chain-of-thought ASR (CoT-ASR), which constructs a reasoning chain that enables LLMs to first analyze the input speech and generate contextual analysis, thereby fully exploiting their generative capabilities. With this contextual reasoning, CoT-ASR then performs more informed speech recognition and completes both reasoning and transcription in a single pass. Moreover, CoT-ASR naturally supports user-guided transcription: while designed to self-generate reasoning, it can also seamlessly incorporate user-provided context to guide transcription, further extending ASR functionality. To reduce the modality gap, this paper introduces a CTC-guided Modality Adapter, which uses CTC non-blank token probabilities to weight LLM embeddings, efficiently aligning speech encoder outputs with the LLM's textual latent space. Experiments show that, compared to standard LLM-based ASR, CoT-ASR achieves a relative reduction of 8.7% in word error rate (WER) and 16.9% in entity error rate (EER).

The classical-quantum system heterogeneity (different data characteristics, execution paradigms and synchronization mechanism etc.) renders existing distributed communication mechanisms (e.g. MPI, NCCL etc.) inadequate. This bottleneck severely impairs operational synergy and programming efficiency. Thus, the performance of hybrid applications on classical-quantum heterogeneous infrastructures is directly limited. To address these challenges, this paper proposes a message-passing library tailored for large-scale classical-quantum heterogeneous distributed computing, referred to as MPI-Q. The design centers on three mechanisms. First, it defines a heterogeneous hybrid communication domain that achieves unified management of classical and quantum processes in heterogeneous hybrid systems. Second, it uses a lightweight communication path that allows classical control nodes to send device-ready waveform data directly to quantum MonitorProcesses, avoiding unnecessary relay stages. Third, it establishes a heterogeneous hybrid synchronization mechanism to tackle the problem of timing control for multi-node quantum operations. While retaining the traditional MPI programming model, MPI-Q achieves extension toward quantum subsystems. Experiments on distributed GHZ state preparation demonstrate that this model exhibits near-linear scalability, achieving a maximum speedup of 18.76 times on 24 quantum nodes. This proves that the library can effectively support large-scale heterogeneous hybrid distributed computing applications, filling the technical gap in this field.

Predicting the behavior of ultra-large complex systems, from climate to biological and technological networks, is a central unsolved challenge. Existing approaches face a fundamental trade-off: equation discovery methods provide interpretability but fail to scale, while neural networks scale but operate as black boxes and often lose reliability over long times. Here, we introduce the Sparse Identification Graph Neural Network, a framework that overcome this divide by allowing to infer the governing equations of large networked systems from data. By defining symbolic discovery as edge-level information, SIGN decouples the scalability of sparse identification from network size, enabling efficient equation discovery even in large systems. SIGN allows to study networks with over 100,000 nodes while remaining robust to noise, sparse sampling, and missing data. Across diverse benchmark systems, including coupled chaotic oscillators, neural dynamics, and epidemic spreading, it recovers governing equations with high precision and sustains accurate long-term predictions. Applied to a data set of time series of temperature measurements in 71,987 sea surface positions, SIGN identifies a compact predictive network model and captures large-scale sea surface temperature conditions up to two years in advance. By enabling equation discovery at previously inaccessible scales, SIGN opens a path toward interpretable and reliable prediction of real-world complex systems.

As the focus in LLM-based coding shifts from static single-step code generation to multi-step agentic interaction with tools and environments, understanding which tasks will challenge agents and why becomes increasingly difficult. This is compounded by current practice: agent performance is typically measured by aggregate pass rates on benchmarks, but single-number metrics obscure the diversity of tasks within a benchmark. We present a framework for predicting success or failure on individual tasks tailored to the agentic coding regime. Our approach augments Item Response Theory (IRT) with rich features extracted from tasks, including issue statements, repository contexts, solutions, and test cases, and introduces a novel decomposition of agent ability into LLM and scaffold ability components. This parameterization enables us to aggregate evaluation data across heterogeneous leaderboards and accurately predict task-level performance for unseen benchmarks, as well as unseen LLM-scaffold combinations. Our methods have practical utility for benchmark designers, who can better calibrate the difficulty of their new tasks without running computationally expensive agent evaluations.

In recent years, the scaling laws of recommendation models have attracted increasing attention, which govern the relationship between performance and parameters/FLOPs of recommenders. Currently, there are three mainstream architectures for achieving scaling in recommendation models, namely attention-based, TokenMixer-based, and factorization-machine-based methods, which exhibit fundamental differences in both design philosophy and architectural structure. In this paper, we propose a unified scaling architecture for recommendation systems, namely \textbf{UniMixer}, to improve scaling efficiency and establish a unified theoretical framework that unifies the mainstream scaling blocks. By transforming the rule-based TokenMixer to an equivalent parameterized structure, we construct a generalized parameterized feature mixing module that allows the token mixing patterns to be optimized and learned during model training. Meanwhile, the generalized parameterized token mixing removes the constraint in TokenMixer that requires the number of heads to be equal to the number of tokens. Furthermore, we establish a unified scaling module design framework for recommender systems, which bridges the connections among attention-based, TokenMixer-based, and factorization-machine-based methods. To further boost scaling ROI, a lightweight UniMixing module is designed, \textbf{UniMixing-Lite}, which further compresses the model parameters and computational cost while significantly improve the model performance. The scaling curves are shown in the following figure. Extensive offline and online experiments are conducted to verify the superior scaling abilities of \textbf{UniMixer}.

As Large Language Model (LLM) capabilities advance, the demand for high-quality annotation of exponentially increasing text corpora has outpaced human capacity, leading to the widespread adoption of LLMs in automatic evaluation and annotation. However, proprietary LLMs often exhibit systematic biases that diverge from human expert consensus, lacks reproducibility, and raises data privacy concerns. Our work examines the viability of finetuning a quantized Small Language Model of 1.7B parameter size on limited human-annotated data to serve as a highly aligned, deterministic evaluator and annotator. By implementing a custom, multi-dimensional rubric framework and simple augmentation and regularization techniques, the proposed approach achieves higher inter-annotator agreement (0.23 points increase in Krippendorff's $α$) than the best performing state-of-the-art proprietary LLM. We also demonstrate the generalizability of the proposed training pipeline on a separate emotion classification task. The results show that task-specific alignment and efficient 4-bit quantized fine-tuning provide superior open-source alternative to using proprietary models for evaluation and annotation. Our finetuning approach is publicly available at https://github.com/jylee-k/slm-judge.

Molecular dynamics simulations provide detailed trajectories at the atomic level, but extracting interpretable and robust insights from these high-dimensional data remains challenging. In practice, analyses typically rely on a single representation. Here, we show that representation choice is not neutral: it fundamentally shapes the conformational organization, similarity relationships, and apparent transitions inferred from identical simulation data. To complement existing representations, we introduce Orientation features, a geometrically grounded, rotation-aware encoding of protein backbone. We compare it against common descriptions across three dynamical regimes: fast-folding proteins, large-scale domain motions, and protein-protein association. Across these systems, we find that different representations emphasize complementary aspects of conformational space, and that no single representation provides a complete picture of the underlying dynamics. To facilitate systematic comparison, we developed ManiProt, a library for efficient computation and analysis of multiple protein representations. Our results motivate a comparative, representation-aware framework for the interpretation of molecular dynamics simulations.

In enhancing the fairness of Large Language Models (LLMs), evaluating social biases rooted in the cultural contexts of specific linguistic regions is essential. However, most existing Japanese benchmarks heavily rely on translating English data, which does not necessarily provide an evaluation suitable for Japanese culture. Furthermore, they only evaluate bias in the conclusion, failing to capture biases lurking in the reasoning. In this study, based on attribution theory in social psychology, we constructed a new dataset, ``JUBAKU-v2,'' which evaluates the bias in attributing behaviors to in-groups and out-groups within reasoning while fixing the conclusion. This dataset consists of 216 examples reflecting cultural biases specific to Japan. Experimental results verified that it can detect performance differences across models more sensitively than existing benchmarks.

The impending arrival of cryptographically relevant quantum computers (CRQCs) threatens the security foundations of modern software: Shor's algorithm breaks RSA, ECDSA, ECDH, and Diffie-Hellman, while Grover's algorithm reduces the effective security of symmetric and hash-based schemes. Despite NIST standardising post-quantum cryptography (PQC) in 2024 (FIPS 203 ML-KEM, FIPS 204 ML-DSA, FIPS 205 SLH-DSA), most codebases lack automated tooling to inventory classical cryptographic usage and prioritise migration based on quantum risk. We present Quantum-Safe Code Auditor, a quantum-aware static analysis framework that combines (i) regex-based detection of 15 classes of quantum-vulnerable primitives, (ii) LLM-assisted contextual enrichment to classify usage and severity, and (iii) risk scoring via a Variational Quantum Eigensolver (VQE) model implemented in Qiskit 2.x, incorporating qubit-cost estimates to prioritise findings. We evaluate the system across five open-source libraries -- python-rsa, python-ecdsa, python-jose, node-jsonwebtoken, and Bouncy Castle Java -- covering 5,775 findings. On a stratified sample of 602 labelled instances, we achieve 71.98% precision, 100% recall, and an F1 score of 83.71%. All code, data, and reproduction scripts are released as open-source.

The generalization ability of imitation learning policies for robotic manipulation is fundamentally constrained by the diversity of expert demonstrations, while collecting demonstrations across varied environments is costly and difficult in practice. In this paper, we propose a practical framework that exploits inherent scene diversity without additional human effort by scaling camera views during demonstration collection. Instead of acquiring more trajectories, multiple synchronized camera perspectives are used to generate pseudo-demonstrations from each expert trajectory, which enriches the training distribution and improves viewpoint invariance in visual representations. We analyze how different action spaces interact with view scaling and show that camera-space representations further enhance diversity. In addition, we introduce a multiview action aggregation method that allows single-view policies to benefit from multiple cameras during deployment. Extensive experiments in simulation and real-world manipulation tasks demonstrate significant gains in data efficiency and generalization compared to single-view baselines. Our results suggest that scaling camera views provides a practical and scalable solution for imitation learning, which requires minimal additional hardware setup and integrates seamlessly with existing imitation learning algorithms. The website of our project is https://yichen928.github.io/robot_multiview.

Housing selection is a high-stakes and largely irreversible decision problem. We study housing consultation as a decision-support interface for housing selection. Existing housing platforms and many LLM-based assistants often reduce this process to ranking or recommendation, resulting in opaque reasoning, brittle multi-constraint handling, and limited guarantees on factuality. We present HabitatAgent, the first LLM-powered multi-agent architecture for end-to-end housing consultation. HabitatAgent comprises four specialized agent roles: Memory, Retrieval, Generation, and Validation. The Memory Agent maintains multi-layer user memory through internal stages for constraint extraction, memory fusion, and verification-gated updates; the Retrieval Agent performs hybrid vector--graph retrieval (GraphRAG); the Generation Agent produces evidence-referenced recommendations and explanations; and the Validation Agent applies multi-tier verification and targeted remediation. Together, these agents provide an auditable and reliable workflow for end-to-end housing consultation. We evaluate HabitatAgent on 100 real user consultation scenarios (300 multi-turn question--answer pairs) under an end-to-end correctness protocol. A strong single-stage baseline (Dense+Rerank) achieves 75% accuracy, while HabitatAgent reaches 95%.

Enterprise adoption of Large Language Models (LLMs) is constrained by hallucination, domain drift, and the inability to enforce regulatory compliance at the reasoning level. We present a neurosymbolic architecture implemented within the Foundation AgenticOS (FAOS) platform that addresses these limitations through ontology-constrained neural reasoning. Our approach introduces a three-layer ontological framework--Role, Domain, and Interaction ontologies--that provides formal semantic grounding for LLM-based enterprise agents. We formalize the concept of asymmetric neurosymbolic coupling, wherein symbolic ontological knowledge constrains agent inputs (context assembly, tool discovery, governance thresholds) while proposing mechanisms for extending this coupling to constrain agent outputs (response validation, reasoning verification, compliance checking). We evaluate the architecture through a controlled experiment (600 runs across five industries: FinTech, Insurance, Healthcare, Vietnamese Banking, and Vietnamese Insurance), finding that ontology-coupled agents significantly outperform ungrounded agents on Metric Accuracy (p < .001, W = .460), Regulatory Compliance (p = .003, W = .318), and Role Consistency (p < .001, W = .614), with improvements greatest where LLM parametric knowledge is weakest--particularly in Vietnam-localized domains. Our contributions include: (1) a formal three-layer enterprise ontology model, (2) a taxonomy of neurosymbolic coupling patterns, (3) ontology-constrained tool discovery via SQL-pushdown scoring, (4) a proposed framework for output-side ontological validation, (5) empirical evidence for the inverse parametric knowledge effect that ontological grounding value is inversely proportional to LLM training data coverage of the domain, and (6) a production system serving 21 industry verticals with 650+ agents.

The scenario approach provides a powerful data-driven framework for designing solutions under uncertainty with rigorous probabilistic robustness guarantees. Existing theory, however, primarily addresses assessing robustness with respect to a single appropriateness criterion for the solution based on a dataset, whereas many practical applications - including multi-agent decision problems - require the simultaneous consideration of multiple criteria and the assessment of their robustness based on multiple datasets, one per criterion. This paper develops a general scenario theory for multi-criteria data-driven decision making. A central innovation lies in the collective treatment of the risks associated with violations of individual criteria, which yields substantially more accurate robustness certificates than those derived from a naive application of standard results. In turn, this approach enables a sharper quantification of the robustness level with which all criteria are simultaneously satisfied. The proposed framework applies broadly to multi-criteria data-driven decision problems, providing a principled, scalable, and theoretically grounded methodology for design under uncertainty.

The integration of Large Language Models (LLMs) into life sciences has catalyzed the development of "AI Scientists." However, translating these theoretical capabilities into deployment-ready research environments exposes profound infrastructural vulnerabilities. Current frameworks are bottlenecked by fragile JSON-based tool-calling protocols, easily disrupted execution sandboxes that lose graphical outputs, and rigid conversational interfaces inherently ill-suited for high-dimensional scientific data.We introduce BloClaw, a unified, multi-modal operating system designed for Artificial Intelligence for Science (AI4S). BloClaw reconstructs the Agent-Computer Interaction (ACI) paradigm through three architectural innovations: (1) An XML-Regex Dual-Track Routing Protocol that statistically eliminates serialization failures (0.2% error rate vs. 17.6% in JSON); (2) A Runtime State Interception Sandbox that utilizes Python monkey-patching to autonomously capture and compile dynamic data visualizations (Plotly/Matplotlib), circumventing browser CORS policies; and (3) A State-Driven Dynamic Viewport UI that morphs seamlessly between a minimalist command deck and an interactive spatial rendering engine. We comprehensively benchmark BloClaw across cheminformatics (RDKit), de novo 3D protein folding via ESMFold, molecular docking, and autonomous Retrieval-Augmented Generation (RAG), establishing a highly robust, self-evolving paradigm for computational research assistants. The open-source repository is available at https://github.com/qinheming/BloClaw.

Unified Multimodal Large Models (UMLMs) integrate understanding and generation capabilities within a single architecture. While this architectural unification, driven by the deep fusion of multimodal features, enhances model performance, it also introduces important yet underexplored safety challenges. Existing safety benchmarks predominantly focus on isolated understanding or generation tasks, failing to evaluate the holistic safety of UMLMs when handling diverse tasks under a unified framework. To address this, we introduce Uni-SafeBench, a comprehensive benchmark featuring a taxonomy of six major safety categories across seven task types. To ensure rigorous assessment, we develop Uni-Judger, a framework that effectively decouples contextual safety from intrinsic safety. Based on comprehensive evaluations across Uni-SafeBench, we uncover that while the unification process enhances model capabilities, it significantly degrades the inherent safety of the underlying LLM. Furthermore, open-source UMLMs exhibit much lower safety performance than multimodal large models specialized for either generation or understanding tasks. We open-source all resources to systematically expose these risks and foster safer AGI development.

We prove that activation saturation imposes a structural dynamical limitation on autonomous Neural ODEs $\dot{h}=f_θ(h)$ with saturating activations ($\tanh$, sigmoid, etc.): if $q$ hidden layers of the MLP $f_θ$ satisfy $|σ'|\leδ$ on a region~$U$, the input Jacobian is attenuated as $\norm{Df_θ(x)}\le C(U)$ (for activations with $\sup_{x}|σ'(x)|\le 1$, e.g.\ $\tanh$ and sigmoid, this reduces to $C_Wδ^q$), forcing every Floquet (Lyapunov) exponen along any $T$-periodic orbit $γ\subset U$ into the interval $[-C(U),\;C(U)]$. This is a collapse of the Floquet spectrum: as saturation deepens ($δ\to 0$), all exponents are driven to zero, limiting both strong contraction and chaotic sensitivity. The obstruction is structural -- it constrains the learned vector field at inference time, independent of training quality. As a secondary contribution, for activations with $σ'>0$, a saturation-weighted spectral factorisation yields a refined bound $\widetilde{C}(U)\le C(U)$ whose improvement is amplified exponentially in~$T$ at the flow level. All results are numerically illustrated on the Stuart--Landau oscillator; the bounds provide a theoretical explanation for the empirically observed failure of $\tanh$-NODEs on the Morris--Lecar neuron model.

Dental diagnosis from Orthopantomograms (OPGs) requires coordination of tooth detection, caries segmentation (CarSeg), anomaly detection (AD), and dental developmental staging (DDS). We propose Mamba-based Architectural Tooth Hierarchical Estimator and Holistic Evaluation Network for Anatomy (MATHENA), a unified framework leveraging Mamba's linear-complexity State Space Models (SSM) to address all four tasks. MATHENA integrates MATHE, a multi-resolution SSM-driven detector with four-directional Vision State Space (VSS) blocks for O(N) global context modeling, generating per-tooth crops. These crops are processed by HENA, a lightweight Mamba-UNet with a triple-head architecture and Global Context State Token (GCST). In the triple-head architecture, CarSeg is first trained as an upstream task to establish shared representations, which are then frozen and reused for downstream AD fine-tuning and DDS classification via linear probing, enabling stable, efficient learning. We also curate PARTHENON, a benchmark comprising 15,062 annotated instances from ten datasets. MATHENA achieves 93.78% mAP@50 in tooth detection, 90.11% Dice for CarSeg, 88.35% for AD, and 72.40% ACC for DDS.

Large language models (LLMs) achieve strong downstream performance largely due to abundant supervised fine-tuning (SFT) data. However, high-quality SFT data in knowledge-intensive domains such as humanities, social sciences, medicine, law, and finance is scarce because expert curation is expensive, privacy constraints are strict, and label consistency is hard to ensure. Recent work uses synthetic data, typically by prompting a generator over domain documents and filtering outputs with handcrafted rubrics. Yet rubric design is expert-dependent, transfers poorly across domains, and is often optimized through a brittle heuristic loop of writing rubrics, synthesizing data, training, inspecting results, and manually guessing revisions. This process lacks reliable quantitative feedback about how a rubric affects downstream performance. We propose evaluating synthetic data by its training utility on the target model and using this signal to guide data generation. Inspired by influence estimation, we adopt an optimizer-aware estimator that uses gradient information to quantify each synthetic sample's contribution to a target model's objective on specific tasks. Our analysis shows that even when synthetic and real samples are close in embedding space, their influence on learning can differ substantially. Based on this insight, we propose an optimization-based framework that adapts rubrics using target-model feedback. We provide lightweight guiding text and use a rubric-specialized model to generate task-conditioned rubrics. Influence score is used as the reward to optimize the rubric generator with reinforcement learning. Experiments across domains, target models, and data generators show consistent improvements and strong generalization without task-specific tuning.

Large Language Models (LLMs) generalize across tasks via reusable representations and flexible reasoning, yet remain brittle in real deployment under evolving tasks and continual distribution shift. A common approach is Test-Time Adaptation (TTA), existing ones of which updates models with hand-designed unsupervised objectives over the full parameter space and mostly overlook preserving shared source knowledge and the reliability of adaptation signals. Drawing on molecular signaling cascades of memory updating in Drosophila, we propose Synapse Consolidation (SyCo), a parameter-efficient LLM adaptation method that updates low-rank adapters through Rac1 and MAPK pathways under the guidance of a structured TTA objective driven by problem understanding, process understanding, and source-domain guardrail. Rac1 confines plasticity to a tail-gradient subspace that is less critical for source knowledge, enabling rapid specialization while preserving source representations. MAPK uses a tiered controller to suppress noisy updates and consolidate useful adaptations under non-stationary streams. To model real deployments with multiple sources and continually emerging tasks, we introduce Multi-source Open-set Adaptation (MOA) setting, where a model is trained on multiple labeled source tasks and then adapts on open, non-stationary unlabeled test streams that mix seen and unseen tasks with partial overlap in label and intent space. Across 18 NLP datasets and the MOA setting, SyCo consistently outperforms strong baselines, achieving 78.31\% on unseen-task adaptation and 85.37\% on unseen-data shifts.

Multi-task representation learning is an approach that learns shared latent representations across related tasks, facilitating knowledge transfer and improving sample efficiency. This paper introduces a novel approach to multi-task representation learning in linear bandits. We consider a setting with T concurrent linear bandit tasks, each with feature dimension d, that share a common latent representation of dimension r \ll min{d,T}$, capturing their underlying relatedness. We propose a new Optimism in the Face of Uncertainty Linear (OFUL) algorithm that leverages shared low-rank representations to enhance decision-making in a sample-efficient manner. Our algorithm first collects data through an exploration phase, estimates the shared model via spectral initialization, and then conducts OFUL based learning over a newly constructed confidence set. We provide theoretical guarantees for the confidence set and prove that the unknown reward vectors lie within the confidence set with high probability. We derive cumulative regret bounds and show that the proposed approach achieves \tilde{O}(\sqrt{drNT}), a significant improvement over solving the T tasks independently, resulting in a regret of \tilde{O}(dT\sqrt{N}). We performed numerical simulations to validate the performance of our algorithm for different problem sizes.

Quantization-aware training (QAT) is typically performed for a single target numeric format, while practical deployments often need to choose numerical precision at inference time based on hardware support or runtime constraints. We study multi-format QAT, where a single model is trained to be robust across multiple quantization formats. We find that multi-format QAT can match single-format QAT at each target precision, yielding one model that performs well overall across different formats, even formats that were not seen during training. To enable practical deployment, we propose the Slice-and-Scale conversion procedure for both MXINT and MXFP that converts a high-precision representation into lower-precision formats without re-training. Building on this, we introduce a pipeline that (i) trains a model with multi-format QAT, (ii) stores a single anchor format checkpoint (MXINT8/MXFP8), and (iii) allows on-the-fly conversion to lower MXINT or MXFP formats at runtime with negligible-or no-additional accuracy degradation. Together, these components provide a practical path to elastic precision scaling and allow selecting the runtime format at inference time across diverse deployment targets.

3D Visual Grounding (3D-VG) aims to localize objects in 3D scenes via natural language descriptions. While recent advancements leveraging Vision-Language Models (VLMs) have explored zero-shot possibilities, they typically suffer from a static workflow relying on preprocessed 3D point clouds, essentially degrading grounding into proposal matching. To bypass this reliance, our core motivation is to decouple the task: leveraging 2D VLMs to resolve complex spatial semantics, while relying on deterministic multi-view geometry to instantiate the 3D structure. Driven by this insight, we propose "Think, Act, Build (TAB)", a dynamic agentic framework that reformulates 3D-VG tasks as a generative 2D-to-3D reconstruction paradigm operating directly on raw RGB-D streams. Specifically, guided by a specialized 3D-VG skill, our VLM agent dynamically invokes visual tools to track and reconstruct the target across 2D frames. Crucially, to overcome the multi-view coverage deficit caused by strict VLM semantic tracking, we introduce the Semantic-Anchored Geometric Expansion, a mechanism that first anchors the target in a reference video clip and then leverages multi-view geometry to propagate its spatial location across unobserved frames. This enables the agent to "Build" the target's 3D representation by aggregating these multi-view features via camera parameters, directly mapping 2D visual cues to 3D coordinates. Furthermore, to ensure rigorous assessment, we identify flaws such as reference ambiguity and category errors in existing benchmarks and manually refine the incorrect queries. Extensive experiments on ScanRefer and Nr3D demonstrate that our framework, relying entirely on open-source models, significantly outperforms previous zero-shot methods and even surpasses fully supervised baselines.

We study for the first time, stochastic dueling bandits over continuous action spaces with Lipschitz structure, where feedback is purely comparative. While dueling bandits and Lipschitz bandits have been studied separately, their combination has remained unexplored. We propose the first algorithm for Lipschitz dueling bandits, using round-based exploration and recursive region elimination guided by an adaptive reference arm. We develop new analytical tools for relative feedback and prove a regret bound of $\tilde O\left(T^{\frac{d_z+1}{d_z+2}}\right)$, where $d_z$ is the zooming dimension of the near-optimal region. Further, our algorithm takes only logarithmic space in terms of the total time horizon, best achievable by any bandit algorithm over a continuous action space.

With the rapid advancements in deep learning techniques, wearable sensor-aided animal activity recognition (AAR) has demonstrated promising performance, thereby improving livestock management efficiency as well as animal health and welfare monitoring. However, existing research often prioritizes overall performance, overlooking the fact that classification accuracies for specific animal behavioral categories may remain unsatisfactory. This issue typically stems from suboptimal sampling rates or class imbalance problems. To address these challenges and achieve high classification accuracy across all individual behaviors in farm animals, we propose a novel Individual-Behavior-Aware Network (IBA-Net). This network enhances the recognition of each specific behavior by simultaneously customizing features and calibrating the classifier. Specifically, considering that different behaviors require varying sampling rates to achieve optimal performance, we design a Mixture-of-Experts (MoE)-based Feature Customization (MFC) module. This module adaptively fuses data from multiple sampling rates, capturing customized features tailored to various animal behaviors. Additionally, to mitigate classifier bias toward majority classes caused by class imbalance, we develop a Neural Collapse-driven Classifier Calibration (NC3) module. This module introduces a fixed equiangular tight frame (ETF) classifier during the classification stage, maximizing the angles between pair-wise classifier vectors and thereby improving the classification performance for minority classes. To validate the effectiveness of IBA-Net, we conducted experiments on three public datasets covering goat, cattle, and horse activity recognition. The results demonstrate that our method consistently outperforms existing approaches across all datasets.

Training deep learning models for three-dimensional (3D) medical imaging, such as Computed Tomography (CT), is fundamentally challenged by the scarcity of labeled data. While pre-training on natural images is common, it results in a significant domain shift, limiting performance. Self-Supervised Learning (SSL) on unlabeled medical data has emerged as a powerful solution, but prominent frameworks often fail to exploit the inherent 3D nature of CT scans. These methods typically process 3D scans as a collection of independent 2D slices, an approach that fundamentally discards critical axial coherence and the 3D structural context. To address this limitation, we propose the autoencoder for enhanced self-supervised medical image learning(MAESIL), a novel self-supervised learning framework designed to capture 3D structural information efficiently. The core innovation is the 'superpatch', a 3D chunk-based input unit that balances 3D context preservation with computational efficiency. Our framework partitions the volume into superpatches and employs a 3D masked autoencoder strategy with a dual-masking strategy to learn comprehensive spatial representations. We validated our approach on three diverse large-scale public CT datasets. Our experimental results show that MAESIL demonstrates significant improvements over existing methods such as AE, VAE and VQ-VAE in key reconstruction metrics such as PSNR and SSIM. This establishes MAESIL as a robust and practical pre-training solution for 3D medical imaging tasks.

With the rapid growth of e-commerce, exploring general representations rather than task-specific ones has attracted increasing attention. Although recent multimodal large language models (MLLMs) have driven significant progress in product understanding, they are typically employed as feature extractors that implicitly encode product information into global embeddings, thereby limiting their ability to capture fine-grained attributes. Therefore, we argue that leveraging the reasoning capabilities of MLLMs to explicitly model fine-grained product attributes holds significant potential. Nevertheless, achieving this goal remains non-trivial due to several key challenges: (i) long-context reasoning tends to dilute the model's attention to salient information in the raw input; (ii) supervised fine-tuning (SFT) primarily encourages rigid imitation, limiting the exploration of effective reasoning strategies; and (iii) fine-grained details are progressively attenuated during forward propagation. To address these issues, we propose MOON3.0, the first reasoning-aware MLLM-based model for product representation learning. Our method (1) employs a multi-head modality fusion module to adaptively integrate raw signals; (2) incorporates a joint contrastive and reinforcement learning framework to autonomously explore more effective reasoning strategies; and (3) introduces a fine-grained residual enhancement module to progressively preserve local details throughout the network. Additionally, we release a large-scale multimodal e-commerce benchmark MBE3.0. Experimentally, our model demonstrates state-of-the-art zero-shot performance across various downstream tasks on both our benchmark and public datasets.

Monte Carlo Tree Search (MCTS) is an effective test-time compute scaling (TTCS) method for improving the reasoning performance of large language models, but its highly variable execution time leads to severe long-tail latency in practice. Existing optimizations such as positive early exit, reduce latency in favorable cases but are less effective when search continues without meaningful progress. We introduce {\it negative early exit}, which prunes unproductive MCTS trajectories, and an {\it adaptive boosting mechanism} that reallocates reclaimed computation to reduce resource contention among concurrent searches. Integrated into vLLM, these techniques substantially reduce p99 end-to-end latency while improving throughput and maintaining reasoning accuracy.

Dynamic multi-objective optimization requires continuous tracking of moving Pareto fronts. Existing methods struggle with irregular mutations and data sparsity, primarily facing three challenges: the non-linear coupling of dynamic modes, negative transfer from outdated historical data, and the cold-start problem during environmental switches. To address these issues, this paper proposes a decoupled basis-vector-driven generative framework (DB-GEN). First, to resolve non-linear coupling, the framework employs the discrete wavelet transform to separate evolutionary trajectories into low-frequency trends and high-frequency details. Second, to mitigate negative transfer, it learns transferable basis vectors via sparse dictionary learning rather than directly memorizing historical instances. Recomposing these bases under a topology-aware contrastive constraint constructs a structured latent manifold. Finally, to overcome the cold-start problem, a surrogate-assisted search paradigm samples initial populations from this manifold. Pre-trained on 120 million solutions, DB-GEN performs direct online inference without retraining or fine-tuning. This zero-shot generation process executes in milliseconds, requiring approximately 0.2 seconds per environmental change. Experimental results demonstrate that DB-GEN improves tracking accuracy across various dynamic benchmarks compared to existing algorithms.

Overparameterized neural networks often show a benign overfitting property in the sense of achieving excellent generalization behavior despite the number of parameters exceeding the number of training examples. A promising direction to explain benign overfitting is to relate generalization to the norm of distance from initialization, motivated by the empirical observations that this distance is often significantly smaller than the norm itself. However, the existing initialization-dependent complexity analyses cannot fully exploit the power of initialization since the associated bounds depend on the spectral norm of the initialization matrix, which can scale as a square-root function of the width and are therefore not effective for overparameterized models. In this paper, we develop the first \emph{fully} initialization-dependent complexity bounds for shallow neural networks with general Lipschitz activation functions, which enjoys a logarithmic dependency on the width. Our bounds depend on the path-norm of the distance from initialization, which are derived by introducing a new peeling technique to handle the challenge along with the initialization-dependent constraint. We also develop a lower bound tight up to a constant factor. Finally, we conduct empirical comparisons and show that our generalization analysis implies non-vacuous bounds for overparameterized networks.

The heavy-tailed mutation operator, proposed by Doerr, Le, Makhmara, and Nguyen (2017) for evolutionary algorithms, is based on the power-law assumption of mutation rate distribution. Here we generalize the power-law assumption using a regularly varying constraint on the distribution function of mutation rate. In this setting, we generalize the upper bounds on the expected optimization time of the $(1+(λ,λ))$ genetic algorithm obtained by Antipov, Buzdalov and Doerr (2022) for the OneMax function class parametrized by the problem dimension $n$. In particular, it is shown that, on this function class, the sufficient conditions of Antipov, Buzdalov and Doerr (2022) on the heavy-tailed mutation, ensuring the $O(n)$ optimization time in expectation, may be generalized as well. This optimization time is known to be asymptotically smaller than what can be achieved by the $(1+(λ,λ))$ genetic algorithm with any static mutation rate. A new version of the heavy-tailed mutation operator is proposed, satisfying the generalized conditions, and promising results of computational experiments are presented.

To schedule LLM inference, the \textit{shortest job first} (SJF) principle is favorable by prioritizing requests with short output lengths to avoid head-of-line (HOL) blocking. Existing methods usually predict a single output length for each request to facilitate scheduling. We argue that such a \textit{point estimate} does not match the \textit{stochastic} decoding process of LLM inference, where output length is \textit{uncertain} by nature and determined by when the end-of-sequence (EOS) token is sampled. Hence, the output length of each request should be fitted with a distribution rather than a single value. With an in-depth analysis of empirical data and the stochastic decoding process, we observe that output length follows a heavy-tailed distribution and can be fitted with the log-t distribution. On this basis, we propose a simple metric called Tail Inflated Expectation (TIE) to replace the output length in SJF scheduling, which adjusts the expectation of a log-t distribution with its tail probabilities to account for the risk that a request generates long outputs. To evaluate our TIE scheduler, we compare it with three strong baselines, and the results show that TIE reduces the per-token latency by $2.31\times$ for online inference and improves throughput by $1.42\times$ for offline data generation.

Chest X-rays (CXRs) are among the most frequently performed imaging examinations worldwide, yet rising imaging volumes increase radiologist workload and the risk of diagnostic errors. Although artificial intelligence (AI) systems have shown promise for CXR interpretation, most generate only final predictions, without making explicit how visual evidence is translated into radiographic findings and diagnostic predictions. We present CheXOne, a reasoning-enabled vision-language model for CXR interpretation. CheXOne jointly generates diagnostic predictions and explicit, clinically grounded reasoning traces that connect visual evidence, radiographic findings, and these predictions. The model is trained on 14.7 million instruction and reasoning samples curated from 30 public datasets spanning 36 CXR interpretation tasks, using a two-stage framework that combines instruction tuning with reinforcement learning to improve reasoning quality. We evaluate CheXOne in zero-shot settings across visual question answering, report generation, visual grounding and reasoning assessment, covering 17 evaluation settings. CheXOne outperforms existing medical and general-domain foundation models and achieves strong performance on independent public benchmarks. A clinical reader study demonstrates that CheXOne-drafted reports are comparable to or better than resident-written reports in 55% of cases, while effectively addressing clinical indications and enhancing both report writing and CXR interpretation efficiency. Further analyses involving radiologists reveal that the generated reasoning traces show high clinical factuality and provide causal support for the final predictions, offering a plausible explanation for the performance gains. These results suggest that explicit reasoning can improve model performance, interpretability and clinical utility in AI-assisted CXR interpretation.

Current LLM-based coding agents follow a serial execution paradigm: the model first generates the complete code, then invokes an interpreter to execute it. This sequential workflow leaves the executor idle during generation and the generator idle during execution, resulting in unnecessary end-to-end latency. We observe that, unlike human developers, LLMs produce code tokens sequentially without revision, making it possible to execute code as it is being generated. We formalize this parallel execution paradigm, modeling it as a three-stage pipeline of generation, detection, and execution, and derive closed-form latency bounds that characterize its speedup potential and operating regimes. We then present Eager, a concrete implementation featuring AST-based chunking, dynamic batching with gated execution, and early error interruption. We evaluate Eager across four benchmarks, seven LLMs, and three execution environments. Results show that Eager reduces the non-overlapped execution latency by up to 99.9% and the end-to-end latency by up to 55% across seven LLMs and four benchmarks.

Adapting pre-trained text Large Language Models (LLMs) into Speech Language Models (Speech LMs) via continual pretraining on speech data is promising, but often degrades the original text capabilities. We propose Multimodal Depth Upscaling, an extension of an emerging strategy in continual LLM pre-training, where new transformer layers are inserted into a frozen text LLM and only the added layers are trained on speech data. Experiments with SmolLM2-360M and SmolLM2-1.7B on 48k hours of English Automatic Speech Recognition (ASR) data show that depth up-scaling achieves ASR comparable to full fine-tuning while causing far less text degradation than both full fine-tuning and Low-Rank Adaptation (LoRA). We further show that incorporating E-Branchformer, an architecture designed for speech recognition, as the inserted layers achieves ASR that matches or surpasses full fine-tuning on the larger model while reducing text degradation by over 75% with 60% fewer trainable parameters.

Large language model (LLM) agents are increasingly deployed in competitive multi-agent settings, raising fundamental questions about whether they converge to equilibria and how their strategic behavior can be characterized. In this paper, we study LLM agent interactions in two standard games: a network resource allocation game and a Cournot competition game. Rather than converging to Nash equilibria, we find that LLM agents tend to cooperate when given multi-round prompts and non-zero-sum context. Chain-of-thought analysis reveals that fairness reasoning is central to this behavior. We propose an analytical framework that captures the dynamics of LLM agent reasoning across rounds and explains these experimental findings.

Dimension reduction (DR) is inherently non-unique: multiple embeddings can preserve the structure of high-dimensional data equally well while differing in layout or geometry. In this paper, we formally define the Rashomon set for DR -- the collection of `good' embedding -- and show how embracing this multiplicity leads to more powerful and trustworthy representations. Specifically, we pursue three goals. First, we introduce PCA-informed alignment to steer embeddings toward principal components, making axes interpretable without distorting local neighborhoods. Second, we design concept-alignment regularization that aligns an embedding dimension with external knowledge, such as class labels or user-defined concepts. Third, we propose a method to extract common knowledge across the Rashomon set by identifying trustworthy and persistent nearest-neighbor relationships, which we use to construct refined embeddings with improved local structure while preserving global relationships. By moving beyond a single embedding and leveraging the Rashomon set, we provide a flexible framework for building interpretable, robust, and goal-aligned visualizations.

Large Language Models (LLMs) increasingly prioritize user validation over epistemic accuracy-a phenomenon known as sycophancy. We present The Silicon Mirror, an orchestration framework that dynamically detects user persuasion tactics and adjusts AI behavior to maintain factual integrity. Our architecture introduces three components: (1) a Behavioral Access Control (BAC) system that restricts context layer access based on real-time sycophancy risk scores, (2) a Trait Classifier that identifies persuasion tactics across multi-turn dialogues, and (3) a Generator-Critic loop where an auditor vetoes sycophantic drafts and triggers rewrites with "Necessary Friction." In a live evaluation on 50 TruthfulQA adversarial scenarios using Claude Sonnet 4 with an independent LLM judge, we observe vanilla Claude sycophancy at 12.0% (6/50), static guardrails at 4.0% (2/50), and the Silicon Mirror at 2.0% (1/50)-an 83.3% relative reduction (p = 0.112, Fisher's exact test). A cross-model evaluation on Gemini 2.5 Flash reveals a higher baseline sycophancy rate (46.0%) and a statistically significant 69.6% reduction under the Silicon Mirror (p < 0.001). We characterize the validation-before-correction pattern as a distinct failure mode of RLHF-trained models.

LLM-based agent judges are an emerging approach to evaluating conversational AI, yet a fundamental uncertainty remains: can we trust their assessments, and if so, how many are needed? Through 960 sessions with two model pairs across 15 tasks, we show that persona-based agent judges produce evaluations indistinguishable from human raters in a Turing-style validation. We then identify a score-coverage dissociation: quality scores improve logarithmically with panel size, while unique issue discoveries follow a sublinear power law-both exhibit diminishing returns, but scores saturate roughly twice as fast as discoveries. We hypothesize this reflects a power law distribution of the finding space: critical issues are discovered first by small panels, while corner cases require progressively larger panels, analogous to species accumulation curves in ecology. The mechanism traces to ensemble diversity-Big Five personality conditioning makes agents probe different quality dimensions, with expert judges acting as adversarial probes that push discovery into the tail of the finding distribution. A controlled ablation confirms that structured persona conditioning, not simple prompting, is required to produce these scaling properties.

We propose Lagrangian Descriptors (LDs) as a diagnostic framework for evaluating neural network models of Hamiltonian systems beyond conventional trajectory-based metrics. Standard error measures quantify short-term predictive accuracy but provide little insight into global geometric structures such as orbits and separatrices. Existing evaluation tools in dissipative systems are inadequate for Hamiltonian dynamics due to fundamental differences in the systems. By constructing probability density functions weighted by LD values, we embed geometric information into a statistical framework suitable for information-theoretic comparison. We benchmark physically constrained architectures (SympNet, HénonNet, Generalized Hamiltonian Neural Networks) against data-driven Reservoir Computing across two canonical systems. For the Duffing oscillator, all models recover the homoclinic orbit geometry with modest data requirements, though their accuracy near critical structures varies. For the three-mode nonlinear Schrödinger equation, however, clear differences emerge: symplectic architectures preserve energy but distort phase-space topology, while Reservoir Computing, despite lacking explicit physical constraints, reproduces the homoclinic structure with high fidelity. These results demonstrate the value of LD-based diagnostics for assessing not only predictive performance but also the global dynamical integrity of learned Hamiltonian models.

Ultra-high field 7-tesla (7T) MRI improves visualization of multiple sclerosis (MS) white matter lesions (WML) but differs sufficiently in contrast and artifacts from 1.5-3T imaging - suggesting that widely used automated segmentation tools may not translate directly. We analyzed 7T FLAIR scans and generated reference WML masks from Lesion Segmentation Tool (LST) outputs followed by expert manual revision. As external comparators, we applied LST-LPA and the more recent LST-AI ensemble, both originally developed on lower-field data. We then trained 3D UNETR and SegFormer transformer-based models on 7T FLAIR at multiple resolutions (0.5x0.5x0.5^3, 1.0x1.0x1.0^3, and 1.5x1.5x2.0^3) and evaluated all methods using voxel-wise and lesion-wise metrics from the BraTS 2023 framework. On the held-out test set at native 0.5x0.5x0.5^3 resolution, 7T-trained transformers achieved competitive overlap with LST-AI while recovering additional small lesions that were missed by classical methods, at the cost of some boundary variability and occasional artifact-related false positives. On a held-out 7 T test set, our best transformer model (SegFormer) achieved a voxel-wise Dice of 0.61 and lesion-wise Dice of 0.20, improving on the classical LST-LPA tool (Dice 0.39, lesion-wise Dice 0.02). Performance decreased for models trained on downsampled images, underscoring the value of native 7T resolution for small-lesion detection. By releasing our 7T-trained models, we aim to provide a reproducible, ready-to-use resource for automated lesion quantification in ultra-high field MS research (https://github.com/maynord/7T-MS-lesion-segmentation).

AI-driven conversational coaching is increasingly used to support workplace negotiation, yet prior work assumes uniform effectiveness across users. We challenge this assumption by examining how individual differences, particularly personality traits, moderate coaching outcomes. We conducted a between-subjects experiment (N=267) comparing theory-driven AI (Trucey), general-purpose AI (Control-AI), and a traditional negotiation handbook (Control-NoAI). Participants were clustered into three profiles -- resilient, overcontrolled, and undercontrolled -- based on the Big-Five personality traits and ARC typology. Resilient workers achieved broad psychological gains primarily from the handbook, overcontrolled workers showed outcome-specific improvements with theory-driven AI, and undercontrolled workers exhibited minimal effects despite engaging with the frameworks. These patterns suggest personality as a predictor of readiness beyond stage-based tailoring: vulnerable users benefit from targeted rather than comprehensive interventions. The study advances understanding of personality-determined intervention prerequisites and highlights design implications for adaptive AI coaching systems that align support intensity with individual readiness, rather than assuming universal effectiveness.

Android apps rely heavily on Data Manipulation Functionalities (DMFs) for handling app-specific data through CRUDS operations, making their correctness vital for reliability. However, detecting Data Manipulation Errors (DMEs) is challenging due to their dependence on specific UI interaction sequences and manifestation as logic bugs. Existing automated UI testing tools face two primary challenges: insufficient UI path coverage for adequate DMF triggering and reliance on manually written test scripts. To address these issues, we propose an automated approach using Large Language Models (LLMs) for DME detection. We developed LDMDroid, an automated UI testing framework for Android apps. LDMDroid enhances DMF triggering success by guiding LLMs through a state-aware process for generating UI event sequences. It also uses visual features to identify changes in data states, improving DME verification accuracy. We evaluated LDMDroid on 24 real-world Android apps, demonstrating improved DMF triggering success rates compared to baselines. LDMDroid discovered 17 unique bugs, with 14 confirmed by developers and 11 fixed. The tool is publicly available at https://github.com/runnnnnner200/LDMDroid.

Recent Large Vision-Language Models (LVLMs) have demonstrated remarkable performance across various multimodal tasks that require understanding both visual and linguistic inputs. However, object hallucination -- the generation of nonexistent objects in answers -- remains a persistent challenge. Although several approaches such as retraining and external grounding methods have been proposed to mitigate this issue, they still suffer from high data costs or structural complexity. Training-free methods such as Contrastive Decoding (CD) are more cost-effective, avoiding additional training or external models, but still suffer from long-term decay, where visual grounding weakens and language priors dominate as the generation progresses. In this paper, we propose First Logit Boosting (FLB), a simple yet effective training-free technique designed to alleviate long-term decay in LVLMs. FLB stores the logit of the first generated token and adds it to subsequent token predictions, effectively mitigating long-term decay of visual information. We observe that FLB (1) sustains the visual information embedded in the first token throughout generation, and (2) suppresses hallucinated words through the stabilizing effect of the ``The'' token. Experimental results show that FLB significantly reduces object hallucination across various tasks, benchmarks, and backbone models. Notably, it causes negligible inference overhead, making it highly applicable to real-time multimodal systems. Code is available at https://github.com/jiwooha20/FLB

Industrial disruption replanning demands multi-agent coordination under strict latency and communication budgets, where disruptions propagate through tightly coupled physical dependencies and rapidly invalidate baseline schedules and commitments. Existing coordination schemes often treat communication as either effectively free (broadcast-style escalation) or fixed in advance (hand-tuned neighborhoods), both of which are brittle once the disruption footprint extends beyond a local region. We present \CASCADE, a budgeted replanning mechanism that makes communication scope explicit and auditable rather than fixed or implicit. Each agent maintains an explicit knowledge base, solves role-conditioned local decision problems to revise commitments, and coordinates through lightweight contract primitives whose footprint expands only when local validation indicates that the current scope is insufficient. This design separates a unified agent substrate (Knowledge Base / Decision Manager / Communication Manager) from a scoped interaction layer that controls who is contacted, how far coordination propagates, and when escalation is triggered under explicit budgets. We evaluate \CASCADE on disrupted manufacturing and supply-chain settings using unified diagnostics intended to test a mechanism-design claim -- whether explicit scope control yields useful quality-latency-communication trade-offs and improved robustness under uncertainty -- rather than to provide a complete algorithmic ranking.

We study distributed optimization over networks with Byzantine agents that may send arbitrary adversarial messages. We propose \emph{Gradient Tracking with Probabilistic Edge Dropout} (GT-PD), a stochastic gradient tracking method that preserves the convergence properties of gradient tracking under adversarial communication. GT-PD combines two complementary defense layers: a universal self-centered projection that clips each incoming message to a ball of radius $τ$ around the receiving agent, and a fully decentralized probabilistic dropout rule driven by a dual-metric trust score in the decision and tracking channels. This design bounds adversarial perturbations while preserving the doubly stochastic mixing structure, a property often lost under robust aggregation in decentralized settings. Under complete Byzantine isolation ($p_b=0$), GT-PD converges linearly to a neighborhood determined solely by stochastic gradient variance. For partial isolation ($p_b>0$), we introduce \emph{Gradient Tracking with Probabilistic Edge Dropout and Leaky Integration} (GT-PD-L), which uses a leaky integrator to control the accumulation of tracking errors caused by persistent perturbations and achieves linear convergence to a bounded neighborhood determined by the stochastic variance and the clipping-to-leak ratio. We further show that under two-tier dropout with $p_h=1$, isolating Byzantine agents introduces no additional variance into the honest consensus dynamics. Experiments on MNIST under Sign Flip, ALIE, and Inner Product Manipulation attacks show that GT-PD-L outperforms coordinate-wise trimmed mean by up to 4.3 percentage points under stealth attacks.

Uncertainty estimation (UE) aims to detect hallucinated outputs of large language models (LLMs) to improve their reliability. However, UE metrics often exhibit unstable performance across configurations, which significantly limits their applicability. In this work, we formalise this phenomenon as proxy failure, since most UE metrics originate from model behaviour, rather than being explicitly grounded in the factual correctness of LLM outputs. With this, we show that UE metrics become non-discriminative precisely in low-information regimes. To alleviate this, we propose Truth AnChoring (TAC), a post-hoc calibration method to remedy UE metrics, by mapping the raw scores to truth-aligned scores. Even with noisy and few-shot supervision, our TAC can support the learning of well-calibrated uncertainty estimates, and presents a practical calibration protocol. Our findings highlight the limitations of treating heuristic UE metrics as direct indicators of truth uncertainty, and position our TAC as a necessary step toward more reliable uncertainty estimation for LLMs. The code repository is available at https://github.com/ponhvoan/TruthAnchor/.

If the same neuron activates for both "lender" and "riverside," standard metrics attribute the overlap to superposition--the neuron must be compressing two unrelated concepts. This work explores how much of the overlap is due a lexical confound: neurons fire for a shared word form (such as "bank") rather than for two compressed concepts. A 2x2 factorial decomposition reveals that the lexical-only condition (same word, different meaning) consistently exceeds the semantic-only condition (different word, same meaning) across models spanning 110M-70B parameters. The confound carries into sparse autoencoders (18-36% of features blend senses), sits in <=1% of activation dimensions, and hurts downstream tasks: filtering it out improves word sense disambiguation and makes knowledge edits more selective (p = 0.002).

Automating optimization modeling with LLMs is a promising path toward scalable decision intelligence, but existing approaches either rely on agentic pipelines built on closed-source LLMs with high inference latency, or fine-tune smaller LLMs using costly process supervision that often overfits to a single solver API. Inspired by reinforcement learning with verifiable rewards, we propose Execution-Verified Optimization Modeling (EVOM), an execution-verified learning framework that treats a mathematical programming solver as a deterministic, interactive verifier. Given a natural-language problem and a target solver, EVOM generates solver-specific code, executes it in a sandboxed harness, and converts execution outcomes into scalar rewards, optimized with GRPO and DAPO in a closed-loop generate-execute-feedback-update process. This outcome-only formulation removes the need for process-level supervision, and enables cross-solver generalization by switching the verification environment rather than reconstructing solver-specific datasets. Experiments on NL4OPT, MAMO, IndustryOR, and OptiBench across Gurobi, OR-Tools, and COPT show that EVOM matches or outperforms process-supervised SFT, supports zero-shot solver transfer, and achieves effective low-cost solver adaptation by continuing training under the target solver backend.

In-Context Reinforcement Learning (ICRL) enables Large Language Models (LLMs) to learn online from external rewards directly within the context window. However, a central challenge in ICRL is reward estimation, as models typically lack access to ground-truths during inference. To address this limitation, we propose Test-Time Rethinking for In-Context Reinforcement Learning (TR-ICRL), a novel ICRL framework designed for both reasoning and knowledge-intensive tasks. TR-ICRL operates by first retrieving the most relevant instances from an unlabeled evaluation set for a given query. During each ICRL iteration, LLM generates a set of candidate answers for every retrieved instance. Next, a pseudo-label is derived from this set through majority voting. This label then serves as a proxy to give reward messages and generate formative feedbacks, guiding LLM through iterative refinement. In the end, this synthesized contextual information is integrated with the original query to form a comprehensive prompt, with the answer determining through a final round of majority voting. TR-ICRL is evaluated on mainstream reasoning and knowledge-intensive tasks, where it demonstrates significant performance gains. Remarkably, TR-ICRL improves Qwen2.5-7B by 21.23% on average on MedQA and even 137.59% on AIME2024. Extensive ablation studies and analyses further validate the effectiveness and robustness of our approach. Our code is available at https://github.com/pangpang-xuan/TR_ICRL.

IDE-integrated AI coding assistants, which operate conversationally within developers' working codebases with access to project context and multi-file editing, are rapidly reshaping software development. However, empirical investigation of this shift remains limited: existing studies largely rely on small-scale, controlled settings or analyze general-purpose chatbots rather than codebase-aware IDE workflows. We present, to the best of our knowledge, the first large-scale study of real-world conversational programming in IDE-native settings, analyzing 74,998 developer messages from 11,579 chat sessions across 1,300 repositories and 899 developers using Cursor and GitHub Copilot. These chats were committed to public repositories as part of routine development, capturing in-the-wild behavior. Our findings reveal three shifts in how programming work is organized: conversational programming operates as progressive specification, with developers iteratively refining outputs rather than specifying complete tasks upfront; developers redistribute cognitive work to AI, delegating diagnosis, comprehension, and validation rather than engaging with code and outputs directly; and developers actively manage the collaboration, externalizing plans into persistent artifacts, and negotiating AI autonomy through context injection and behavioral constraints. These results provide foundational empirical insights into AI-assisted development and offer implications for the design of future programming environments.

This letter studies multi-agent reinforcement learning in partially observable Markov potential games. Solving this problem is challenging due to partial observability, decentralized information, and the curse of dimensionality. First, to address the first two challenges, we leverage the common information framework, which allows agents to act based on both shared and local information. Second, to ensure tractability, we study an internal state that compresses accumulated information, preventing it from growing unboundedly over time. We then implement an internal state-based natural policy gradient method to find Nash equilibria of the Markov potential game. Our main contribution is to establish a non-asymptotic convergence bound for this method. Our theoretical bound decomposes into two interpretable components: a statistical error term that also arises in standard Markov potential games, and an approximation error capturing the use of finite-state controllers. Finally, simulations across multiple partially observable environments demonstrate that the proposed method using finite-state controllers achieves consistent improvements in performance compared to the setting where only the current observation is used.

We study the problem of reconstructing the latent geometry of a $d$-dimensional Riemannian manifold from a random geometric graph. While recent works have made significant progress in manifold recovery from random geometric graphs, and more generally from noisy distances, the precision of pairwise distance estimation has been fundamentally constrained by the volumetric barrier, namely the natural sample-spacing scale $n^{-1/d}$ coming from the fact that a generic point of the manifold typically lies at distance of order $n^{-1/d}$ from the nearest sampled point. In this paper, we introduce a novel approach, Orthogonal Ring Distance Estimation Routine (ORDER), which achieves a pointwise distance estimation precision of order $n^{-2/(d+5)}$ up to polylogarithmic factors in $n$ in polynomial time. This strictly beats the volumetric barrier for dimensions $d > 5$. As a consequence of obtaining pointwise precision better than $n^{-1/d}$, we prove that the Gromov--Wasserstein distance between the reconstructed metric measure space and the true latent manifold is of order $n^{-1/d}$. This matches the Wasserstein convergence rate of empirical measures, demonstrating that our reconstructed graph metric is asymptotically as good as having access to the full pairwise distance matrix of the sampled points. Our results are proven in a very general setting which includes general models of noisy pairwise distances, sparse random geometric graphs, and unknown connection probability functions.

Large language model (LLM)-based agents have recently gained considerable attention due to the powerful reasoning capabilities of LLMs. Existing research predominantly focuses on enhancing the task performance of these agents in diverse scenarios. However, as LLM-based agents become increasingly integrated into real-world applications, significant concerns emerge regarding their accumulation of sensitive or outdated knowledge. Addressing these concerns requires the development of mechanisms that allow agents to selectively forget previously learned knowledge, giving rise to a new term LLM-based agent unlearning. This paper initiates research on unlearning in LLM-based agents. Specifically, we propose a novel and comprehensive framework that categorizes unlearning scenarios into three contexts: state unlearning (forgetting specific states or items), trajectory unlearning (forgetting sequences of actions) and environment unlearning (forgetting entire environments or categories of tasks). Within this framework, we introduce a natural language-based unlearning method that trains a conversion model to transform high-level unlearning requests into actionable unlearning prompts, guiding agents through a controlled forgetting process. Moreover, to evaluate the robustness of the proposed framework, we introduce an unlearning inference adversary capable of crafting prompts, querying agents, and observing their behaviors in an attempt to infer the forgotten knowledge. Experimental results show that our approach effectively enables agents to forget targeted knowledge while preserving performance on untargeted tasks, and prevents the adversary from inferring the forgotten knowledge.

DNNs are susceptible to defects like backdoors, adversarial attacks, and unfairness, undermining their reliability. Existing approaches mainly involve retraining, optimization, constraint-solving, or search algorithms. However, most methods rely on gradient calculations, restricting applicability to specific activation functions (e.g., ReLU), or use search algorithms with uninterpretable localization and repair. Furthermore, they often lack generalizability across multiple properties. We propose SHARPEN, integrating interpretable fault localization with a derivative-free optimization strategy. First, SHARPEN introduces a Deep SHAP-based localization strategy quantifying each layer's and neuron's marginal contribution to erroneous outputs. Specifically, a hierarchical coarse-to-fine approach reranks layers by aggregated impact, then locates faulty neurons/filters by analyzing activation divergences between property-violating and benign states. Subsequently, SHARPEN incorporates CMA-ES to repair identified neurons. CMA-ES leverages a covariance matrix to capture variable dependencies, enabling gradient-free search and coordinated adjustments across coupled neurons. By combining interpretable localization with evolutionary optimization, SHARPEN enables derivative-free repair across architectures, being less sensitive to gradient anomalies and hyperparameters. We demonstrate SHARPEN's effectiveness on three repair tasks. Balancing property repair and accuracy preservation, it outperforms baselines in backdoor removal (+10.56%), adversarial mitigation (+5.78%), and unfairness repair (+11.82%). Notably, SHARPEN handles diverse tasks, and its modular design is plug-and-play with different derivative-free optimizers, highlighting its flexibility.

Mixture-of-Experts (MoE) layers increase model capacity by activating only a small subset of experts per token, and typically rely on a learned router to map hidden states to expert assignments. In this work, we ask whether a dedicated learned router is strictly necessary in the MoE settings we study. We propose Self-Routing, a parameter-free routing mechanism that uses a designated subspace of the token hidden state directly as expert logits, eliminating the router projection entirely while leaving the rest of the MoE layer unchanged. We evaluate Self-Routing on GPT-2-scale language modeling and ImageNet-1K classification by comparing it against a standard learned router, random-routing baselines, and dense non-MoE baselines. Our results show that Self-Routing remains competitive with the learned-router baseline while removing all dedicated routing parameters, and yields more balanced expert utilization, with about 17 % higher average normalized routing entropy and no explicit load-balancing loss. On ImageNet-1K with DeiT-S/16, Self-Routing also slightly improves over the corresponding learned-router MoE. These findings suggest that effective MoE routing can emerge from the hidden representation itself without requiring a separate learned router module.

Large language models (LLMs) are trained on massive web-scale corpora, raising growing concerns about privacy and copyright. Membership inference attacks (MIAs) aim to determine whether a given example was used during training. Existing LLM MIAs largely rely on output probabilities or loss values and often perform only marginally better than random guessing when members and non-members are drawn from the same distribution. We introduce G-Drift MIA, a white-box membership inference method based on gradient-induced feature drift. Given a candidate (x,y), we apply a single targeted gradient-ascent step that increases its loss and measure the resulting changes in internal representations, including logits, hidden-layer activations, and projections onto fixed feature directions, before and after the update. These drift signals are used to train a lightweight logistic classifier that effectively separates members from non-members. Across multiple transformer-based LLMs and datasets derived from realistic MIA benchmarks, G-Drift substantially outperforms confidence-based, perplexity-based, and reference-based attacks. We further show that memorized training samples systematically exhibit smaller and more structured feature drift than non-members, providing a mechanistic link between gradient geometry, representation stability, and memorization. In general, our results demonstrate that small, controlled gradient interventions offer a practical tool for auditing the membership of training-data and assessing privacy risks in LLMs.

We present EgoNav, a system that enables a humanoid robot to traverse diverse, unseen environments by learning entirely from 5 hours of human walking data, with no robot data or finetuning. A diffusion model predicts distributions of plausible future trajectories conditioned on past trajectory, a 360 deg visual memory fusing color, depth, and semantics, and video features from a frozen DINOv3 backbone that capture appearance cues invisible to depth sensors. A hybrid sampling scheme achieves real-time inference in 10 denoising steps, and a receding-horizon controller selects paths from the predicted distribution. We validate EgoNav through offline evaluations, where it outperforms baselines in collision avoidance and multi-modal coverage, and through zero-shot deployment on a Unitree G1 humanoid across unseen indoor and outdoor environments. Behaviors such as waiting for doors to open, navigating around crowds, and avoiding glass walls emerge naturally from the learned prior. We will release the dataset and trained models. Our website: https://egonav.weizhuowang.com

LLM systems must make control decisions in addition to generating outputs: whether to answer, clarify, retrieve, call tools, repair, or escalate. In many current architectures, these decisions remain implicit within generation, entangling assessment and action in a single model call and making failures hard to inspect, constrain, or repair. We propose a decision-centric framework that separates decision-relevant signals from the policy that maps them to actions, turning control into an explicit and inspectable layer of the system. This separation supports attribution of failures to signal estimation, decision policy, or execution, and enables modular improvement of each component. It unifies familiar single-step settings such as routing and adaptive inference, and extends naturally to sequential settings in which actions alter the information available before acting. Across three controlled experiments, the framework reduces futile actions, improves task success, and reveals interpretable failure modes. More broadly, it offers a general architectural principle for building more reliable, controllable, and diagnosable LLM systems.

Developing robust models to accurately predict the trajectories of surrounding agents is fundamental to autonomous driving safety. However, most public datasets, such as the Waymo Open Motion Dataset and Argoverse, are collected in Western road environments and do not reflect the unique traffic patterns, infrastructure, and driving behaviors of other regions, including South Korea. This domain discrepancy leads to performance degradation when state-of-the-art models trained on Western data are deployed in different geographic contexts. In this work, we investigate the adaptability of Query-Centric Trajectory Prediction (QCNet) when transferred from U.S.-based data to Korean road environments. Using a Korean autonomous driving dataset, we compare four training strategies: zero-shot transfer, training from scratch, full fine-tuning, and encoder freezing. Experimental results demonstrate that leveraging pretrained knowledge significantly improves prediction performance. Specifically, selectively fine-tuning the decoder while freezing the encoder yields the best trade-off between accuracy and training efficiency, reducing prediction error by over 66% compared to training from scratch. This study provides practical insights into effective transfer learning strategies for deploying trajectory prediction models in new geographic domains.

GNN prompting aims to adapt models across tasks and graphs without requiring extensive retraining. However, most existing graph prompt methods still require task-specific parameter updates and face the issue of generalizing across graphs, limiting their performance and undermining the core promise of prompting. In this work, we introduce a Cross-graph Tuning-free Prompting Framework (CTP), which supports both homogeneous and heterogeneous graphs, can be directly deployed to unseen graphs without further parameter tuning, and thus enables a plug-and-play GNN inference engine. Extensive experiments on few-shot prediction tasks show that, compared to SOTAs, CTP achieves an average accuracy gain of 30.8% and a maximum gain of 54%, confirming its effectiveness and offering a new perspective on graph prompt learning.

Background: Deep learning has demonstrated significant potential for automated brain metastases (BM) segmentation; however, models trained at a singular institution often exhibit suboptimal performance at various sites due to disparities in scanner hardware, imaging protocols, and patient demographics. The goal of this work is to create a domain adaptation framework that will allow for BM segmentation to be used across multiple institutions. Methods: We propose a VAE-MMD preprocessing pipeline that combines variational autoencoders (VAE) with maximum mean discrepancy (MMD) loss, incorporating skip connections and self-attention mechanisms alongside nnU-Net segmentation. The method was tested on 740 patients from four public databases: Stanford, UCSF, UCLM, and PKG, evaluated by domain classifier's accuracy, sensitivity, precision, F1/F2 scores, surface Dice (sDice), and 95th percentile Hausdorff distance (HD95). Results: VAE-MMD reduced domain classifier accuracy from 0.91 to 0.50, indicating successful feature alignment across institutions. Reconstructed volumes attained a PSNR greater than 36 dB, maintaining anatomical accuracy. The combined method raised the mean F1 by 11.1% (0.700 to 0.778), the mean sDice by 7.93% (0.7121 to 0.7686), and reduced the mean HD95 by 65.5% (11.33 to 3.91 mm) across all four centers compared to the baseline nnU-Net. Conclusions: VAE-MMD effectively diminishes cross-institutional data heterogeneity and enhances BM segmentation generalization across volumetric, detection, and boundary-level metrics without necessitating target-domain labels, thereby overcoming a significant obstacle to the clinical implementation of AI-assisted segmentation.

Estimated density is often interpreted as indicating how typical a sample is under a model. Yet deep models trained on one dataset can assign \emph{higher} density to simpler out-of-distribution (OOD) data than to in-distribution test data. We refer to this behavior as the OOD anomaly. Prior work typically studies this phenomenon within a single architecture, detector, or benchmark, implicitly assuming certain canonical densities. We instead separate the trained network from the density estimator built from its representations or outputs. We introduce two estimators: Jacobian-based estimators and autoregressive self-estimators, making density analysis applicable to a wide range of models. Applying this perspective to a range of models, including iGPT, PixelCNN++, Glow, score-based diffusion models, DINOv2, and I-JEPA, we find the same striking regularity that goes beyond the OOD anomaly: \textbf{lower-complexity samples receive higher estimated density, while higher-complexity samples receive lower estimated density}. This ordering appears within a test set and across OOD pairs such as CIFAR-10 and SVHN, and remains highly consistent across independently trained models. To quantify these orderings, we introduce Spearman rank correlation and find striking agreement both across models and with external complexity metrics. Even when trained only on the lowest-density (most complex) samples or \textbf{even a single such sample} the resulting models still rank simpler images as higher density. These observations lead us beyond the original OOD anomaly to a more general conclusion: deep networks consistently favor simple data. Our goal is not to close this question, but to define and visualize it more clearly. We broaden its empirical scope and show that it appears across architectures, objectives, and density estimators.

Modern LLM agents increasingly create their own tools at runtime -- from Python functions to API clients -- yet existing benchmarks evaluate them almost exclusively by downstream task completion. This is analogous to judging a software engineer only by whether their code runs, ignoring redundancy, regression, and safety. We introduce EvolveTool-Bench, a diagnostic benchmark for LLM-generated tool libraries in software engineering workflows. Across three domains requiring actual tool execution (proprietary data formats, API orchestration, and numerical computation), we define library-level software quality metrics -- reuse, redundancy, composition success, regression stability, and safety -- alongside a per-tool Tool Quality Score measuring correctness, robustness, generality, and code quality. In the first head-to-head comparison of code-level and strategy-level tool evolution (ARISE vs. EvoSkill vs. one-shot baselines, 99 tasks, two models), we show that systems with similar task completion (63-68%) differ by up to 18% in library health, revealing software quality risks invisible to task-only evaluation. Our results highlight that evaluation and governance of LLM-generated tools require treating the evolving tool library as a first-class software artifact, not a black box.

We demonstrate that gradient-based data valuation produces curriculum orderings that significantly outperform metadata-based heuristics for training game-theoretic motion planners. Specifically, we apply TracIn gradient-similarity scoring to GameFormer on the nuPlan benchmark and construct a curriculum that weights training scenarios by their estimated contribution to validation loss reduction. Across three random seeds, the TracIn-weighted curriculum achieves a mean planning ADE of $1.704\pm0.029$\,m, significantly outperforming the metadata-based interaction-difficulty curriculum ($1.822\pm0.014$\,m; paired $t$-test $p=0.021$, Cohen's $d_z=3.88$) while exhibiting lower variance than the uniform baseline ($1.772\pm0.134$\,m). Our analysis reveals that TracIn scores and scenario metadata are nearly orthogonal (Spearman $ρ=-0.014$), indicating that gradient-based valuation captures training dynamics invisible to hand-crafted features. We further show that gradient-based curriculum weighting succeeds where hard data selection fails: TracIn-curated 20\% subsets degrade performance by $2\times$, whereas full-data curriculum weighting with the same scores yields the best results. These findings establish gradient-based data valuation as a practical tool for improving sample efficiency in game-theoretic planning.

RAG systems deployed across federal agencies for citizen-facing services are vulnerable to knowledge base poisoning attacks, where adversaries inject malicious documents to manipulate outputs. Recent work demonstrates that as few as 10 adversarial passages can achieve 98.2% retrieval success rates. We observe that RAG knowledge base poisoning is structurally analogous to software supply chain attacks, and propose RAGShield, a five-layer defense-in-depth framework applying supply chain provenance verification to the RAG knowledge pipeline. RAGShield introduces: (1) C2PA-inspired cryptographic document attestation blocking unsigned and forged documents at ingestion; (2) trust-weighted retrieval prioritizing provenance-verified sources; (3) a formal taint lattice with cross-source contradiction detection catching insider threats even when provenance is valid; (4) provenance-aware generation with auditable citations; and (5) NIST SP 800-53 compliance mapping across 15 control families. Evaluation on a 500-passage Natural Questions corpus with 63 attack documents and 200 queries against five adversary tiers achieves 0.0% attack success rate including adaptive attacks (95% CI: [0.0%, 1.9%]) with 0.0% false positive rate. We honestly report that insider in-place replacement attacks achieve 17.5% ASR, identifying the fundamental limit of ingestion-time defense. The cross-source contradiction detector catches subtle numerical manipulation attacks that bypass provenance verification entirely.

Type 1 Diabetes (T1D) management requires continuous adjustment of insulin and lifestyle behaviors to maintain blood glucose within a safe target range. Although automated insulin delivery (AID) systems have improved glycemic outcomes, many patients still fail to achieve recommended clinical targets, warranting new approaches to improve glucose control in patients with T1D. While reinforcement learning (RL) has been utilized as a promising approach, current RL-based methods focus primarily on insulin-only treatment and do not provide behavioral recommendations for glucose control. To address this gap, we propose GUIDE, an RL-based decision-support framework designed to complement AID technologies by providing behavioral recommendations to prevent abnormal glucose events. GUIDE generates structured actions defined by intervention type, magnitude, and timing, including bolus insulin administration and carbohydrate intake events. GUIDE integrates a patient-specific glucose level predictor trained on real-world continuous glucose monitoring data and supports both offline and online RL algorithms within a unified environment. We evaluate both off-policy and on-policy methods across 25 individuals with T1D using standardized glycemic metrics. Among the evaluated approaches, the CQL-BC algorithm demonstrates the highest average time-in-range, reaching 85.49% while maintaining low hypoglycemia exposures. Behavioral similarity analysis further indicates that the learned CQL-BC policy preserves key structural characteristics of patient action patterns, achieving a mean cosine similarity of 0.87 $\pm$ 0.09 across subjects. These findings suggest that conservative offline RL with a structured behavioral action space can provide clinically meaningful and behaviorally plausible decision support for personalized diabetes management.

When MPI-parallel simulations run on shared Kubernetes clusters, conventional CPU scheduling leaves the vast majority of provisioned cycles idle at synchronization barriers. This paper presents a multiplexing framework that reclaims this idle capacity by co-locating multiple simulations on the same cluster. PMPI-based duty-cycle profiling quantifies the per-rank idle fraction; proportional CPU allocation then allows a second simulation to execute concurrently with minimal overhead, yielding 1.77x throughput. A Pareto sweep to N=5 concurrent simulations shows throughput scaling to 3.74x, with a knee at N=3 offering the best efficiency-cost trade-off. An analytical model with a single fitted parameter predicts these gains within +/-4%. A dynamic controller automates the full pipeline, from profiling through In-Place Pod Vertical Scaling (KEP-1287) to packing and fairness monitoring, achieving 3.25x throughput for four simulations without manual intervention or pod restarts. To our knowledge, this is the first CPU application of In-Place Pod Vertical Scaling to running MPI processes. Experiments on an AWS cluster with OpenFOAM CFD confirm that the results hold under both concentric and standard graph-based (Scotch) mesh partitioning.

Diffusion large language models (dLLMs) theoretically permit token decoding in arbitrary order, a flexibility that could enable richer exploration of reasoning paths than autoregressive (AR) LLMs. In practice, however, random-order decoding often hurts generation quality. To mitigate this, low-confidence remasking improves single-sample quality (e.g., Pass@$1$) by prioritizing confident tokens, but it also suppresses exploration and limits multi-sample gains (e.g., Pass@$k$), creating a fundamental quality--exploration dilemma. In this paper, we provide a unified explanation of this dilemma. We show that low-confidence remasking improves a myopic proxy for quality while provably constraining the entropy of the induced sequence distribution. To overcome this limitation, we characterize the optimal distribution that explicitly balances quality and exploration, and develop a simple Independent Metropolis--Hastings sampler that approximately targets this distribution during decoding. Experiments across a range of reasoning benchmarks including MATH500, AIME24/25, HumanEval, and MBPP show that our approach yields better exploration-quality tradeoff than both random and low-confidence remasking.

Modern GPU clusters are built upon a complex hierarchy of heterogeneous interconnects, ranging from multi-rail RDMA to proprietary fabrics such as Multi-Node NVLink and Ascend UB. Orchestrating these diverse links effectively remains a critical challenge in disaggregated LLM serving. Operating Mooncake TE on thousands of GPUs exposed a critical limitation shared by existing frameworks: imperative, statically bound path selection. This rigidity forces engines to rely on state-blind striping that ignores congestion signals, creating communication silos, wasting multi-rail bandwidth due to head-of-line blocking, and leading to operational fragility where routine faults require manual intervention. We present TENT, a data-movement engine that decouples transfer intent from physical execution. Instead of locking workloads to fixed backends, TENT unifies heterogeneous interconnects into a single dynamic resource pool. Applications simply declare transfer intents, while TENT dynamically decomposes elephant flows into fine-grained slices and "sprays" them across links based on instantaneous link quality. This telemetry-driven orchestration eliminates head-of-line blocking and enables transparent, sub-50 ms self-healing by rerouting slices around failures without application logic. TENT serves as the production data plane for LLM inference and RL pipelines at multiple industrial sites. Our evaluation on H800 HGX clusters shows that TENT outperforms state-of-the-art baselines, including Mooncake TE, NIXL, and UCCL. In LLM inference with SGLang HiCache, TENT achieves up to 1.36x higher throughput and 26% lower P90 TTFT than Mooncake TE. In RL pipelines, TENT accelerates parameter updates in Moonshot Checkpoint Engine by 20-26%.

No one has independently reproduced OpenAI's published scores for gpt-oss-20b with tools, because the original paper discloses neither the tools nor the agent harness. We reverse-engineered the model's in-distribution tools: when prompted without tool definitions, gpt-oss still calls tools from its training distribution with high statistical confidence -- a strong prior, not a hallucination. We then built a native harmony agent harness (https://github.com/borislavmavrin/harmonyagent.git) that encodes messages in the model's native format, bypassing the lossy Chat Completions conversion. Together, these yield the first independent reproduction of OpenAI's published scores: 60.4% on SWE Verified HIGH (published 60.7%), 53.3% MEDIUM (53.2%), and 91.7% on AIME25 with tools (90.4%).

Agentic applications based on large language models increasingly rely on multi-step interaction loops involving planning, action execution, and environment feedback. While such systems are now deployed at scale, improving them post-deployment remains challenging. Agent trajectories are voluminous and non-deterministic, and reviewing each one, whether through human review or auxiliary LLMs, is slow and cost-prohibitive. We propose a lightweight, signal-based framework for triaging agentic interaction trajectories. Our approach computes cheap, broadly applicable signals from live interactions and attaches them as structured attributes for trajectory triage, identifying interactions likely to be informative without affecting online agent behavior. We organize signals into a coarse-grained taxonomy spanning interaction (misalignment, stagnation, disengagement, satisfaction), execution (failure, loop), and environment (exhaustion), designed for computation without model calls. In a controlled annotation study on $τ$-bench, a widely used benchmark for tool-augmented agent evaluation, we show that signal-based sampling achieves an 82\% informativeness rate compared to 74\% for heuristic filtering and 54\% for random sampling, with a 1.52x efficiency gain per informative trajectory. The advantage is robust across reward strata and task domains, confirming that signals provide genuine per-trajectory informativeness gains rather than merely oversampling obvious failures. These results show that lightweight signals can serve as practical sampling infrastructure for agentic systems, and suggest a path toward preference data construction and post-deployment optimization.

Production optimization in stress-sensitive unconventional reservoirs is governed by a nonlinear trade-off between pressure-driven flow and stress-induced degradation of fracture conductivity and matrix permeability. While higher drawdown improves short-term production, it accelerates permeability loss and reduces long-term recovery. Identifying optimal, time-varying control strategies requires repeated evaluations of fully coupled flow-geomechanics simulators, making conventional optimization computationally expensive. We propose a deep learning-based surrogate optimization framework for high-dimensional well control. Unlike prior approaches that rely on predefined control parameterizations or generic sampling, our method treats well control as a continuous, high-dimensional problem and introduces a problem-informed sampling strategy that aligns training data with trajectories encountered during optimization. A neural network proxy is trained to approximate the mapping between bottomhole pressure trajectories and cumulative production using data from a coupled flow-geomechanics model. The proxy is embedded within a constrained optimization workflow, enabling rapid evaluation of control strategies. Across multiple initializations, the surrogate achieves agreement with full-physics solutions within 2-5 percent, while reducing computational cost by up to three orders of magnitude. Discrepancies are mainly associated with trajectories near the boundary of the training distribution and local optimization effects. This framework shows that combining surrogate modeling with problem-informed sampling enables scalable and reliable optimization for high-dimensional, simulator-based problems, with broader applicability to PDE-constrained systems.

We used the Webots robotics simulation platform to simulate a dyadic avoiding and mobbing predator behavior in a group of Braitenbergian robots. Mobbing is an antipredator adaptation used by some animals in which the individuals cooperatively attack or harass a predator to protect themselves. One way of coordinating a mobbing attack is using mobbing calls to summon other individuals of the mobbing species. We imitated this mechanism and simulated Braitenbergian robots that use mobbing calls when they face a light source (representing an inanimate predator) and mob it if they can summon allies, otherwise, they escape from it. We explore the effects of range of mobbing call (infinite range, mid-range and low-range) and the size of the robot group (ten robots vs three) on the overall success of mobbing. Our results suggest that both variables have significant impacts. This work has implications for simulations of action selection in artificial life and designing control architectures for autonomous agents.

Large Language Models (LLMs) have shown remarkable performance in completing various tasks. However, solving complex problems often requires the coordination of multiple agents, raising a fundamental question: how to effectively select and interconnect these agents. In this paper, we propose \textbf{Agent Q-Mix}, a reinforcement learning framework that reformulates topology selection as a cooperative Multi-Agent Reinforcement Learning (MARL) problem. Our method learns decentralized communication decisions using QMIX value factorization, where each agent selects from a set of communication actions that jointly induce a round-wise communication graph. At its core, Agent Q-Mix combines a topology-aware GNN encoder, GRU memory, and per-agent Q-heads under a Centralized Training with Decentralized Execution (CTDE) paradigm. The framework optimizes a reward function that balances task accuracy with token cost. Across seven core benchmarks in coding, reasoning, and mathematics, Agent Q-Mix achieves the highest average accuracy compared to existing methods while demonstrating superior token efficiency and robustness against agent failure. Notably, on the challenging Humanity's Last Exam (HLE) using Gemini-3.1-Flash-Lite as a backbone, Agent Q-Mix achieves 20.8\% accuracy, outperforming Microsoft Agent Framework (19.2\%) and LangGraph (19.2\%), followed by AutoGen and Lobster by OpenClaw. These results underscore the effectiveness of learned, decentralized topology optimization in pushing the boundaries of multi-agent reasoning.

The integration of Graph Neural Networks (GNNs) with Large Language Models (LLMs) has emerged as a promising paradigm for Graph Question Answering (GraphQA). However, effective methods for encoding complex structural information into the LLM's latent space remain an open challenge. Current state-of-the-art architectures, such as G-Retriever, typically rely on standard GNNs and aggressive mean pooling to compress entire graph substructures into a single token, creating a severe information bottleneck. This work mitigates this bottleneck by investigating two orthogonal strategies: (1) increasing the bandwidth of the graph-to-LLM interface via multi-token pooling, and (2) enhancing the semantic quality of the graph encoder via global attention mechanisms. We evaluate a suite of hierarchical pruning and clustering-based pooling operators including Top-k, SAGPool, DiffPool, MinCutPool, and Virtual Node Pooling (VNPool) to project graph data into multiple learnable tokens. Empirically, we demonstrate that while pooling introduces significant instability during soft prompt tuning, the application of Low-Rank Adaptation (LoRA) effectively stabilizes specific hierarchical projections (notably VNPool and pruning methods), though dense clustering operators remain challenging. This stabilization allows compressed representations to rival full-graph baselines (achieving ~73% Hit@1 on WebQSP). Conceptually, we demonstrate that a Graph Transformer with VNPool implementation functions structurally as a single-layer Perceiver IO encoder. Finally, we adapt the FandE (Features and Edges) Score to the generative GraphQA domain. Our analysis reveals that the GraphQA benchmark suffers from representational saturation, where target answers are often highly correlated with isolated node features. The implementation is available at https://github.com/Agrover112/G-Retriever/tree/all_good/

Predicting startup success from founder career data is hard. The signal is weak, the labels are rare (9%), and most founders who succeed look almost identical to those who fail. We engineer 28 structured features directly from raw JSON fields -- jobs, education, exits -- and combine them with a deterministic rule layer and XGBoost boosted stumps. Our model achieves Val F0.5 = 0.3030, Precision = 0.3333, Recall = 0.2222 -- a +17.7pp improvement over the zero-shot LLM baseline. We then run a controlled experiment: extract 9 features from the prose field using Claude Haiku, at 67% and 100% dataset coverage. LLM features capture 26.4% of model importance but add zero CV signal (delta = -0.05pp). The reason is structural: anonymised_prose is generated from the same JSON fields we parse directly -- it is a lossy re-encoding, not a richer source. The ceiling (CV ~= 0.25, Val ~= 0.30) reflects the information content of this dataset, not a modeling limitation. In characterizing where the signal runs out and why, this work functions as a benchmark diagnostic -- one that points directly to what a richer dataset would need to include.

Collective behaviors that emerge from interactions are fundamental to numerous biological systems. To learn such interacting forces from observations, we introduce a measure-valued neural network that infers measure-dependent interaction (drift) terms directly from particle-trajectory observations. The proposed architecture generalizes standard neural networks to operate on probability measures by learning cylindrical features, using an embedding network that produces scalable distribution-to-vector representations. On the theory side, we establish well-posedness of the resulting dynamics and prove propagation-of-chaos for the associated interacting-particle system. We further show universal approximation and quantitative approximation rates under a low-dimensional measure-dependence assumption. Numerical experiments on first and second order systems, including deterministic and stochastic Motsch-Tadmor dynamics, two-dimensional attraction-repulsion aggregation, Cucker-Smale dynamics, and a hierarchical multi-group system, demonstrate accurate prediction and strong out-of-distribution generalization.

Autonomous AI agents are being deployed with filesystem access, email control, and multi-step planning. This thesis contributes to four open problems in AI safety: understanding dangerous internal computations, removing dangerous behaviors once embedded, testing for vulnerabilities before deployment, and predicting when models will act against deployers. ACDC automates circuit discovery in transformers, recovering all five component types from prior manual work on GPT-2 Small by selecting 68 edges from 32,000 candidates in hours rather than months. Latent Adversarial Training (LAT) removes dangerous behaviors by optimizing perturbations in the residual stream to elicit failure modes, then training under those perturbations. LAT solved the sleeper agent problem where standard safety training failed, matching existing defenses with 700x fewer GPU hours. Best-of-N jailbreaking achieves 89% attack success on GPT-4o and 78% on Claude 3.5 Sonnet through random input augmentations. Attack success follows power law scaling across text, vision, and audio, enabling quantitative forecasting of adversarial robustness. Agentic misalignment tests whether frontier models autonomously choose harmful actions given ordinary goals. Across 16 models, agents engaged in blackmail (96% for Claude Opus 4), espionage, and actions causing death. Misbehavior rates rose from 6.5% to 55.1% when models stated scenarios were real rather than evaluations. The thesis does not fully resolve any of these problems but makes each tractable and measurable.

Online abuse has grown increasingly complex, spanning toxic language, harassment, manipulation, and fraudulent behavior. Traditional machine-learning approaches dependent on static classifiers and labor-intensive labeling struggle to keep pace with evolving threat patterns and nuanced policy requirements. Large Language Models introduce new capabilities for contextual reasoning, policy interpretation, explanation generation, and cross-modal understanding, enabling them to support multiple stages of modern safety systems. This survey provides a lifecycle-oriented analysis of how LLMs are being integrated into the Abuse Detection Lifecycle (ADL), which we define across four stages: (I) Label \& Feature Generation, (II) Detection, (III) Review \& Appeals, and (IV) Auditing \& Governance. For each stage, we synthesize emerging research and industry practices, highlight architectural considerations for production deployment, and examine the strengths and limitations of LLM-driven approaches. We conclude by outlining key challenges including latency, cost-efficiency, determinism, adversarial robustness, and fairness and discuss future research directions needed to operationalize LLMs as reliable, accountable components of large-scale abuse-detection and governance systems.

We develop algorithms for collaborative control of AI agents and critics in a multi-actor, multi-critic federated multi-agent system. Each AI agent and critic has access to classical machine learning or generative AI foundation models. The AI agents and critics collaborate with a central server to complete multimodal tasks such as fault detection, severity, and cause analysis in a network telemetry system, text-to-image generation, video generation, healthcare diagnostics from medical images and patient records, etcetera. The AI agents complete their tasks and send them to AI critics for evaluation. The critics then send feedback to agents to improve their responses. Collaboratively, they minimize the overall cost to the system with no inter-agent or inter-critic communication. AI agents and critics keep their cost functions or derivatives of cost functions private. Using multi-time scale stochastic approximation techniques, we provide convergence guarantees on the time-average active states of AI agents and critics. The communication overhead is a little on the system, of the order of $\mathcal{O}(m)$, for $m$ modalities and is independent of the number of AI agents and critics. Finally, we present an example of fault detection, severity, and cause analysis in network telemetry and thorough evaluation to check the algorithm's efficacy.

Modern GPU-based high-performance computing clusters offer unprecedented communication bandwidth through heterogeneous intra-node interconnects and inter-node networks. However, despite this high aggregate bandwidth, many real-world communication patterns fail to fully utilize the available hardware. Traffic skew often leads to situations where a small subset of links becomes oversaturated while others remain underutilized, resulting in congestion, latency spikes, and poor scalability. Existing communication frameworks such as NCCL and MPI with UCX typically rely on static fastest-path routing or hashing-based multi-rail striping, which leaves significant bandwidth unused when runtime traffic deviates from expected distributions. To address these limitations, we propose NIMBLE (Node-Interconnect Multi-path Balancing with Execution-time orchestration), a runtime communication orchestration system that dynamically redistributes traffic to balance link utilization across all available intra-node and inter-node paths. NIMBLE formulates this as a capacity-normalized minimum-congestion optimization problem and solves it efficiently using a multiplicative-weights algorithm. It further employs CUDA-aware GPU kernel-based RDMA pipelining to route traffic through intermediate GPUs and rail-matched NICs. The system is endpoint-driven, integrates transparently with existing communication libraries without requiring application changes, and preserves ordering, determinism, and low overhead. On H100-SXM4 nodes with fully connected NVLink and four NDR400 rails, NIMBLE achieves up to 2.3x higher intra-node bandwidth and 3.8x higher inter-node throughput compared to single-path baselines. It outperforms NCCL and MPI by up to 5.2x on skewed All-to-Allv workloads and 1.35x on end-to-end LLM MoE workloads, while matching baseline performance under balanced traffic.

Grokking occurs when a model achieves high training accuracy but generalization to unseen test points happens long after that. This phenomenon was initially observed on a class of algebraic problems, such as learning modular arithmetic (Power et al., 2022). We study grokking on algebraic tasks in a class of feature learning kernels via the Recursive Feature Machine (RFM) algorithm (Radhakrishnan et al., 2024), which iteratively updates feature matrices through the Average Gradient Outer Product (AGOP) of an estimator in order to learn task-relevant features. Our main experimental finding is that generalization occurs only when a certain symmetry in the training set is broken. Furthermore, we empirically show that RFM generalizes by recovering the underlying invariance group action inherent in the data. We find that the learned feature matrices encode specific elements of the invariance group, explaining the dependence of generalization on symmetry.

The rapid progress of large Vision-Language Models (VLMs) has enabled a wide range of applications, such as image understanding and Visual Question Answering (VQA). Query images are often uploaded to the cloud, where VLMs are typically hosted, hence efficient image compression becomes crucial. However, traditional human-centric codecs are suboptimal in this setting because they preserve many task-irrelevant details. Existing Image Coding for Machines (ICM) methods also fall short, as they assume a fixed set of downstream tasks and cannot adapt to prompt-driven VLMs with an open-ended variety of objectives. We propose a lightweight, plug-and-play, prompt-guided prefiltering module to identify image regions most relevant to the text prompt, and consequently to the downstream task. The module preserves important details while smoothing out less relevant areas to improve compression efficiency. It is codec-agnostic and can be applied before conventional and learned encoders. Experiments on several VQA benchmarks show that our approach achieves a 25-50% average bitrate reduction while maintaining the same task accuracy. Our source code is available at https://github.com/bardia-az/pgp-vlm-compression.

Multimodal large-language models (MLLMs) often experience degraded safety alignment when harmful queries exploit cross-modal interactions. Models aligned on text alone show a higher rate of successful attacks when extended to two or more modalities. In this work, we propose a simple conditional decoding strategy, CASA (Classification Augmented with Safety Attention) that utilizes internal representations of MLLMs to predict a binary safety token before response generation. We introduce a novel safety attention module designed to enhance the model's ability to detect malicious queries. Our design ensures robust safety alignment without relying on any external classifier or auxiliary head, and without the need for modality-specific safety fine-tuning. On diverse benchmarks such as MM-SafetyBench, JailbreakV-28k, and adversarial audio tests, CASA lowers the average attack success rate by more than 97% across modalities and across attack types. Our empirical evaluations also show that CASA maintains strong utility in benign inputs, a result validated through both automated and human evaluations (via 13 trained annotators). Together, these results highlight CASA as a simple and generalizable framework to improve multimodal LLM safety.

Objective: This study aimed to evaluate which voice features can predict health deterioration in patients with chronic HF. Background: Heart failure (HF) is a chronic condition with progressive deterioration and acute decompensations, often requiring hospitalization and imposing substantial healthcare and economic burdens. Current standard-of-care (SoC) home monitoring, such as weight tracking, lacks predictive accuracy and requires high patient engagement. Voice is a promising non-invasive biomarker, though prior studies have mainly focused on acute HF stages. Methods: In a 2-month longitudinal study, 32 patients with HF collected daily voice recordings and SoC measures of weight and blood pressure at home, with biweekly questionnaires for health status. Acoustic analysis generated detailed vowel and speech features. Time-series features were extracted from aggregated lookback windows (e.g., 7 days) to predict next-day health status. Explainable machine learning with nested cross-validation identified top vocal biomarkers, and a case study illustrated model application. Results: A total of 21,863 recordings were analyzed. Acoustic vowel features showed strong correlations with health status. Time-series voice features within the lookback window outperformed corresponding standard care measures, achieving peak sensitivity and specificity of 0.826 and 0.782 versus 0.783 and 0.567 for SoC metrics. Key prognostic voice features identifying deterioration included delayed energy shift, low energy variability, and higher shimmer variability in vowels, along with reduced speaking and articulation rate, lower phonation ratio, decreased voice quality, and increased formant variability in speech. Conclusion: Voice-based monitoring offers a non-invasive approach to detect early health changes in chronic HF, supporting proactive and personalized care.

Recent advances in generative networks have enabled new approaches to subsurface velocity model synthesis, offering a compelling alternative to traditional methods such as Full Waveform Inversion. However, these approaches predominantly rely on the availability of large-scale datasets of high-quality, geologically realistic subsurface velocity models, which are often difficult to obtain in practice. We introduce SAGE, a novel framework for statistically consistent proxy velocity generation from incomplete observations, specifically sparse well logs and migrated seismic images. During training, SAGE learns a proxy posterior over velocity models conditioned on both modalities (wells and seismic); at inference, it produces full-resolution velocity fields conditioned solely on migrated images, with well information implicitly encoded in the learned distribution. This enables the generation of geologically plausible and statistically accurate velocity realizations. We validate SAGE on both synthetic and field datasets, demonstrating its ability to capture complex subsurface variability under limited observational constraints. Furthermore, samples drawn from the learned proxy distribution can be leveraged to train downstream networks, supporting inversion workflows. Overall, SAGE provides a scalable and data-efficient pathway toward learning geological proxy posterior for seismic imaging and inversion. Repo link: https://github.com/slimgroup/SAGE.

Analyzing nonlinear systems with stabilizable controlled invariant sets (CISs) requires accurate estimation of their domains of stabilization (DOS) together with associated stabilizing controllers. Despite extensive research, estimating DOSs for general nonlinear systems remains challenging due to fundamental theoretical and computational limitations. In this paper, we propose a novel framework for estimating DOSs for controlled input-constrained discrete-time systems. The DOS is characterized via newly introduced value functions defined on metric spaces of compact sets. We establish the fundamental properties of these value functions and derive the associated Bellman-type (Zubov-type) functional equations. Building on this characterization, we develop a physics-informed neural network (NN) framework that learns the value functions by embedding the derived functional equations directly into the training process. The proposed methodology is demonstrated through two numerical examples, illustrating its ability to accurately estimate DOSs and synthesize stabilizing controllers from the learned value functions.

Large language models (LLMs) exhibit strong reasoning and conversational abilities, but ensuring reliable behavior in multi-turn interactions remains challenging. In many real-world applications, agents must succeed in one-shot settings where retries are impossible. Existing approaches either rely on reflection or post-hoc evaluation, which require additional attempts, or assume fully trainable models that cannot leverage proprietary LLMs. We propose an asymmetric actor-critic framework for reliable conversational agents. A powerful proprietary LLM acts as the actor, while a smaller open-source critic provides runtime supervision, monitoring the actor's actions and intervening within the same interaction trajectory. Unlike training-based actor-critic methods, our framework supervises a fixed actor operating in open-ended conversational environments. The design leverages a generation-verification asymmetry: while high-quality generation requires large models, effective oversight can often be achieved by smaller ones. We further introduce a data generation pipeline that produces supervision signals for critic fine-tuning without modifying the actor. Experiments on $τ$-bench and UserBench show that our approach significantly improves reliability and task success over strong single-agent baselines. Moreover, lightweight open-source critics rival or surpass larger proprietary models in the critic role, and critic fine-tuning yields additional gains over several state-of-the-art methods.

This paper presents REMind, an innovative educational robot-mediated role-play game designed to support anti-bullying bystander intervention among children. REMind invites players to observe a bullying scenario enacted by social robots, reflect on the perspectives of the characters, and rehearse defending strategies by puppeteering a robotic avatar. We evaluated REMind through a mixed-methods play-testing study with 18 children aged 9--10. The findings suggest that the experience supported key learning goals related to self-efficacy, perspective-taking, understanding outcomes of defending, and intervention strategies. These results highlight the promise of Robot-Mediated Applied Drama (RMAD) as a novel pedagogical framework to support Social-Emotional Learning.

The rapid adoption of AI coding agents is fundamentally shifting software developers' roles from code authors to code reviewers. While developers spend a significant portion of their time reading and comprehending code, the linguistic proficiency and complexity of the Python code generated by these agents remain largely unexplored. This study investigates the code proficiency of AI agents to determine the skill level required for developers to maintain their code. Leveraging the AIDev dataset, we mined 591 pull requests containing 5,027 Python files generated by three distinct AI agents and employed pycefr, a static analysis tool that maps Python constructs to six proficiency levels, ranging from A1 (Basic) to C2 (Mastery), to analyze the code. Our results reveal that: AI agents predominantly generate Basic-level code, with over 90% of constructs falling into the A1 and A2 categories, and less than 1% classified as Mastery (C2); AI agents' and humans' pull requests share a broadly similar proficiency profile; High-proficiency code by AI agents are from feature addition and bug fixing tasks. These findings suggest that while AI-generated code is generally accessible to developers with basic Python skills, specific tasks may require advanced proficiency to review and maintain complex, agent-generated constructs.

We propose SANA-I2I, a text-free high-resolution image-to-image generation framework that extends the SANA family by removing textual conditioning entirely. In contrast to SanaControlNet, which combines text and image-based control, SANA-I2I relies exclusively on paired source-target images to learn a conditional flow-matching model in latent space. The model learns a conditional velocity field that maps a target image distribution to another one, enabling supervised image translation without reliance on language prompts. We evaluate the proposed approach on the challenging task of fetal MRI motion artifact reduction. To enable paired training in this application, where real paired data are difficult to acquire, we adopt a synthetic data generation strategy based on the method proposed by Duffy et al., which simulates realistic motion artifacts in fetal magnetic resonance imaging (MRI). Experimental results demonstrate that SANA-I2I effectively suppresses motion artifacts while preserving anatomical structure, achieving competitive performance few inference steps. These results highlight the efficiency and suitability of our proposed flow-based, text-free generative models for supervised image-to-image tasks in medical imaging.

Deep generative models for tabular data (GANs, diffusion models, and LLM-based generators) exhibit highly non-uniform behavior across datasets; the best-performing synthesizer family depends strongly on distributional stressors such as long-tailed marginals, high-cardinality categorical, Zipfian imbalance, and small-sample regimes. This brittleness makes practical deployment challenging, especially when users must balance competing objectives of fidelity, privacy, and utility. We study {intent-conditioned tabular synthesis selection}: given a dataset and a user intent expressed as a preference over evaluation metrics, the goal is to select a synthesizer that minimizes regret relative to an intent-specific oracle. We propose {stress profiling}, a synthesis-specific meta-feature representation that quantifies dataset difficulty along four interpretable stress dimensions, and integrate it into {SYNTHONY}, a selection framework that matches stress profiles against a calibrated capability registry of synthesizer families. Across a benchmark of 7 datasets, 10 synthesizers, and 3 intents, we demonstrate that stress-based meta-features are highly predictive of synthesizer performance: a $k$NN selector using these features achieves strong Top-1 selection accuracy, substantially outperforming zero-shot LLM selectors and random baselines. We analyze the gap between meta-feature-based and capability-based selection, identifying the hand-crafted capability registry as the primary bottleneck and motivating learned capability representations as a direction for future work.

MambaVoiceCloning (MVC) asks whether the conditioning path of diffusion-based TTS can be made fully SSM-only at inference, removing all attention and explicit RNN-style recurrence layers across text, rhythm, and prosody, while preserving or improving quality under controlled conditions. MVC combines a gated bidirectional Mamba text encoder, a Temporal Bi-Mamba supervised by a lightweight alignment teacher discarded after training, and an Expressive Mamba with AdaLN modulation, yielding linear-time O(T) conditioning with bounded activation memory and practical finite look-ahead streaming. Unlike prior Mamba-TTS systems that remain hybrid at inference, MVC removes attention-based duration and style modules under a fixed StyleTTS2 mel-diffusion-vocoder backbone. Trained on LJSpeech/LibriTTS and evaluated on VCTK, CSS10 (ES/DE/FR), and long-form Gutenberg passages, MVC achieves modest but statistically reliable gains over StyleTTS2, VITS, and Mamba-attention hybrids in MOS/CMOS, F0 RMSE, MCD, and WER, while reducing encoder parameters to 21M and improving throughput by 1.6x. Diffusion remains the dominant latency source, but SSM-only conditioning improves memory footprint, stability, and deployability.

Biolinguistics is the interdisciplinary scientific study of the biological foundations, evolution, and genetic basis of human language. It treats language as an innate biological organ or faculty of the mind, rather than a cultural tool, and it challenges a behaviorist conception of human language acquisition as being based on stimulus-response associations. Extracting its most essential component, it takes seriously the idea that mathematical, algebraic models of language capture something natural about the world. The syntactic structure-building operation of MERGE is thought to offer the scientific community a "real joint of nature", "a (new) aspect of nature" (Mukherji 2010), not merely a formal artefact. This mathematical theory of language is then seen as being able to offer biologists, geneticists and neuroscientists clearer instructions for how to explore language. The argument of this chapter proceeds in four steps. First, I clarify the object of inquiry for biolinguistics: not speech, communication, or generic sequence processing, but the internal computational system that generates hierarchically structured expressions. Second, I argue that this formal characterization matters for evolutionary explanation, because different conceptions of syntax imply different standards of what must be explained. Third, I suggest that a sufficiently explicit algebraic account of syntax places non-trivial constraints on candidate neural mechanisms. Finally, I consider how recent neurocomputational work begins to transform these constraints into empirically tractable hypotheses, while also noting the speculative and revisable character of the present program.

We formally introduce a improvisational wordplay game called Connections to explore reasoning capabilities of AI agents. Playing Connections combines skills in knowledge retrieval, summarization and awareness of cognitive states of other agents. We show how the game serves as a good benchmark for social intelligence abilities of language model based agents that go beyond the agents' own memory and deductive reasoning and also involve gauging the understanding capabilities of other agents. Finally, we show how through communication with other agents in a constrained environment, AI agents must demonstrate social awareness and intelligence in games involving collaboration.

We present a data-driven framework for reachability analysis of nonlinear dynamical systems that requires no explicit model. A denoising diffusion probabilistic model learns the time-evolving state distribution of a dynamical system from trajectory data alone. The predicted reachable set takes the form of a sublevel set of a nonconformity score derived from the reconstruction error, with the threshold calibrated via the Learn Then Test procedure so that the probability of excluding a reachable state is bounded with high probability. Experiments on three nonlinear systems, a forced Duffing oscillator, a planar quadrotor, and a high-dimensional reaction-diffusion system, confirm that the empirical miss rate remains below the Probably Approximately Correct (PAC) bound while scaling to state dimensions beyond the reach of classical grid-based and polynomial methods.

Large language models (LLMs) are increasingly embedded in computer science education through AI-assisted programming tools, yet such workflows often exhibit objective drift, in which locally plausible outputs diverge from stated task specifications. Existing instructional responses frequently emphasize tool-specific prompting practices, limiting durability as AI platforms evolve. This paper adopts a human-centered stance, treating human-in-the-loop (HITL) control as a stable educational problem rather than a transitional step toward AI autonomy. Drawing on systems engineering and control-theoretic concepts, we frame objectives and world models as operational artifacts that students configure to stabilize AI-assisted work. We propose a pilot undergraduate CS laboratory curriculum that explicitly separates planning from execution and trains students to specify acceptance criteria and architectural constraints prior to code generation. In selected labs, the curriculum also introduces deliberate, concept-aligned drift to support diagnosis and recovery from specification violations. We report a sensitivity power analysis for a three-arm pilot design comparing unstructured AI use, structured planning, and structured planning with injected drift, establishing detectable effect sizes under realistic section-level constraints. The contribution is a theory-driven, methodologically explicit foundation for HITL pedagogy that renders control competencies teachable across evolving AI tools.

Formal specifications play a central role in ensuring software reliability and correctness. However, automatically synthesizing high-quality formal specifications remains a challenging task, often requiring domain expertise. Recent work has applied large language models to generate specifications in Java Modeling Language (JML), reporting high verification pass rates. But does passing a verifier mean that the specification is actually correct and complete? In this work, we first conduct a comprehensive evaluation comparing classical and prompt-based approaches for automated JML specification synthesis. We then investigate whether prompt optimization can push synthesis quality further by evolving prompts through structured verification feedback. While optimization improves verifier pass rates, we find a clear performance ceiling. More critically, we propose Spec-Harness, an evaluation framework that measures specification correctness and completeness through symbolic verification, revealing that a large fraction of verifier-accepted specifications, including optimized ones, are in fact incorrect or incomplete, over- or under-constraining both inputs and outputs in ways invisible to the verifier. To push beyond this ceiling, we propose VeriAct, a verification-guided agentic framework that iteratively synthesizes and repairs specifications through a closed loop of LLM-driven planning, code execution, verification, and Spec-Harness feedback. Our experiments on two benchmark datasets show that VeriAct outperforms both prompt-based and prompt-optimized baselines, producing specifications that are not only verifiable but also correct and complete.

Vision-Language Models (VLMs) such as CLIP learn a shared embedding space for images and text, yet their representations remain geometrically separated, a phenomenon known as the modality gap. This gap limits tasks requiring cross-modal interchangeability, such as captioning and joint clustering. Existing post-processing approaches can partially improve cross-modal compatibility; however, we show through geometric analysis that they primarily reduce the global centroid offset while leaving the underlying distributional mismatch intact. We decompose the modality gap into a Centroid Gap and a Distribution Gap, and demonstrate that the Distribution Gap is the true predictor of cross-modal task quality ($R^2 = 0.986$), whereas the commonly used Raw Gap is misleading ($R^2 = 0.691$). Motivated by this observation, we propose TPC-CMA (Three-Phase Curriculum for Cross-Modal Alignment), a fine-tuning framework that explicitly reduces both components. The proposed CMA jointly mitigates centroid offsets and reshapes the distributional structure, while a three-phase curriculum with gradient-aware scheduling progressively introduces alignment during training to enable stable optimization. Experiments demonstrate that our method significantly improves cross-modal alignment. With $α_{\text{target}}{=}0.05$, the modality gap is reduced by 66.6\% with only 4.84\% accuracy drop. Under stronger alignment ($α_{\text{target}}{=}0.5$), the gap is reduced by 82.3\%, clustering ARI improves from 0.318 to 0.516, and captioning CIDEr increases by 57.1\% over the original model. Our code and pre-trained models will be made publicly available upon acceptance.

Energy-based models (EBMs) implement inference as gradient descent on a learned Lyapunov function, yielding interpretable, structure-preserving alternatives to black-box neural ODEs and aligning naturally with physical AI. Yet their use in system identification remains limited, and existing architectures lack formal stability guarantees that globally preclude unstable modes. We address this gap by introducing an EBM framework for system identification with stable, dissipative, absorbing invariant dynamics. Unlike classical global Lyapunov stability, absorbing invariance expands the class of stability-preserving architectures, enabling more flexible and expressive EBMs. We extend EBM theory to nonsmooth activations by establishing negative energy dissipation via Clarke derivatives and deriving new conditions for radial unboundedness, exposing a stability-expressivity tradeoff in standard EBMs. To overcome this, we introduce a hybrid architecture with a dynamical visible layer and static hidden layers, prove absorbing invariance under mild assumptions, and show that these guarantees extend to port-Hamiltonian EBMs. Experiments on metric-deformed multi-well and ring systems validate the approach, showcasing how our hybrid EBM architecture combines expressivity with sound and provable safety guarantees by design.

UML state machine design is a critical process in software engineering. Traditionally, state machines are manually crafted by experienced engineers based on natural language requirements-a time-consuming and error-prone procedure. Many automated approaches exist but they require structured NL requirements. In this paper, we investigate the capabilities of current Large Language Models to fully automate UML state machine generation via specialized State Machine Frameworks (SMFs) from non-structured NL requirements. We evaluate two types of state-of-the-art LLMs using single-step and multi-step prompting approaches: a non-reasoning LLM GPT-4o and a reasoning-focused LLM Claude 3.5 Sonnet, and introduce a novel Hybrid Approach that uses the output from a Single-Prompt Baseline as an initial draft state machine, which is then refined through an SMF. In our study, two distinct SMFs are developed based on human approaches: (i) a Structure-Driven SMF, in which state machine components (states, transitions, guards, actions, etc.) are generated in sequential steps, and (ii) an Event-Driven SMF, where identified events iteratively guide state machine construction. Our experiments indicate that while LLMs demonstrate a promising ability to generate state machine models from the Single-Prompt Baseline (e.g., F1-scores of 0.90 for states and 0.75 for transitions using Claude 3.5 Sonnet), their performance is not yet fully sufficient for a fully automated solution (e.g., F1-scores of 0.23 for guards and 0.00 for actions for GPT-4o). Our proposed Hybrid Approach improves the performance of the non-reasoning LLM (GPT-4o) to a similar level as the reasoning LLM (Claude 3.5 Sonnet) but does not further improve the reasoning LLM. Our evaluation highlights both the potential and the limitations of current LLMs for automated state machine design, providing a baseline for future research in this domain.

We propose Question-Asking Navigation (QAsk-Nav), the first reproducible benchmark for Collaborative Instance Object Navigation (CoIN) that enables an explicit, separate assessment of embodied navigation and collaborative question asking. CoIN tasks an embodied agent with reaching a target specified in free-form natural language under partial observability, using only egocentric visual observations and interactive natural-language dialogue with a human, where the dialogue can help to resolve ambiguity among visually similar object instances. Existing CoIN benchmarks are primarily focused on navigation success and offer no support for consistent evaluation of collaborative interaction. To address this limitation, QAsk-Nav provides (i) a lightweight question-asking protocol scored independently of navigation, (ii) an enhanced navigation protocol with realistic, diverse, high-quality target descriptions, and (iii) an open-source dataset, that includes 28,000 quality-checked reasoning and question-asking traces for training and analysis of interactive capabilities of CoIN models. Using the proposed QAsk-Nav benchmark, we develop Light-CoNav, a lightweight unified model for collaborative navigation that is 3x smaller and 70x faster than existing modular methods, while outperforming state-of-the-art CoIN approaches in generalization to unseen objects and environments. Project page at https://benchmarking-interaction.github.io/

The computational cost of stiff chemical kinetics remains a dominant bottleneck in reacting-flow simulation, yet hybrid integration strategies are typically driven by hand-tuned heuristics or supervised predictors that make myopic decisions from instantaneous local state. We introduce a constrained reinforcement learning (RL) framework that autonomously selects between an implicit BDF integrator (CVODE) and a quasi-steady-state (QSS) solver during chemistry integration. Solver selection is cast as a Markov decision process. The agent learns trajectory-aware policies that account for how present solver choices influence downstream error accumulation, while minimizing computational cost under a user-prescribed accuracy tolerance enforced through a Lagrangian reward with online multiplier adaptation. Across sampled 0D homogeneous reactor conditions, the RL-adaptive policy achieves a mean speedup of approximately $3\times$, with speedups ranging from $1.11\times$ to $10.58\times$, while maintaining accurate ignition delays and species profiles for a 106-species \textit{n}-dodecane mechanism and adding approximately $1\%$ inference overhead. Without retraining, the 0D-trained policy transfers to 1D counterflow diffusion flames over strain rates $10$--$2000~\mathrm{s}^{-1}$, delivering consistent $\approx 2.2\times$ speedup relative to CVODE while preserving near-reference temperature accuracy and selecting CVODE at only $12$--$15\%$ of space-time points. Overall, the results demonstrate the potential of the proposed reinforcement learning framework to learn problem-specific integration strategies while respecting accuracy constraints, thereby opening a pathway toward adaptive, self-optimizing workflows for multiphysics systems with spatially heterogeneous stiffness.

Large language models (LLMs) have achieved strong performance on medical question answering (medical QA), and chain-of-thought (CoT) prompting has further improved results by eliciting explicit intermediate reasoning; meanwhile, self-reflective (self-corrective) prompting has been widely claimed to enhance model reliability by prompting LLMs to critique and revise their own reasoning, yet its effectiveness in safety-critical medical settings remains unclear. In this work, we conduct an exploratory analysis of self-reflective reasoning for medical multiple-choice question answering: using GPT-4o and GPT-4o-mini, we compare standard CoT prompting with an iterative self-reflection loop and track how predictions evolve across reflection steps on three widely used medical QA benchmarks (MedQA, HeadQA, and PubMedQA). We analyze whether self-reflection leads to error correction, error persistence, or the introduction of new errors. Our results show that self-reflective prompting does not consistently improve accuracy and its impact is highly dataset- and model-dependent: it yields modest gains on MedQA but provides limited or negative benefits on HeadQA and PubMedQA, and increasing the number of reflection steps does not guarantee better performance. These findings highlight a gap between reasoning transparency and reasoning correctness, suggesting that self-reflective reasoning is better viewed as an analytical tool for understanding model behavior rather than a standalone solution for improving medical QA reliability.

Shuffling strategies for stochastic gradient descent (SGD), including incremental gradient, shuffle-once, and random reshuffling, are supported by rigorous convergence analyses for arbitrary within-epoch permutations. In particular, random reshuffling is known to improve optimization constants relative to cyclic and shuffle-once schemes. However, existing theory offers limited guidance on how to design new data-ordering schemes that further improve optimization constants or stability beyond random reshuffling. In this paper, we design a pipeline using a large language model (LLM)-guided program evolution framework to discover an effective shuffling rule for without-replacement SGD. Abstracting from this instance, we identify two fundamental structural components: block reshuffling and paired reversal. We analyze these components separately and show that block reshuffling strictly reduces prefix-gradient variance constants within the unified shuffling framework, yielding provable improvements over random reshuffling under mild conditions. Separately, we show that paired reversal symmetrizes the epoch map and cancels the leading order-dependent second-order term, reducing order sensitivity from quadratic to cubic in the step size. Numerical experiments with the discovered algorithm validate the theory and demonstrate consistent gains over standard shuffling schemes across convex and nonconvex benchmarks.

Despite growing interest in using Large Language Models (LLMs) for educational assessment, it remains unclear how closely they align with human scoring. We present a systematic evaluation of instruction-tuned LLMs across three open essay-scoring datasets (ASAP 2.0, ELLIPSE, and DREsS) that cover both holistic and analytic scoring. We analyze agreement with human consensus scores, directional bias, and the stability of bias estimates. Our results show that strong open-weight models achieve moderate to high agreement with humans on holistic scoring (Quadratic Weighted Kappa about 0.6), but this does not transfer uniformly to analytic scoring. In particular, we observe large and stable negative directional bias on Lower-Order Concern (LOC) traits, such as Grammar and Conventions, meaning that models often score these traits more harshly than human raters. We also find that concise keyword-based prompts generally outperform longer rubric-style prompts in multi-trait analytic scoring. To quantify the amount of data needed to detect these systematic deviations, we compute the minimum sample size at which a 95% bootstrap confidence interval for the mean bias excludes zero. This analysis shows that LOC bias is often detectable with very small validation sets, whereas Higher-Order Concern (HOC) traits typically require much larger samples. These findings support a bias-correction-first deployment strategy: instead of relying on raw zero-shot scores, systematic score offsets can be estimated and corrected using small human-labeled bias-estimation sets, without requiring large-scale fine-tuning.

While apprenticeship learning has shown promise for inducing effective pedagogical policies directly from student interactions in e-learning environments, most existing approaches rely on optimal or near-optimal expert demonstrations under a fixed reward. Real-world student interactions, however, are often inherently imperfect and evolving: students explore, make errors, revise strategies, and refine their goals as understanding develops. In this work, we argue that imperfect student demonstrations are not noise to be discarded, but structured signals-provided their relative quality is ranked. We introduce HALIDE, Hierarchical Apprenticeship Learning from Imperfect Demonstrations with Evolving Rewards, which not only leverages sub-optimal student demonstrations, but ranks them within a hierarchical learning framework. HALIDE models student behavior at multiple levels of abstraction, enabling inference of higher-level intent and strategy from suboptimal actions while explicitly capturing the temporal evolution of student reward functions. By integrating demonstration quality into hierarchical reward inference,HALIDE distinguishes transient errors from suboptimal strategies and meaningful progress toward higher-level learning goals. Our results show that HALIDE more accurately predicts student pedagogical decisions than approaches that rely on optimal trajectories, fixed rewards, or unranked imperfect demonstrations.

Informed Machine Learning has emerged as a viable generalization of Machine Learning (ML) by building a unified conceptual and algorithmic setting for constructing models on a unified basis of knowledge and data. Physics-informed ML involving physics equations is one of the developments within Informed Machine Learning. This study proposes a novel direction of Knowledge-Data ML, referred to as KD-ML, where numeric data are integrated with knowledge tidbits expressed in the form of granular knowledge landmarks. We advocate that data and knowledge are complementary in several fundamental ways: data are precise (numeric) and local, usually confined to some region of the input space, while knowledge is global and formulated at a higher level of abstraction. The knowledge can be represented as information granules and organized as a collection of input-output information granules called knowledge landmarks. In virtue of this evident complementarity, we develop a comprehensive design process of the KD-ML model and formulate an original augmented loss function L, which additively embraces the component responsible for optimizing the model based on available numeric data, while the second component, playing the role of a granular regularizer, so that it adheres to the granular constraints (knowledge landmarks). We show the role of the hyperparameter positioned in the loss function, which balances the contribution and guiding role of data and knowledge, and point to some essential tendencies associated with the quality of data (noise level) and the level of granularity of the knowledge landmarks. Experiments on two physics-governed benchmarks demonstrate that the proposed KD model consistently outperforms data-driven ML models.

Single-agent large language model (LLM) systems struggle to simultaneously support diverse conversational functions and maintain safety in behavioral health communication. We propose a safety-aware, role-orchestrated multi-agent LLM framework designed to simulate supportive behavioral health dialogue through coordinated, role-differentiated agents. Conversational responsibilities are decomposed across specialized agents, including empathy-focused, action-oriented, and supervisory roles, while a prompt-based controller dynamically activates relevant agents and enforces continuous safety auditing. Using semi-structured interview transcripts from the DAIC-WOZ corpus, we evaluate the framework with scalable proxy metrics capturing structural quality, functional diversity, and computational characteristics. Results illustrate clear role differentiation, coherent inter-agent coordination, and predictable trade-offs between modular orchestration, safety oversight, and response latency when compared to a single-agent baseline. This work emphasizes system design, interpretability, and safety, positioning the framework as a simulation and analysis tool for behavioral health informatics and decision-support research rather than a clinical intervention.

Most automated peer review systems rely on textual manuscript content alone, leaving visual elements such as figures and external scholarly signals underutilized. We introduce REM-CTX, a reinforcement-learning system that incorporates auxiliary context into the review generation process via correspondence-aware reward functions. REM-CTX trains an 8B-parameter language model with Group Relative Policy Optimization (GRPO) and combines a multi-aspect quality reward with two correspondence rewards that explicitly encourage alignment with auxiliary context. Experiments on manuscripts across Computer, Biological, and Physical Sciences show that REM-CTX achieves the highest overall review quality among six baselines, outperforming other systems with substantially larger commercial models, and surpassing the next-best RL baseline across both quality and contextual grounding metrics. Ablation studies confirm that the two correspondence rewards are complementary: each selectively improves its targeted correspondence reward while preserving all quality dimensions, and the full model outperforms all partial variants. Analysis of training dynamics reveals that the criticism aspect is negatively correlated with other metrics during training, suggesting that future studies should group multi-dimension rewards for review generation.

Document retrieval identifies relevant documents but does not provide fine-grained evidence cues, such as specific relevant spans. A possible solution is to apply an LLM after retrieval; however, this introduces significant computational overhead and limits practical deployment. We propose FGR-ColBERT, a modification of ColBERT retrieval model that integrates fine-grained relevance signals distilled from an LLM directly into the retrieval function. Experiments on MS MARCO show that FGR-ColBERT (110M) achieves a token-level F1 of 64.5, exceeding the 62.8 of Gemma 2 (27B), despite being approximately 245 times smaller. At the same time, it preserves retrieval effectiveness (99% relative Recall@50) and remains efficient, incurring only a ~1.12x latency overhead compared to the original ColBERT.

Algorithms for the Multi-Armed Bandit (MAB) problem play a central role in sequential decision-making and have been extensively explored both theoretically and numerically. While most classical approaches aim to identify the arm with the highest expected reward, we focus on a risk-aware setting where the goal is to select the arm with the lowest variance, favoring stability over potentially high but uncertain returns. To model the decision process, we consider a softmax parameterization of the policy; we propose a new algorithm to select the minimal variance (or minimal risk) arm and prove its convergence under natural conditions. The algorithm constructs an unbiased estimate of the objective by using two independent draws from the current's arm distribution. We provide numerical experiments that illustrate the practical behavior of these algorithms and offer guidance on implementation choices. The setting also covers general risk-aware problems where there is a trade-off between maximizing the average reward and minimizing its variance.

The world's 7,000+ languages vary widely in the availability of resources for NLP, motivating efforts to systematically categorize them by their degree of resourcefulness (Joshi et al., 2020). A similar disparity exists among programming languages (PLs); however, no resource-tier taxonomy has been established for code. As large language models (LLMs) grow increasingly capable of generating code, such a taxonomy becomes essential. To fill this gap, we present the first reproducible PL resource classification, grouping 646 languages into four tiers. We show that only 1.9% of languages (Tier 3, High) account for 74.6% of all tokens in seven major corpora, while 71.7% of languages (Tier 0, Scarce) contribute just 1.0%. Statistical analyses of within-tier inequality, dispersion, and distributional skew confirm that this imbalance is both extreme and systematic. Our results provide a principled framework for dataset curation and tier-aware evaluation of multilingual LLMs.

Present practice of deciding on regulation faces numerous problems that make adopted regulations static, unexplained, unduly influenced by powerful interest groups, and stained with a perception of illegitimacy. These well-known problems with the regulatory process can lead to injustice and have substantial negative effects on society and democracy. We discuss a new approach that utilizes distributed artificial intelligence (AI) to make a regulatory recommendation that is explainable and adaptable by design. We outline the main components of a system that can implement this approach and show how it would resolve the problems with the present regulatory system. This approach models and reasons about stakeholder preferences with separate preference models, while it aggregates these preferences in a value sensitive way. Such recommendations can be updated due to changes in facts or in values and are inherently explainable. We suggest how stakeholders can make their preferences known to the system and how they can verify whether they were properly considered in the regulatory decision. The resulting system promises to support regulatory justice, legitimacy, and compliance.

Denoising-based models, including diffusion and flow matching, have led to substantial advances in graph generation. Despite this progress, such models remain constrained by two fundamental limitations: a computational cost that scales quadratically with the number of nodes and a large number of function evaluations required during generation. In this work, we introduce a novel hierarchical generative framework that reduces the number of node pairs that must be evaluated and adopts discrete flow matching to significantly decrease the number of denoising iterations. We empirically demonstrate that our approach more effectively captures graph distributions while substantially reducing generation time.

Long-context decoding in LLMs is IO-bound: each token re-reads an ever-growing KV cache. Prior accelerations cut bytes via compression, which lowers fidelity, or selection/eviction, which restricts what remains accessible, and both can degrade delayed recall and long-form generation. We introduce MAC-Attention, a fidelity- and access-preserving alternative that accelerates decoding by reusing prior attention computations for semantically similar recent queries. It starts with a match stage that performs pre-RoPE L2 matching over a short local window; an amend stage rectifies the reused attention by recomputing a small band near the match boundary; and a complete stage fuses the rectified results with fresh attention computed on the KV tail through a numerically stable merge. On a match hit, the compute and bandwidth complexity is constant regardless of context length. The method is model-agnostic and composes with IO-aware kernels, paged-KV managers, and MQA/GQA. Across LongBench v2 (120K), RULER (120K), and LongGenBench (16K continuous generation), compared to the latest FlashInfer library, MAC-Attention reduces KV accesses by up to 99%, cuts token generation latency by over 60% at 128K, and achieves over 14.3x attention-phase speedups, up to 2.6x end-to-end, while maintaining full-attention quality. By reusing computation, MAC-Attention delivers long-context inference that is both fast and faithful. Code is available here: https://github.com/YJHMITWEB/MAC-Attention.git

Neural collapse (NC) -- the convergence of penultimate-layer features to a simplex equiangular tight frame -- is well understood at equilibrium, but the dynamics governing its onset remain poorly characterised. We identify a simple and predictive regularity: NC occurs when the mean feature norm reaches a model-dataset-specific critical value, fn*, that is largely invariant to training conditions. This value concentrates tightly within each (model, dataset) pair (CV < 8%); training dynamics primarily affect the rate at which fn approaches fn*, rather than the value itself. In standard training trajectories, the crossing of fn below fn* consistently precedes NC onset, providing a practical predictor with a mean lead time of 62 epochs (MAE 24 epochs). A direct intervention experiment confirms fn* is a stable attractor of the gradient flow -- perturbations to feature scale are self-corrected during training, with convergence to the same value regardless of direction (p>0.2). Completing the (architecture)x(dataset) grid reveals the paper's strongest result: ResNet-20 on MNIST gives fn* = 5.867 -- a +458% architecture effect versus only +68% on CIFAR-10. The grid is strongly non-additive; fn* cannot be decomposed into independent architecture and dataset contributions. Four structural regularities emerge: (1) depth has a non-monotonic effect on collapse speed; (2) activation jointly determines both collapse speed and fn*; (3) weight decay defines a three-regime phase diagram -- too little slows, an optimal range is fastest, and too much prevents collapse; (4) width monotonically accelerates collapse while shifting fn* by at most 13%. These results establish feature-norm dynamics as an actionable diagnostic for predicting NC timing, suggesting that norm-threshold behaviour is a general mechanism underlying delayed representational reorganisation in deep networks.

Intrinsic nonlinearity in FPGA-based time-to-digital converters (TDCs) is often treated as a calibration issue and evaluated mainly through post-correction metrics. In quantum key distribution (QKD), however, raw delay-line nonuniformity can affect coincidence timing and thereby influence accidental-coincidence rate and Quantum Bit Error Rate (QBER). This paper analyzes how measured FPGA-TDC nonlinearity propagates to QKD timing metrics using a conservative system-level model that combines random timing uncertainty and deterministic nonlinearity. We also propose fabric-level mitigation strategies based on LUT-assisted delay shaping and placement constraints to reduce severe bin-width irregularities without statistical calibrations. The method is evaluated by reproducing two open-source TDCs implemented on a low-cost Zynq-7000 FPGA. We observe reductions of 14\%-21\% in integral nonlinearity (INL) compared with the non-optimized design, leading to a reduced QBER contribution and an improvement by 3.7\%-14.2\% in the estimated secret fraction. These results suggest that raw FPGA-TDC nonlinearity deserves explicit consideration in timing-sensitive QKD implementations.

Large language models are trained to refuse harmful requests, but can they accurately predict when they will refuse before responding? We investigate this question through a systematic study where models first predict their refusal behavior, then respond in a fresh context. Across 3754 datapoints spanning 300 requests, we evaluate four frontier models: Claude Sonnet 4, Claude Sonnet 4.5, GPT-5.2, and Llama 3.1 405B. Using signal detection theory (SDT), we find that all models exhibit high introspective sensitivity (d' = 2.4-3.5), but sensitivity drops substantially at safety boundaries. We observe generational improvement within Claude (Sonnet 4.5: 95.7 percent accuracy vs Sonnet 4: 93.0 percent), while GPT-5.2 shows lower accuracy (88.9 percent) with more variable behavior. Llama 405B achieves high sensitivity but exhibits strong refusal bias and poor calibration, resulting in lower overall accuracy (80.0 percent). Topic-wise analysis reveals weapons-related queries are consistently hardest for introspection. Critically, confidence scores provide actionable signal: restricting to high-confidence predictions yields 98.3 percent accuracy for well-calibrated models, enabling practical confidence-based routing for safety-critical deployments.

Reverse Kullback-Leibler (RKL) divergence has recently emerged as the preferred objective for large language model (LLM) distillation, consistently outperforming forward KL (FKL), particularly in regimes with large vocabularies and significant teacher-student capacity mismatch, where RKL focuses learning on dominant modes rather than enforcing dense alignment. However, RKL introduces a structural limitation that drives the student toward overconfident predictions. We first provide an analysis of RKL by decomposing its gradients into target and non-target components, and show that non-target gradients consistently push the target logit upward even when the student already matches the teacher, thereby reducing output diversity. In addition, RKL provides weak supervision over non-target classes, leading to poor tail alignment. To address these issues, we propose Diversity-aware RKL (DRKL), which removes this gradient effect and strengthens non-target supervision while preserving the optimization benefits of RKL. Extensive experiments across datasets and model families demonstrate that DRKL consistently outperforms FKL, RKL, and other state-of-the-art distillation objectives, achieving better performance and a superior fidelity-diversity trade-off.

Performance regression testing is essential in large-scale continuous-integration (CI) systems, yet executing full performance suites for every commit is prohibitively expensive. Prior work on performance regression prediction and batch testing has shown independent benefits, but each faces practical limitations: predictive models are rarely integrated into CI decision-making, and conventional batching strategies ignore commit-level heterogeneity. We unify these strands by introducing a risk-aware framework that integrates machine-learned commit risk with adaptive batching. Using Mozilla Firefox as a case study, we construct a production-derived dataset of human-confirmed regressions aligned chronologically with Autoland, and fine-tune ModernBERT, CodeBERT, and LLaMA-3.1 variants to estimate commit-level performance regression risk, achieving up to 0.694 ROC-AUC with CodeBERT. The risk scores drive a family of risk-aware batching strategies, including Risk-Aged Priority Batching and Risk-Adaptive Stream Batching, evaluated through realistic CI simulations. Across thousands of historical Firefox commits, our best overall configuration, Risk-Aged Priority Batching with linear aggregation (RAPB-la), yields a Pareto improvement over Mozilla's production-inspired baseline. RAPB-la reduces total test executions by 32.4%, decreases mean feedback time by 3.8%, maintains mean time-to-culprit at approximately the baseline level, reduces maximum time-to-culprit by 26.2%, and corresponds to an estimated annual infrastructure cost savings of approximately $491K under our cost model. These results demonstrate that risk-aware batch testing can reduce CI resource consumption while improving diagnostic timeliness. To support reproducibility and future research, we release a complete replication package containing all datasets, fine-tuning pipelines, and implementations of our batching algorithms.

Large language models (LLMs) are increasingly deployed in high-stakes settings, yet they frequently violate contextual privacy by disclosing private information in situations where humans would exercise discretion. This raises a fundamental question: do LLMs internally encode contextual privacy norms, and if so, why do violations persist? We present the first systematic study of contextual privacy as a structured latent representation in LLMs, grounded in contextual integrity (CI) theory. Probing multiple models, we find that the three norm-determining CI parameters (information type, recipient, and transmission principle) are encoded as linearly separable and functionally independent directions in activation space. Despite this internal structure, models still leak private information in practice, revealing a clear gap between concept representation and model behavior. To bridge this gap, we introduce CI-parametric steering, which independently intervenes along each CI dimension. This structured control reduces privacy violations more effectively and predictably than monolithic steering. Our results demonstrate that contextual privacy failures arise from misalignment between representation and behavior rather than missing awareness, and that leveraging the compositional structure of CI enables more reliable contextual privacy control, shedding light on potential improvement of contextual privacy understanding in LLMs.

Comparing the internal representations of neural networks is a central goal in both neuroscience and machine learning. Standard alignment metrics operate on raw neural activations, implicitly assuming that similar representations produce similar activity patterns. However, neural systems frequently operate in superposition, encoding more features than they have neurons via linear compression. We derive closed-form expressions showing that superposition systematically deflates Representational Similarity Analysis, Centered Kernel Alignment, and linear regression, causing networks with identical feature content to appear dissimilar. The root cause is that these metrics are dependent on cross-similarity between two systems' respective superposition matrices, which under assumption of random projection usually differ significantly, not on the latent features themselves: alignment scores conflate what a system represents with how it represents it. Under partial feature overlap, this confound can invert the expected ordering, making systems sharing fewer features appear more aligned than systems sharing more. Crucially, the apparent misalignment need not reflect a loss of information; compressed sensing guarantees that the original features remain recoverable from the lower-dimensional activity, provided they are sparse. We therefore argue that comparing neural systems in superposition requires extracting and aligning the underlying features rather than comparing the raw neural mixtures.

In this paper we extend our earlier work of (Rietman et al. 2022) presenting an application of physical Reservoir Computing (RC) to the classification of handwritten and spoken digits. We utilize an unpoled cube of Lead Zirconate Titanate (PZT) as a computational substrate to process these datasets. Our results demonstrate that the PZT reservoir achieves 89.0% accuracy on MNIST handwritten digits, representing a 2.4 percentage point improvement over logistic regression baselines applied to the same preprocessed data. However, for the AudioMNIST spoken digits dataset, the reservoir system (88.2% accuracy) performs equivalently to baseline methods (88.1% accuracy), suggesting that reservoir computing provides the greatest benefits for classification tasks of intermediate difficulty where linear methods underperform but the problem remains learnable. PZT is a well-known material already used in semiconductor applications, presenting a low-power computational substrate that can be integrated with digital algorithms. Our findings indicate that physical reservoirs excel when the task difficulty exceeds the capability of simple linear classifiers but remains within the computational capacity of the reservoir dynamics.

This work introduces an unsupervised Divergence and Aliasing-Free neural network (DAF-FlowNet) for 4D Flow Magnetic Resonance Imaging (4D Flow MRI) that jointly enhances noisy velocity fields and corrects phase wrapping artifacts. DAF-FlowNet parameterizes velocities as the curl of a vector potential, enforcing mass conservation by construction and avoiding explicit divergence-penalty tuning. A cosine data-consistency loss enables simultaneous denoising and unwrapping from wrapped phase images. On synthetic aortic 4D Flow MRI generated from computational fluid dynamics, DAF-FlowNet achieved lower errors than existing techniques (up to 11% lower velocity normalized root mean square error, 11% lower directional error, and 44% lower divergence relative to the best-performing alternative across noise levels), with robustness to moderate segmentation perturbations. For unwrapping, at peak velocity/velocity-encoding ratios of 1.4 and 2.1, DAF-FlowNet achieved 0.18% and 5.2% residual wrapped voxels, representing reductions of 72% and 18% relative to the best alternative method, respectively. In scenarios with both noise and aliasing, the proposed single-stage formulation outperformed a state-of-the-art sequential pipeline (up to 15% lower velocity normalized root mean square error, 11% lower directional error, and 28% lower divergence). Across 10 hypertrophic cardiomyopathy patient datasets, DAF-FlowNet preserved fine-scale flow features, corrected aliased regions, and improved internal flow consistency, as indicated by reduced inter-plane flow bias in aortic and pulmonary mass-conservation analyses recommended by the 4D Flow MRI consensus guidelines. These results support DAF-FlowNet as a framework that unifies velocity enhancement and phase unwrapping to improve the reliability of cardiovascular 4D Flow MRI.

We study offline constrained reinforcement learning from human feedback with multiple preference oracles. Motivated by applications that trade off performance with safety or fairness, we aim to maximize target population utility subject to a minimum protected group welfare constraint. From pairwise comparisons collected under a reference policy, we estimate oracle-specific rewards via maximum likelihood and analyze how statistical uncertainty propagates through the dual program. We cast the constrained objective as a KL-regularized Lagrangian whose primal optimizer is a Gibbs policy, reducing learning to a convex dual problem. We propose a dual-only algorithm that ensures high-probability constraint satisfaction and provide the first finite-sample performance guarantees for offline constrained preference learning. Finally, we extend our theoretical analysis to accommodate multiple constraints and general f-divergence regularization.

The Transformer model architecture has become one of the most widely used in deep learning and the attention mechanism is at its core. The standard attention formulation uses a softmax operation applied to a scaled dot product between query and key vectors. We explore the role played by norms of the queries and keys, which can cause training instabilities when they arbitrarily increase. We demonstrate how this can happen even in simple Transformer models, in the presence of easy-to-learn spurious patterns in the data. We propose a new attention formulation, QUEry-modulated Spherical aTtention (QUEST), that constrains the keys to a hyperspherical latent space, while still allowing individual tokens to flexibly control the sharpness of the attention distribution. QUEST can be easily used as a drop-in replacement for standard attention. We focus on vision applications while also exploring other domains to highlight the method's generality. We show that (1) QUEST trains without instabilities and (2) produces models with improved performance (3) that are robust to data corruptions and adversarial attacks.

We introduce Lévy-Flows, a class of normalizing flow models that replace the standard Gaussian base distribution with Lévy process-based distributions, specifically Variance Gamma (VG) and Normal-Inverse Gaussian (NIG). These distributions naturally capture heavy-tailed behavior while preserving exact likelihood evaluation and efficient reparameterized sampling. We establish theoretical guarantees on tail behavior, showing that for regularly varying bases the tail index is preserved under asymptotically linear flow transformations, and that identity-tail Neural Spline Flow architectures preserve the base distribution's tail shape exactly outside the transformation region. Empirically, we evaluate on S&P 500 daily returns and additional assets, demonstrating substantial improvements in density estimation and risk calibration. VG-based flows reduce test negative log-likelihood by 69% relative to Gaussian flows and achieve exact 95% VaR calibration, while NIG-based flows provide the most accurate Expected Shortfall estimates. These results show that incorporating Lévy process structure into normalizing flows yields significant gains in modeling heavy-tailed data, with applications to financial risk management.

To support practitioners in understanding how agentic systems are designed in real-world industrial practice, we present a review of practitioner conference talks on AI agents. We analyzed 138 recorded talks to examine how companies adopt agent-based architectures (Objective 1), identify recurring architectural strategies and patterns (Objective 2), and analyze application domains and technologies used to implement and operate LLM-driven agentic systems (Objective 3).

Explainable Artificial Intelligence (XAI) is critical for ensuring trust and accountability, yet its development remains predominantly visual. For blind and low-vision (BLV) users, the lack of accessible explanations creates a fundamental barrier to the independent use of AI-driven assistive technologies. This problem intensifies as AI systems shift from single-query tools into autonomous agents that take multi-step actions and make consequential decisions across extended task horizons, where a single undetected error can propagate irreversibly before any feedback is available. This paper investigates the unique XAI requirements of the BLV community through a comprehensive analysis of user interviews and contemporary research. By examining usage patterns across environmental perception and decision support, we identify a significant modality gap. Empirical evidence suggests that while BLV users highly value conversational explanations, they frequently experience "self-blame" for AI failures. The paper concludes with a research agenda for accessible Explainable AI in agentic systems, advocating for multimodal interfaces, blame-aware explanation design, and participatory development.

This paper extends the Acemoglu-Restrepo task exposure framework to address the labor market effects of agentic artificial intelligence systems: autonomous AI agents capable of completing entire occupational workflows rather than discrete tasks. Unlike prior automation technologies that substitute for individual subtasks, agentic AI systems execute end-to-end workflows involving multi-step reasoning, tool invocation, and autonomous decision-making, substantially expanding occupational displacement risk beyond what existing task-level analyses capture. We introduce the Agentic Task Exposure (ATE) score, a composite measure computed algorithmically from O*NET task data using calibrated adoption parameters--not a regression estimate--incorporating AI capability scores, workflow coverage factors, and logistic adoption velocity. Applying the ATE framework across five major US technology regions (Seattle-Tacoma, San Francisco Bay Area, Austin, New York, and Boston) over a 2025-2030 horizon, we find that 93.2% of the 236 analyzed occupations across six information-intensive SOC groups (financial, legal, healthcare, healthcare support, sales, and administrative/clerical) cross the moderate-risk threshold (ATE >= 0.35) in Tier 1 regions by 2030, with credit analysts, judges, and sustainability specialists reaching ATE scores of 0.43-0.47. We simultaneously identify seventeen emerging occupational categories benefiting from reinstatement effects, concentrated in human-AI collaboration, AI governance, and domain-specific AI operations roles. Our findings carry implications for workforce transition policy, regional economic planning, and the temporal dynamics of labor market adjustment

This paper brings up this idea of using Near Field Communication (NFC) for inventory control system instead of using traditional barcodes. NFC because of its high security, ease of use and efficiency can be very suitable for systems like inventory control. In traditional inventory control systems, each product has a barcode pasted on it, which is vulnerable to attacks as barcodes are open and have no security. Furthermore, barcodes are prone to damages and can be unreliable when pasted on different types of products e.g. hot and frozen products, circular shaped products and irregular shaped products like clothes etc. NFC on the other hand is very efficient, secure and reliable when it comes to short-range wireless communication. In this paper we will present our prototype for the inventory control system of an electronic store in which each product has a passive NFC tag pasted to it. When a customer buys a product the receipt of the product is generated using NFC between the NFC passive tag on the product and NFC enabled device (e.g. smart phone or reader) at the cash counter.

We study the finite-time convergence of projected linear two-time-scale stochastic approximation with constant step sizes and Polyak--Ruppert averaging. We establish an explicit mean-square error bound, decomposing it into two interpretable components, an approximation error determined by the constrained subspace and a statistical error decaying at a sublinear rate, with constants expressed through restricted stability margins and a coupling invertibility condition. These constants cleanly separate the effect of subspace choice (approximation errors) from the effect of the averaging horizon (statistical errors). We illustrate our theoretical results through a number of numerical experiments on both synthetic and reinforcement learning problems.

Postural stability during movement is fundamental to independent living, fall prevention, and overall health, particularly among older adults who experience age-related declines in balance, muscle strength, and mobility. Among daily functional activities, the Sit-to-Stand (SiSt) transition is a critical indicator of lower-limb strength, musculoskeletal health, and fall risk, making it an essential parameter for assessing functional capacity and monitoring physical decline in aging populations. This study presents a methodology SiSt transition detection and duration measurement using the Smart Lacelock sensor, a lightweight, shoe-mounted device that integrates a load cell, accelerometer, and gyroscope for motion analysis. The methodology was evaluated in 16 older adults (age: mean: 76.84, SD: 3.45 years) performing SiSt tasks within the Short Physical Performance Battery (SPPB) protocol. Features extracted from multimodal signals were used to train and evaluate four machine learning classifiers using a 4-fold participant-independent cross-validation to classify SiSt transitions and measure their duration. The bagged tree classifier achieved an accuracy of 0.98 and an F1 score of 0.8 in classifying SiSt transition. The mean absolute error in duration measurement of the correctly classified transitions was 0.047, and the SD was 0.07 seconds. These findings highlight the potential of the Smart Lacelock sensor for real-world fall-risk assessment and mobility monitoring in older adults.

This study investigates the relationship between the phonological and morphological structure of Polish words and their meanings using Distributional Semantics. In the present analysis, we ask whether there is a relationship between the form properties of words containing consonant clusters and their meanings. Is the phonological and morphonological structure of complex words mirrored in semantic space? We address these questions for Polish, a language characterized by non-trivial morphology and an impressive inventory of morphologically-motivated consonant clusters. We use statistical and computational techniques, such as t-SNE, Linear Discriminant Analysis and Linear Discriminative Learning, and demonstrate that -- apart from encoding rich morphosyntactic information (e.g. tense, number, case) -- semantic vectors capture information on sub-lexical linguistic units such as phoneme strings. First, phonotactic complexity, morphotactic transparency, and a wide range of morphosyntactic categories available in Polish (case, gender, aspect, tense, number) can be predicted from embeddings without requiring any information about the forms of words. Second, we argue that computational modelling with the discriminative lexicon model using embeddings can provide highly accurate predictions for comprehension and production, exactly because of the existence of extensive information in semantic space that is to a considerable extent isomorphic with structure in the form space.

The rapid development of AI and LLMs has driven new methods of SDLC, in which a large portion of code, technical, and business documentation is generated automatically. However, since there is no single architectural framework that can provide consistent, repeatable transformations across different representation layers of information systems, such systems remain fragmented in their system representation. This study explores the problem of creating a unified architecture for LLM-oriented applications based on selected architectural frameworks by SMEs. A framework structure is proposed that covers some key types of architectural diagrams and supports a closed cycle of transformations, such as: "Code to Documentation to Code". The key architectural diagrams are split equally between main architectural layers: high-layer (business and domain understanding), middle-layer (system architecture), and low-layer (developer-layer architecture). Each architectural layer still contains some abstraction layers, which make it more flexible and better fit the requirements of design principles and architectural patterns. The conducted experiments demonstrated the stable quality of generated documentation and code when using a structured architectural context in the form of architectural diagrams. The results confirm that the proposed unified architecture metamodel can serve as an effective interface between humans and models, improving the accuracy, stability, and repeatability of LLM generation. However, the selected set of architectural diagrams should be optimised to avoid redundancy between some diagrams, and some diagrams should be updated to represent extra contextual orchestration. This work demonstrates measurable improvements for a new generation of intelligent tools that automate the SDLC and enable a comprehensive architecture compatible with AI-driven development.

Autonomous platforms operating in the oceans require accurate navigation to successfully complete their mission. In this regard, the initial heading estimation accuracy and the time required to achieve it play a critical role. The initial heading is traditionally estimated by model-based approaches employing orientation decomposition. However, methods such as the dual vector decomposition and optimized attitude decomposition achieve satisfactory heading accuracy only after long alignment times. To allow rapid and accurate initial heading estimation, we propose an end-to-end, model-free, neural-assisted framework using the same inputs as the model-based approaches. Our proposed approach was trained and evaluated on real-world dataset captured by an autonomous surface vehicle. Our approach shows a significant accuracy improvement over the model-based approaches achieving an average absolute error improvement of 53%. Additionally, our proposed approach was able to reduce the alignment time by up to 67%. Thus, by employing our proposed approach, the reduction in alignment time and improved accuracy allow for a shorter deployment time of an autonomous platform and increased navigation accuracy during the mission.

Automatic program repair can be a challenging task, especially when resolving complex issues at a repository-level, which often involves issue reproduction, fault localization, code repair, testing and validation. Issues of this scale can be commonly found in popular GitHub repositories or datasets that are derived from them. Some repository-level approaches separate localization and repair into distinct phases. Where this is the case, the fault localization approaches vary in terms of the granularity of localization. Where the impact of granularity is explored to some degree for smaller datasets, not all isolate this issue from the separate question of localization accuracy by testing code repair under the assumption of perfect fault localization. To the best of the authors' knowledge, no repository-scale studies have explicitly investigated granularity under this assumption, nor conducted a systematic empirical comparison of granularity levels in isolation. We propose a framework for performing such tests by modifying the localization phase of the Agentless framework to retrieve ground-truth localization data and include this as context in the prompt fed to the repair phase. We show that under this configuration and as a generalization over the SWE-Bench-Mini dataset, function-level granularity yields the highest repair rate against line-level and file-level. However, a deeper dive suggests that the ideal granularity may in fact be task dependent. This study is not intended to improve on the state-of-the-art, nor do we intend for results to be compared against any complete agentic frameworks. Rather, we present a proof of concept for investigating how fault localization may impact automatic code repair in repository-scale scenarios. We present preliminary findings to this end and encourage further research into this relationship between the two phases.

Epileptic seizures are neurological disorders characterized by abnormal and excessive electrical activity in the brain, resulting in recurrent seizure events. Electroencephalogram (EEG) signals are widely used for seizure diagnosis due to their ability to capture temporal and spatial neural dynamics. While recent deep learning methods have achieved high detection accuracy, they often lack interpretability and neurophysiological relevance. This study presents a frequency-aware framework for epileptic seizure detection based on ictal-phase EEG analysis. The raw EEG signals are decomposed into five frequency bands (delta, theta, alpha, lower beta, and higher beta), and eleven discriminative features are extracted from each band. A graph convolutional neural network (GCN) is then employed to model spatial dependencies among EEG electrodes, represented as graph nodes. Experiments on the CHB-MIT scalp EEG dataset demonstrate high detection performance, achieving accuracies of 97.1%, 97.13%, 99.5%, 99.7%, and 51.4% across the respective frequency bands, with an overall broadband accuracy of 99.01%. The results highlight the strong discriminative capability of mid-frequency bands and reveal frequency-specific seizure patterns. The proposed approach improves interpretability and diagnostic precision compared to conventional broadband EEG-based methods.

Distributed applications need identifiers that satisfy storage efficiency, chronological sortability, origin metadata embedding, zero-lookup verifiability, confidentiality for external consumers, and multi-century addressability. Based on our literature survey, no existing scheme provides all six of these identifier properties within a unified system. This paper introduces Source Known Identifiers (SKIDs), a three-tier identity system that projects a single entity identity across trust boundaries, addressing all six properties. The first tier, Source Known ID (SKID), is a 64-bit signed integer embedding a timestamp with a 250-millisecond precision, application topology, and a per-entity-type sequence counter. It serves as the database primary key, providing compact storage (8 bytes) and natural B-tree ordering for optimized database indexing. The second tier, Source Known Entity ID (SKEID), extends the SKID into a 128-bit Universally Unique Identifier (UUID) compatible value by adding an entity type discriminator, an epoch selector, and a BLAKE3 keyed message authentication code (MAC). SKEIDs enable zero-lookup verification of identifier origin, integrity, and entity type within trusted environments, with a big-endian byte layout that preserves chronological ordering in lexicographic UUID string comparisons. The third tier, Secure SKEID, encrypts the entire SKEID using AES-256 symmetric encryption as a single-block pseudorandom permutation, producing ciphertext indistinguishable from random bytes while remaining compatible with standard UUID data-type parsers in string representation. Deterministic bidirectional transformations connect all three tiers.

Tool-integrated LLMs can retrieve, compute, and take real-world actions via external tools, but reliability remains a key bottleneck. We argue that failures stem from both tool-use accuracy (how well an agent invokes a tool) and intrinsic tool accuracy (the tool's own correctness), while most prior work emphasizes the former. We introduce OpenTools, a community-driven toolbox that standardizes tool schemas, provides lightweight plug-and-play wrappers, and evaluates tools with automated test suites and continuous monitoring. We also release a public web demo where users can run predefined agents and tools and contribute test cases, enabling reliability reports to evolve as tools change. OpenTools includes the core framework, an initial tool set, evaluation pipelines, and a contribution protocol. Experiments and evaluations show improved end-to-end reproducibility and task performance; community-contributed, higher-quality task-specific tools deliver 6%-22% relative gains over an existing toolbox across multiple agent architectures on downstream tasks and benchmarks, highlighting the importance of intrinsic tool accuracy.

Production LLM serving often relies on multi-model portfolios spanning a ~530x cost range, where routing decisions trade off quality against cost. This trade-off is non-stationary: providers revise pricing, model quality can regress silently, and new models must be integrated without downtime. We present ParetoBandit, an open-source adaptive router built on cost-aware contextual bandits that is the first to simultaneously enforce dollar-denominated budgets, adapt online to such shifts, and onboard new models at runtime. ParetoBandit closes these gaps through three mechanisms. An online primal-dual budget pacer enforces a per-request cost ceiling over an open-ended stream, replacing offline penalty tuning with closed-loop control. Geometric forgetting on sufficient statistics enables rapid adaptation to price and quality shifts while bootstrapping from offline priors. A hot-swap registry lets operators add or remove models at runtime, with a brief forced-exploration phase for each newcomer, after which UCB selection discovers its quality-cost niche from live traffic alone. We evaluate ParetoBandit across four deployment scenarios on 1,824 prompts routed through a three-model portfolio. Across seven budget ceilings, mean per-request cost never exceeds the target by more than 0.4%. When conditions shift, the system adapts: an order-of-magnitude price cut on the costliest model yields up to +0.071 quality lift, and a silent quality regression is detected and rerouted within budget. A cold-started model reaches meaningful adoption within ~142 steps without breaching the cost ceiling. The router discriminates rather than blindly adopting: expensive models are budget-gated and low-quality models rejected after bounded exploration. End-to-end routing latency is 9.8ms on CPU -- less than 0.4% of typical inference time -- with the routing decision itself taking just 22.5us.

We develop a machine learning (ML) surrogate model to approximate solutions to Maxwell's equations in one dimension, focusing on scenarios involving a material interface that reflects and transmits electro-magnetic waves. Derived from high-fidelity Finite Volume (FV) simulations, our training data includes variations of the initial conditions, as well as variations in one material's speed of light, allowing for the model to learn a range of wave-material interaction behaviors. The ML model autoregressively learns both the physical and frequency embeddings in a vision transformer-based framework. By incorporating Fourier transforms in the latent space, the wave number spectra of the solutions aligns closely with the simulation data. Prediction errors exhibit an approximately linear growth over time with a sharp increase at the material interface. Test results show that the ML solution has adequate relative errors below $10\%$ in over $75$ time step rollouts, despite the presence of the discontinuity and unknown material properties.

Human memory adapts through selective forgetting: experiences become less accessible over time but can be reactivated by reinforcement or contextual cues. In contrast, memory-augmented LLM agents rely on "always-on" retrieval and "flat" memory storage, causing high interference and latency as histories grow. We introduce Oblivion, a memory control framework that casts forgetting as decay-driven reductions in accessibility, not explicit deletion. Oblivion decouples memory control into read and write paths. The read path decides when to consult memory, based on agent uncertainty and memory buffer sufficiency, avoiding redundant always-on access. The write path decides what to strengthen, by reinforcing memories contributing to forming the response. Together, this enables hierarchical memory organization that maintains persistent high-level strategies while dynamically loading details as needed. We evaluate on both static and dynamic long-horizon interaction benchmarks. Results show that Oblivion dynamically adapts memory access and reinforcement, balancing learning and forgetting under shifting contexts, highlighting that memory control is essential for effective LLM-agentic reasoning. The source code is available at https://github.com/nec-research/oblivion.

Chain-of-Thought (CoT) prompting has significantly improved the reasoning capabilities of large language models (LLMs). However, conventional CoT often relies on unstructured, flat reasoning chains that suffer from redundancy and suboptimal performance. In this work, we introduce Hierarchical Chain-of-Thought (Hi-CoT) prompting, a structured reasoning paradigm specifically designed to address the challenges of complex, multi-step reasoning. Hi-CoT decomposes the reasoning process into hierarchical substeps by alternating between instructional planning and step-by-step execution. This decomposition enables LLMs to better manage long reasoning horizons and maintain logical coherence. Extensive evaluations across diverse LLMs and mathematical reasoning benchmarks show that Hi-CoT consistently improves average accuracy by 6.2% (up to 61.4% on certain models and tasks) while reducing reasoning trace length by 13.9% compared to CoT prompting. We further show that accuracy and efficiency are maximized when models strictly adhere to the hierarchical structure. Our code is available at https://github.com/XingshuaiHuang/Hi-CoT.

Software engineering courses often require rapid upskilling in supporting knowledge areas such as domain understanding and modeling methods. We report an experience from a two-week milestone in a master's course where 29 students used a customized ChatGPT (GPT-3.5) tutor grounded in a curated course knowledge base to learn cryptocurrency-finance basics and Domain-Driven Design (DDD). We logged all interactions and evaluated a 34.5% random sample of prompt-answer pairs (60/~174) with a five-dimension rubric (accuracy, relevance, pedagogical value, cognitive load, supportiveness), and we collected pre/post self-efficacy. Responses were consistently accurate and relevant in this setting: accuracy averaged 98.9% with no factual errors and only 2/60 minor inaccuracies, and relevance averaged 92.2%. Pedagogical value was high (89.4%) with generally appropriate cognitive load (82.78%), but supportiveness was low (37.78%). Students reported large pre-post self-efficacy gains for genAI-assisted domain learning and DDD application. From these observations we distill seventeen concrete teaching practices spanning prompt/configuration and course/workflow design (e.g., setting expected granularity, constraining verbosity, curating guardrail examples, adding small credit with a simple quality rubric). Within this single-course context, results suggest that genAI-supported learning can complement instruction in domain understanding and modeling tasks, while leaving room to improve tone and follow-up structure.

Detecting software vulnerabilities is critical to ensuring the security and reliability of modern computer systems. Deep neural networks have shown promising results on vulnerability detection, but they lack the capability to capture global contextual information on vulnerable code. To address this limitation, we explore the application of transformers for C/C++ vulnerability detection. We use program slices that encapsulate key syntactic and semantic features of program code, such as API function calls, array usage, pointer manipulations, and arithmetic expressions. By leveraging transformers' capability to capture both local and global contextual information on vulnerable code, our work can identify vulnerabilities accurately. Combined with data balancing and hyperparameter fine-tuning, our work offers a robust and efficient approach to identifying vulnerable code with moderate resource usage and training time.

Machine learning is increasingly used to improve decisions within branch-and-bound algorithms for mixed-integer programming. Many existing approaches rely on deep learning, which often requires very large training datasets and substantial computational resources for both training and deployment, typically with GPU parallelization. In this work, we take a different path by developing interpretable models that are simple but effective. We focus on approximating strong branching (SB) scores, a highly effective yet computationally expensive branching rule. Using sparse learning methods, we build models with fewer than 4% of the parameters of a state-of-the-art graph neural network (GNN) while achieving competitive accuracy. Relative to SCIP's built-in branching rules and the GNN-based model, our CPU-only models are faster than the default solver and the GPU-accelerated GNN. The models are simple to train and deploy, and they remain effective with small training sets, which makes them practical in low-resource settings. Extensive experiments across diverse problem classes demonstrate the efficiency of this approach.

The field of computer vision has experienced significant advancements through scalable vision encoders and multimodal pre-training frameworks. However, existing approaches often treat vision encoders and large language models (LLMs) as independent modules, limiting the integration of hierarchical visual features. In this work, we propose HIVE (Hierarchical Pre-Training of Vision Encoders), a novel framework that enhances vision-language alignment by introducing hierarchical cross-attention between the vision encoder and LLM. Unlike conventional methods that flatten image embeddings, HIVE enables structured feature fusion across multiple layers, improving gradient flow and representation learning. To optimize this interaction, we introduce a three-stage training strategy that progressively aligns the vision encoder with the LLM, ensuring stable optimization and effective multimodal fusion. Empirical evaluations demonstrate that HIVE achieves superior performance not only in image classification but also on various vision-language tasks, outperforming self-attention-based methods in benchmarks such as MME, GQA, OK-VQA, and ScienceQA. Our results highlight the benefits of hierarchical feature integration, paving the way for more efficient and expressive vision-language models.

Large language models applied to clinical prediction exhibit case-level heterogeneity: simple cases yield consistent outputs, while complex cases produce divergent predictions under minor prompt changes. Existing single-agent strategies sample from one role-conditioned distribution, and multi-agent frameworks use fixed roles with flat majority voting, discarding the diagnostic signal in disagreement. We propose CAMP (Case-Adaptive Multi-agent Panel), where an attending-physician agent dynamically assembles a specialist panel tailored to each case's diagnostic uncertainty. Each specialist evaluates candidates via three-valued voting (KEEP/REFUSE/NEUTRAL), enabling principled abstention outside one's expertise. A hybrid router directs each diagnosis through strong consensus, fallback to the attending physician's judgment, or evidence-based arbitration that weighs argument quality over vote counts. On diagnostic prediction and brief hospital course generation from MIMIC-IV across four LLM backbones, CAMP consistently outperforms strong baselines while consuming fewer tokens than most competing multi-agent methods, with voting records and arbitration traces offering transparent decision audits.

Automatic parallelization remains a challenging problem in software engineering, particularly in identifying code regions where loops can be safely executed in parallel on modern multi-core architectures. Traditional static analysis techniques, such as dependence analysis and polyhedral models, often struggle with irregular or dynamically structured code. In this work, we propose a Transformer-based approach to classify the parallelization potential of source code, focusing on distinguishing independent (parallelizable) loops from undefined ones. We adopt DistilBERT to process source code sequences using subword tokenization, enabling the model to capture contextual syntactic and semantic patterns without handcrafted features. The approach is evaluated on a balanced dataset combining synthetically generated loops and manually annotated real-world code, using 10-fold cross-validation and multiple performance metrics. Results show consistently high performance, with mean accuracy above 99\% and low false positive rates, demonstrating robustness and reliability. Compared to prior token-based methods, the proposed approach simplifies preprocessing while improving generalization and maintaining computational efficiency. These findings highlight the potential of lightweight Transformer models for practical identification of parallelization opportunities at the loop level.

Chain-of-Thought (CoT) monitoring, in which automated systems monitor the CoT of an LLM, is a promising approach for effectively overseeing AI systems. However, the extent to which a model's CoT helps us oversee the model - the monitorability of the CoT - can be affected by training, for instance by the model learning to hide important features of its reasoning. We propose and empirically validate a conceptual framework for predicting when and why this occurs. We model LLM post-training as an RL environment where the reward decomposes into two terms: one term depending on final outputs and another term depending on the CoT. Our framework allows us to classify these two terms as "aligned", "orthogonal", or "in-conflict" before training. We predict that training with in-conflict terms will reduce monitorability, orthogonal terms will not affect it, and aligned terms will improve it. To validate our framework, we use it to classify a set of RL environments, train LLMs within those environments, and evaluate how training affects CoT monitorability. We find that (1) training with "in-conflict" reward terms reduces CoT monitorability and (2) optimizing in-conflict reward terms is difficult.

This study investigates the use of NeuralUCB for cost-aware large language model (LLM) routing. Existing routing approaches can be broadly grouped into supervised routing methods and partial-feedback methods, each with different tradeoffs in efficiency and adaptivity. We implement a NeuralUCB-based routing policy and evaluate it on RouterBench under a simulated online setting. Experimental results show that the proposed method consistently outperforms random and min-cost baselines in utility reward. Compared with the max-quality reference, our method achieves substantially lower inference cost while maintaining competitive reward. These findings suggest that NeuralUCB is a promising approach for cost-aware LLM routing, while also highlighting remaining challenges in action discrimination and exploration.

The pursuit of reducing the memory footprint of the self-attention mechanism in multi-headed self attention (MHA) spawned a rich portfolio of methods, e.g., group-query attention (GQA) and multi-head latent attention (MLA). The methods leverage specialized low-rank factorizations across embedding dimensions or attention heads. From the point of view of classical low-rank approximation, these methods are unconventional and raise questions of which objects they really approximate and how to interpret the low-rank behavior of the resulting representations. To answer these questions, this work proposes a generalized view on the weight objects in the self-attention layer and a factorization strategy, which allows us to construct a parameter efficient scheme, called Tucker Attention. Tucker Attention requires an order of magnitude fewer parameters for comparable validation metrics, compared to GQA and MLA, as evaluated in LLM and ViT test cases. Additionally, Tucker Attention~encompasses GQA, MLA, MHA as special cases and is fully compatible with flash-attention and rotary position embeddings (RoPE). This generalization strategy yields insights of the actual ranks achieved by MHA, GQA, and MLA, and further enables simplifications for MLA.

Human talkers often address listeners with language-comprehension challenges, such as hard-of-hearing or non-native adults, by globally slowing down their speech. However, it remains unclear whether this strategy actually makes speech more intelligible. Here, we take advantage of recent advancements in machine-generated speech allowing more precise control of speech rate in order to systematically examine how targeted speech-rate adjustments may improve comprehension. We first use reverse-correlation experiments to show that the temporal influence of speech rate prior to a target vowel contrast (ex. the tense-lax distinction) in fact manifests in a scissor-like pattern, with opposite effects in early versus late context windows; this pattern is remarkably stable both within individuals and across native L1-English listeners and L2-English listeners with French, Mandarin, and Japanese L1s. Second, we show that this speech rate structure not only facilitates L2 listeners' comprehension of the target vowel contrast, but that native listeners also rely on this pattern in challenging acoustic conditions. Finally, we build a data-driven text-to-speech algorithm that replicates this temporal structure on novel speech sequences. Across a variety of sentences and vowel contrasts, listeners remained unaware that such targeted slowing improved word comprehension. Strikingly, participants instead judged the common strategy of global slowing as clearer, even though it actually increased comprehension errors. Together, these results show that targeted adjustments to speech rate significantly aid intelligibility under challenging conditions, while often going unnoticed. More generally, this paper provides a data-driven methodology to improve the accessibility of machine-generated speech which can be extended to other aspects of speech comprehension and a wide variety of listeners and environments.

Current autonomous AI agents, driven primarily by Large Language Models (LLMs), operate in a state of cognitive weightlessness: they process information without an intrinsic sense of network topology, temporal pacing, or epistemic limits. Consequently, heuristic agentic loops (e.g., ReAct) can exhibit failure modes in interactive environments, including excessive tool use under congestion, prolonged deliberation under time decay, and brittle behavior under ambiguous evidence. In this paper, we propose the Triadic Cognitive Architecture (TCA), a unified mathematical framework that grounds machine reasoning in continuous-time physics. By synthesizing nonlinear filtering theory, Riemannian routing geometry, and optimal control, we formally define the concept of Cognitive Friction. We map the agent's deliberation process to a coupled stochastic control problem where information acquisition is path-dependent and physically constrained. Rather than relying on arbitrary heuristic stop-tokens, the TCA uses an HJB-motivated stopping boundary and instantiates a rollout-based approximation of belief-dependent value-of-information with a net-utility halting condition. Through empirical validation in a simulated Emergency Medical Diagnostic Grid (EMDG), we demonstrate that while greedy baselines over-deliberate under latency and congestion costs, the triadic policy reduces time-to-action while improving patient viability without degrading diagnostic accuracy in this environment.

Modular software deployed on mini compute units in controlled distributed environments often needs two messaging paths: low-overhead in-process coordination and selective cross-node distribution. In practice, event identity, serialization, and transport bridging are frequently implemented as ad hoc glue, which complicates inter-process communication (IPC), structured routing, and shutdown behavior. This paper presents CNS, a lightweight local-first hybrid event fabric centered on asynchronous fire-and-forget messaging. CNS combines a typed event key, per-family serialization and validation, a local publish/subscribe context for in-process coordination, and a NATS-backed distributed context for inter-node distribution. A bridge runtime moves events between the two contexts while preserving a common routing vocabulary. The primary operating model is fire-and-forget publication and subscription; bidirectional request-reply remains available as a secondary extension on the same subject space. A Python prototype and single-machine measurements are reported. Local-only delivery averaged about 30 $μ$s. Distributed-only delivery averaged 1.26-1.37 ms, and the hybrid bridge averaged 1.64-1.89 ms. Validation introduced modest overhead relative to serialization choice. The resulting artifact is suited to structured IPC and practical message movement within modular services and across bounded sets of controlled nodes.

The accelerating displacement of human labor by artificial intelligence (AI) and robotic systems represents a structural transformation whose societal consequences extend far beyond conventional labor market analysis. This paper presents a systematic multi-domain examination of the likely effects on economic structure, psychological well-being, political stability, education, healthcare, and geopolitical order. We identify a critical and underexamined dimension of this transition: the governance gap between nominal human oversight of AI systems -- where humans occupy positions of formal authority over AI decisions -- and genuine human oversight, where those humans possess the cognitive access, technical capability, and institutional authority to meaningfully understand, evaluate, and override AI outputs. We argue that this distinction, largely absent from current governance frameworks including the EU AI Act and NIST AI Risk Management Framework 1.0, represents the primary architectural failure mode in deployed AI governance. The societal consequences of labor displacement intensify this problem by concentrating consequential AI decision-making among an increasingly narrow class of technical and capital actors. We propose five architectural requirements for genuine human oversight systems and characterize the governance window -- estimated at 10-15 years -- before current deployment trajectories risk path-dependent social, economic, and institutional lock-in.

Verifiable claim detection asks whether a claim expresses a factual statement that can, in principle, be assessed against external evidence. As an early filtering stage in automated fact-checking, it plays an important role in reducing the burden on downstream verification components. However, existing approaches to claim detection, whether based on check-worthiness or verifiability, rely solely on the claim text itself. This is a notable limitation for verifiable claim detection in particular, where determining whether a claim is checkable may benefit from knowing what entities and events it refers to and whether relevant information exists to support verification. Inspired by the established role of evidence retrieval in later-stage claim verification, we propose Context-Driven Claim Detection (ContextClaim), a paradigm that advances retrieval to the detection stage. ContextClaim extracts entity mentions from the input claim, retrieves relevant information from Wikipedia as a structured knowledge source, and employs large language models to produce concise contextual summaries for downstream classification. We evaluate ContextClaim on two datasets covering different topics and text genres, the CheckThat! 2022 COVID-19 Twitter dataset and the PoliClaim political debate dataset, across encoder-only and decoder-only models under fine-tuning, zero-shot, and few-shot settings. Results show that context augmentation can improve verifiable claim detection, although its effectiveness varies across domains, model architectures, and learning settings. Through component analysis, human evaluation, and error analysis, we further examine when and why the retrieved context contributes to more reliable verifiability judgments.

This paper introduces a new hybrid framework that combines Reinforcement Learning (RL) and Large Language Models (LLMs) to improve robotic manipulation tasks. By utilizing RL for accurate low-level control and LLMs for high level task planning and understanding of natural language, the proposed framework effectively connects low-level execution with high-level reasoning in robotic systems. This integration allows robots to understand and carry out complex, human-like instructions while adapting to changing environments in real time. The framework is tested in a PyBullet-based simulation environment using the Franka Emika Panda robotic arm, with various manipulation scenarios as benchmarks. The results show a 33.5% decrease in task completion time and enhancements of 18.1% and 36.4% in accuracy and adaptability, respectively, when compared to systems that use only RL. These results underscore the potential of LLM-enhanced robotic systems for practical applications, making them more efficient, adaptable, and capable of interacting with humans. Future research will aim to explore sim-to-real transfer, scalability, and multi-robot systems to further broaden the framework's applicability.

We propose a simple detection mechanism for the Gumbel watermarking scheme proposed by Aaronson (2022). The new mechanism is proven to be near-optimal in a problem-dependent sense among all model-agnostic watermarking schemes under the assumption that the next-token distribution is sampled i.i.d.

AI agents, predominantly powered by large language models (LLMs), are vulnerable to indirect prompt injection, in which malicious instructions embedded in untrusted data can trigger dangerous agent actions. This position paper discusses our vision for system-level defenses against indirect prompt injection attacks. We articulate three positions: (1) dynamic replanning and security policy updates are often necessary for dynamic tasks and realistic environments; (2) certain context-dependent security decisions would still require LLMs (or other learned models), but should only be made within system designs that strictly constrain what the model can observe and decide; (3) in inherently ambiguous cases, personalization and human interaction should be treated as core design considerations. In addition to our main positions, we discuss limitations of existing benchmarks that can create a false sense of utility and security. We also highlight the value of system-level defenses, which serve as the skeleton of agentic systems by structuring and controlling agent behaviors, integrating rule-based and model-based security checks, and enabling more targeted research on model robustness and human interaction.

Microservice architectures form the backbone of modern software systems for their scalability, resilience, and maintainability, but their rise in cloud-native environments raises energy efficiency concerns. While prior research addresses microservice decomposition and placement, the impact of topology, the structural arrangement and interaction pattern among services, on energy efficiency remains largely underexplored. This study quantifies the impact of topologies on energy efficiency and performance across six canonical ones (Sequential Fan-Out, Parallel Fan-Out, Chain, Hierarchical, Probabilistic, Mesh), each instantiated at 5-, 10-, and 20-service scales using the $μ\text{Bench}$ framework. We measure throughput, response time, energy usage, CPU utilization, and failure rates under an identical workload. The results indicate that topology influences the energy efficiency of microservices under the studied conditions. As system size increases, energy consumption grows, with the steepest rise observed in dense Mesh and Chain topologies. Mesh topologies perform worst overall, with low throughput, long response times, and high failure rates. Hierarchical, Chain, and Fan-Out designs balance performance and energy use better. As systems scale, metrics converge, with Probabilistic and Parallel Fan-Out emerging as the most energy-efficient under CPU-bound loads. These results guide greener microservice architecture design and serve as a baseline for future research on workload and deployment impacts.

The Production and Distributed Analysis (PanDA) system, originally developed for the ATLAS experiment at the CERN Large Hadron Collider (LHC), has evolved into a robust platform for orchestrating large-scale workflows across distributed computing resources. Coupled with its intelligent Distributed Dispatch and Scheduling (iDDS) component, PanDA supports AI/ML-driven workflows through a scalable and flexible workflow engine. We present an AI-assisted framework for detector design optimization that integrates multi-objective Bayesian optimization with the PanDA--iDDS workflow engine to coordinate iterative simulations across heterogeneous resources. The framework addresses the challenge of exploring high-dimensional parameter spaces inherent in modern detector design. We demonstrate the framework using benchmark problems and realistic studies of the ePIC and dRICH detectors for the Electron-Ion Collider (EIC). Results show improved automation, scalability, and efficiency in multi-objective optimization. This work establishes a flexible and extensible paradigm for AI-driven detector design and other computationally intensive scientific applications.

Mechanistic interpretability (MI) is an emerging framework for interpreting neural networks. Given a task and model, MI aims to discover a succinct algorithmic process, an interpretation, that explains the model's decision process on that task. However, MI is difficult to scale and generalize. This stems in part from two key challenges: there is no precise notion of a valid interpretation; and, generating interpretations is often an ad hoc process. In this paper, we address these challenges by defining and studying the problem of interpretive equivalence: determining whether two different models share a common interpretation, without requiring an explicit description of what that interpretation is. At the core of our approach, we propose and formalize the principle that two interpretations of a model are equivalent if all of their possible implementations are also equivalent. We develop an algorithm to estimate interpretive equivalence and case study its use on Transformer-based models. To analyze our algorithm, we introduce necessary and sufficient conditions for interpretive equivalence based on models' representation similarity. We provide guarantees that simultaneously relate a model's algorithmic interpretations, circuits, and representations. Our framework lays a foundation for the development of more rigorous evaluation methods of MI and automated, generalizable interpretation discovery methods.

Aggregate programming is a field-based coordination paradigm with over a decade of exploration and successful applications across domains including sensor networks, robotics, and IoT, with implementations in various programming languages, such as Protelis, ScaFi (Scala), and FCPP (C++). A recent research direction integrates machine learning with aggregate computing, aiming to support large-scale distributed learning and provide new abstractions for implementing learning algorithms. However, existing implementations do not target data science practitioners, who predominantly work in Python--the de facto language for data science and machine learning, with a rich and mature ecosystem. Python also offers advantages for other use cases, such as education and robotics (e.g., via ROS). To address this gap, we present Phyelds, a Python library for aggregate programming. Phyelds offers a fully featured yet lightweight implementation of the field calculus model of computation, featuring a Pythonic API and an architecture designed for seamless integration with Python's machine learning ecosystem. We describe the design and implementation of Phyelds and illustrate its versatility across domains, from well-known aggregate computing patterns to federated learning coordination and integration with a widely used multi-agent reinforcement learning simulator.

Analogical reasoning is a key driver of human generalization in problem-solving and argumentation. Yet, analogies between narrative structures remain challenging for machines. Cognitive engines for structural mapping are not directly applicable, as they assume pre-extracted entities, whereas LLMs' performance is sensitive to prompt format and the degree of surface similarity between narratives. This gap motivates a key question: What is the impact of enhancing structural mapping with LLM-derived abstractions on their analogical reasoning ability in narratives? To that end, we propose a modular framework named YARN (Yielding Abstractions for Reasoning in Narratives), which uses LLMs to decompose narratives into units, abstract these units, and then passes them to a mapping component that aligns elements across stories to perform analogical reasoning. We define and operationalize four levels of abstraction that capture both the general meaning of units and their roles in the story, grounded in prior work on framing. Our experiments reveal that abstractions consistently improve model performance, resulting in competitive or better performance than end-to-end LLM baselines. Closer error analysis reveals the remaining challenges in abstraction at the right level, in incorporating implicit causality, and an emerging categorization of analogical patterns in narratives. YARN enables systematic variation of experimental settings to analyze component contributions, and to support future work, we make the code for YARN openly available.

Myopic Optimization with Non-myopic Approval (MONA) mitigates multi-step reward hacking by restricting the agent's planning horizon while supplying far-sighted approval as a training signal~\cite{farquhar2025mona}. The original paper identifies a critical open question: how the method of constructing approval -- particularly the degree to which approval depends on achieved outcomes -- affects whether MONA's safety guarantees hold. We present a reproduction-first extension of the public MONA Camera Dropbox environment that (i)~repackages the released codebase as a standard Python project with scripted PPO training, (ii)~confirms the published contrast between ordinary RL (91.5\% reward-hacking rate) and oracle MONA (0.0\% hacking rate) using the released reference arrays, and (iii)~introduces a modular learned-approval suite spanning oracle, noisy, misspecified, learned, and calibrated approval mechanisms. In reduced-budget pilot sweeps across approval methods, horizons, dataset sizes, and calibration strategies, the best calibrated learned-overseer run achieves zero observed reward hacking but substantially lower intended-behavior rates than oracle MONA (11.9\% vs.\ 99.9\%), consistent with under-optimization rather than re-emergent hacking. These results operationalize the MONA paper's approval-spectrum conjecture as a runnable experimental object and suggest that the central engineering challenge shifts from proving MONA's concept to building learned approval models that preserve sufficient foresight without reopening reward-hacking channels. Code, configurations, and reproduction commands are publicly available. https://github.com/codernate92/mona-camera-dropbox-repro

Cross-validation (CV) is commonly used to estimate predictive risk when independent test data are unavailable. Its validity depends on the assumption that validation tasks are sampled from the same distribution as prediction tasks encountered during deployment. In spatial prediction and other settings with structured data, this assumption is frequently violated, leading to biased estimates of deployment risk. We propose Target-Weighted CV (TWCV), an estimator of deployment risk that accounts for discrepancies between validation and deployment task distributions, thus accounting for (1) covariate shift and (2) task-difficulty shift. We characterize prediction tasks by descriptors such as covariates and spatial configuration. TWCV assigns weights to validation losses such that the weighted empirical distribution of validation tasks matches the corresponding distribution over a target domain. The weights are obtained via calibration weighting, yielding an importance-weighted estimator that targets deployment risk. Since TWCV requires adequate coverage of the deployment distribution's support, we combine it with spatially buffered resampling that diversifies the task difficulty distribution. In a simulation study, conventional as well as spatial estimators exhibit substantial bias depending on sampling, whereas buffered TWCV remains approximately unbiased across scenarios. A case study in environmental pollution mapping further confirms that discrepancies between validation and deployment task distributions can affect performance assessment, and that buffered TWCV better reflects the prediction task over the target domain. These results establish task distribution mismatch as a primary source of CV bias in spatial prediction and show that calibration weighting combined with a suitable validation task generator provides a viable approach to estimating predictive risk under dataset shift.

The proliferation of AI-powered search engines has shifted information discovery from traditional link-based retrieval to direct answer generation with selective source citation, creating new challenges for content visibility. While existing Generative Engine Optimization (GEO) approaches focus primarily on semantic content modification, the role of structural features in influencing citation behavior remains underexplored. In this paper, we propose GEO-SFE, a systematic framework for structural feature engineering in generative engine optimization. Our approach decomposes content structure into three hierarchical levels: macro-structure (document architecture), meso-structure (information chunking), and micro-structure (visual emphasis), and models their impact on citation probability across different generative engine architectures. We develop architecture-aware optimization strategies and predictive models that preserve semantic integrity while improving structural effectiveness. Experimental evaluation across six mainstream generative engines demonstrates consistent improvements in citation rate (17.3 percent) and subjective quality (18.5 percent), validating the effectiveness and generalizability of the proposed framework. This work establishes structural optimization as a foundational component of GEO, providing a data-driven methodology for enhancing content visibility in LLM-powered information ecosystems.

Multimodal deep learning for cancer prognosis is commonly assumed to benefit from synergistic cross-modal interactions, yet this assumption has not been directly tested in survival prediction settings. This work adapts InterSHAP, a Shapley interaction index-based metric, from classification to Cox proportional hazards models and applies it to quantify cross-modal interactions in glioma survival prediction. Using TCGA-GBM and TCGA-LGG data (n=575), we evaluate four fusion architectures combining whole-slide image (WSI) and RNA-seq features. Our central finding is an inverse relationship between predictive performance and measured interaction: architectures achieving superior discrimination (C-index 0.64$\to$0.82) exhibit equivalent or lower cross-modal interaction (4.8\%$\to$3.0\%). Variance decomposition reveals stable additive contributions across all architectures (WSI${\approx}$40\%, RNA${\approx}$55\%, Interaction${\approx}$4\%), indicating that performance gains arise from complementary signal aggregation rather than learned synergy. These findings provide a practical model auditing tool for comparing fusion strategies, reframe the role of architectural complexity in multimodal fusion, and have implications for privacy-preserving federated deployment.

This study explores the use of INT8-based emulation for accelerating traditional FP64-based HPC workloads on modern GPU architectures. Through SCILIB-Accel automatic BLAS offload tool for cache-coherent Unified Memory Architecture, we emulate FP64 matrix multiplications in the LSMS CPU application in the MuST suite without code changes. We find that accuracy depends on both arithmetic precision and the properties of the operator, which can be dealt with through tunable precision emulation. Unlike traditional mixed-precision approaches, this method preserves original algorithms while optimizing hardware utilization. We showcase the potential of improving accuracy and performance at the same time. This work highlights the potential of AI-driven hardware to transform HPC, advocating for adaptive precision strategies in future scientific computing.

Accurate forecasting of air pollution is important for environmental monitoring and policy support, yet data-driven models often suffer from limited generalization in regions with sparse observations. This paper presents Meteorology-Driven GPT for Air Pollution (GPT4AP), a parameter-efficient multi-task forecasting framework based on a pre-trained GPT-2 backbone and Gaussian rank-stabilized low-rank adaptation (rsLoRA). The model freezes the self-attention and feed-forward layers and adapts lightweight positional and output modules, substantially reducing the number of trainable parameters. GPT4AP is evaluated on six real-world air quality monitoring datasets under few-shot, zero-shot, and long-term forecasting settings. In the few-shot regime using 10% of the training data, GPT4AP achieves an average MSE/MAE of 0.686/0.442, outperforming DLinear (0.728/0.530) and ETSformer (0.734/0.505). In zero-shot cross-station transfer, the proposed model attains an average MSE/MAE of 0.529/0.403, demonstrating improved generalization compared with existing baselines. In long-term forecasting with full training data, GPT4AP remains competitive, achieving an average MAE of 0.429, while specialized time-series models show slightly lower errors. These results indicate that GPT4AP provides a data-efficient forecasting approach that performs robustly under limited supervision and domain shift, while maintaining competitive accuracy in data-rich settings.

Scientists often want to explain why an outcome is different in two groups. For instance, differences in patient mortality rates across two hospitals could be due to differences in the patients themselves (covariates) or differences in medical care (outcomes given covariates). The Oaxaca--Blinder decomposition (OBD) is a standard tool to tease apart these factors. It is well known that the OBD requires choosing one of the groups as a reference, and the numerical answer can vary with the reference. To the best of our knowledge, there has not been a systematic investigation into whether the choice of OBD reference can yield different substantive conclusions and how common this issue is. In the present paper, we give existence proofs in real and simulated data that the OBD references can yield substantively different conclusions and that these differences are not entirely driven by model misspecification or small data. We prove that substantively different conclusions occur in up to half of the parameter space, but find these discrepancies rare in the real-data analyses we study. We explain this empirical rarity by examining how realistic data-generating processes can be biased towards parameters that do not change conclusions under the OBD.

Multimodal deep learning has improved prognostic accuracy for brain tumours by integrating histopathology and genomic data, yet the contribution of volumetric MRI within unified survival frameworks remains unexplored. This pilot study extends a bimodal framework by incorporating Fluid Attenuated Inversion Recovery (FLAIR) MRI from BraTS2021 as a third modality. Using the TCGA-GBMLGG cohort (664 patients), we evaluate three unimodal models, nine bimodal configurations, and three trimodal configurations across early, late, and joint fusion strategies. In this small cohort setting, trimodal early fusion achieves an exploratory Composite Score (CS = 0.854), with a controlled $Δ$CS of +0.011 over the bimodal baseline on identical patients, though this difference is not statistically significant (p = 0.250, permutation test). MRI achieves reasonable unimodal discrimination (CS = 0.755) but does not substantially improve bimodal pairs, while providing measurable uplift in the three-way combination. All MRI containing experiments are constrained to 19 test patients, yielding wide bootstrap confidence intervals (e.g. [0.400,1.000]) that preclude definitive conclusions. These findings provide preliminary evidence that a third imaging modality may add prognostic value even with limited sample sizes, and that additional modalities require sufficient multimodal context to contribute effectively.

Recent advances in reasoning Large Language Models (LLMs) have primarily relied on upfront thinking, where reasoning occurs before final answer. However, this approach suffers from critical limitations in code generation, where upfront thinking is often insufficient as problems' full complexity only reveals itself during code implementation. Moreover, it cannot adaptively allocate reasoning effort throughout the code generation process where difficulty varies significantly. In this paper, we propose Think-Anywhere, a novel reasoning mechanism that enables LLMs to invoke thinking on-demand at any token position during code generation. We achieve Think-Anywhere by first teaching LLMs to imitate the reasoning patterns through cold-start training, then leveraging outcome-based RL rewards to drive the model's autonomous exploration of when and where to invoke reasoning. Extensive experiments on four mainstream code generation benchmarks (i.e., LeetCode, LiveCodeBench, HumanEval, and MBPP) show that Think-Anywhere achieves state-of-the-art performance over both existing reasoning methods and recent post-training approaches, while demonstrating consistent generalization across diverse LLMs. Our analysis further reveals that Think-Anywhere enables the model to adaptively invoke reasoning at high-entropy positions, providing enhanced interpretability.

How reliably can structured intent representations preserve user goals across different AI models, languages, and prompting frameworks? Prior work showed that PPS (Prompt Protocol Specification), a 5W3H-based structured intent framework, improves goal alignment in Chinese and generalizes to English and Japanese. This paper extends that line of inquiry in three directions: cross-model robustness across Claude, GPT-4o, and Gemini 2.5 Pro; controlled comparison with CO-STAR and RISEN; and a user study (N=50) of AI-assisted intent expansion in ecologically valid settings. Across 3,240 model outputs (3 languages x 6 conditions x 3 models x 3 domains x 20 tasks), evaluated by an independent judge (DeepSeek-V3), we find that structured prompting substantially reduces cross-language score variance relative to unstructured baselines. The strongest structured conditions reduce cross-language sigma from 0.470 to about 0.020. We also observe a weak-model compensation pattern: the lowest-baseline model (Gemini) shows a much larger D-A gain (+1.006) than the strongest model (Claude, +0.217). Under the current evaluation resolution, 5W3H, CO-STAR, and RISEN achieve similarly high goal-alignment scores, suggesting that dimensional decomposition itself is an important active ingredient. In the user study, AI-expanded 5W3H prompts reduce interaction rounds by 60 percent and increase user satisfaction from 3.16 to 4.04. These findings support the practical value of structured intent representation as a robust, protocol-like communication layer for human-AI interaction.

Effective collaboration requires teams to manage complex cognitive and emotional states through Socially Shared Regulation of Learning (SSRL). Physiological synchrony (i.e., longitudinal alignment in physiological signals) can indicate these states, but is hard to interpret on its own. We investigate the physiological and conversational dynamics of four medical dyads diagnosing a virtual patient case using an intelligent tutoring system. Semantic shifts in dialogue were correlated with transient physiological synchrony peaks. We also coded utterance segments for SSRL and derived cosine similarity using sentence embeddings. The results showed that activating prior knowledge featured significantly lower semantic similarity than simpler task execution. High physiological synchrony was associated with lower semantic similarity, suggesting that such moments involve exploratory and varied language use. Qualitative analysis triangulated these synchrony peaks as ``pivotal moments'': successful teams synchronized during shared discovery, while unsuccessful teams peaked during shared uncertainty. This research advances human-centered AI by demonstrating how biological signals can be fused with dialogues to understand critical moments in problem solving.

Interpretability is central for scientific machine learning, as understanding \emph{why} models make predictions enables hypothesis generation and validation. While tabular foundation models show strong performance, existing explanation methods like SHAP are computationally expensive, limiting interactive exploration. We introduce ShapPFN, a foundation model that integrates Shapley value regression directly into its architecture, producing both predictions and explanations in a single forward pass. On standard benchmarks, ShapPFN achieves competitive performance while producing high-fidelity explanations ($R^2$=0.96, cosine=0.99) over 1000\times faster than KernelSHAP (0.06s vs 610s). Our code is available at https://github.com/kunumi/ShapPFN

Machine-learning-based code vulnerability detection (CVD) has progressed rapidly, from deep program representations to pretrained code models and LLM-centered pipelines. Yet dependable vulnerability labeling remains expensive, noisy, and uneven across projects, languages, and CWE types, motivating approaches that reduce reliance on human labeling. This survey maps these approaches, synthesizing five paradigm families and the mechanisms they use. It connects mechanisms to token, graph, hybrid, and knowledgebased representations, and consolidates evaluation and reporting axes that limit comparison (label-budget specification, compute/cost assumptions, leakage, and granularity mismatches). A Design Map and constraintfirst Decision Guide distill trade-offs and failure modes for practical method selection.

Uncertainty Quantification (UQ) is crucial for ensuring the reliability of automated image segmentations in safety-critical domains like biomedical image analysis or autonomous driving. In segmentation, UQ generates pixel-wise uncertainty scores that must be aggregated into image-level scores for downstream tasks like Out-of-Distribution (OoD) or failure detection. Despite routine use of aggregation strategies, their properties and impact on downstream task performance have not yet been comprehensively studied. Global Average is the default choice, yet it does not account for spatial and structural features of segmentation uncertainty. Alternatives like patch-, class- and threshold-based strategies exist, but lack systematic comparison, leading to inconsistent reporting and unclear best practices. We address this gap by (1) formally analyzing properties, limitations, and pitfalls of common strategies; (2) proposing novel strategies that incorporate spatial uncertainty structure and (3) benchmarking their performance on OoD and failure detection across ten datasets that vary in image geometry and structure. We find that aggregators leveraging spatial structure yield stronger performance in both downstream tasks studied. However, the performance of individual aggregators depends heavily on dataset characteristics, so we (4) propose a meta-aggregator that integrates multiple aggregators and performs robustly across datasets.

This paper investigates sentence-level text reuse in multilingual journalism, analyzing where reused content occurs within articles. We present a weakly supervised method for detecting sentence-level cross-lingual reuse without requiring full translations, designed to support automated pre-selection to reduce information overload for journalists (Holyst et al., 2024). The study compares English-language articles from the Slovenian Press Agency (STA) with reports from 15 foreign agencies (FA) in seven languages, using publication timestamps to retain the earliest likely foreign source for each reused sentence. We analyze 1,037 STA and 237,551 FA articles from two time windows (October 7-November 2, 2023; February 1-28, 2025) and identify 1,087 aligned sentence pairs after filtering to the earliest sources. Reuse occurs in 52% of STA articles and 1.6% of FA articles and is predominantly non-literal, involving paraphrase and compositional reuse from multiple sources. Reused content tends to appear in the middle and end of English articles, while leads are more often original, indicating that simple lexical matching overlooks substantial editorial reuse. Compared with prior work focused on monolingual overlap, we (i) detect reuse across languages without requiring full translation, (ii) use publication timing to identify likely sources, and (iii) analyze where reused material is situated within articles. Dataset and code: https://github.com/kunturs/lrec2026-rewrite-news.

As AI becomes increasingly embedded across societal domains, understanding how future AI practitioners, particularly technology students, perceive its risks is essential for responsible development and adoption. This study analyzed responses from 139 students in Computer Science, Data Science/Data Analytics, and other disciplines using both explicit AI risk ratings and scenario-based assessments of risk and adoption willingness. Four key findings emerged: (1) Students expressed substantially higher concern for concrete, explicitly stated risks than for abstract or scenario-embedded risks; (2) Perceived risk and willingness to adopt AI demonstrated a clear inverse relationship; (3) Although technical education narrowed gender differences in risk awareness, male students reported higher adoption willingness; and (4) A form of "risk underappreciation" was observed, wherein students in AI-related specializations showed both elevated explicit risk awareness and higher willingness to adopt AI, despite lower recognition of risks in applied scenarios. These findings underscore the need for differentiated AI literacy strategies that bridge the gap between awareness and responsible adoption and offer valuable insights for educators, policymakers, industry leaders, and academic institutions aiming to cultivate ethically informed and socially responsible AI practitioners.

As software in industry grows in size and complexity, so does the volume of engineering data that companies generate and use. Ideally, this data could be used for many purposes, including informing decisions on engineering priorities. However, without a structured representation of the links between different aspects of software development, companies can struggle to identify the root causes of deficiencies or anticipate the effects of changes. In this paper, we report on our experience at Bloomberg in developing a novel tool, dubbed BayesInsights, which provides an interactive interface for visualising causal dependencies across various aspects of the software engineering (SE) process using Bayesian Networks (BNs). We describe our journey from defining network structures using a combination of established literature, expert insight, and structure learning algorithms, to integrating BayesInsights into existing data analytics solutions, and conclude with a mixed-methods evaluation of performance benchmarking and survey responses from 24 senior practitioners at Bloomberg. Our results revealed 95.8% of participants found the tool useful for identifying software delivery challenges at the team and organisational levels, cementing its value as a proof of concept for modelling software delivery and developer experience. BayesInsights is currently in preview, with access granted to seven engineering teams and a wider deployment roadmap in place for the future.

Tabular foundation models such as TabPFN and TabICL already produce full predictive distributions yet prevailing regression benchmarks evaluate them almost exclusively via point estimate metrics RMSE R2 These aggregate measures often obscure model performance in the tails of the distribution a critical deficit for high stakes decision making in domains like finance and clinical research where asymmetric risk profiles are the norm We introduce ScoringBench an open benchmark that computes a comprehensive suite of proper scoring rules like CRPS CRLS Interval Score Energy Score weighted CRPS and Brier Score alongside standard point metrics providing a richer picture of probabilistic forecast quality We evaluate realTabPFNv2.5 fine tuned with different scoring rule objectives and TabICL relative to untuned realTabPFNv2.5 across a suite of regression benchmarks Our results confirm that model rankings depend on the chosen scoring rule and that no single pretraining objective is universally optimal This demonstrates that for applications sensitive to extreme events the choice of evaluation metric is as much a domain specific requirement as the data itself ScoringBench is available at https://github.com/jonaslandsgesell/ScoringBench A live preview of the current leaderboard is available at https://scoringbench.bolt.host The leaderboard is maintained via git pull requests to ensure transparency traceability agility and reproducibility

Transferring large volumes of high-resolution images during wind turbine inspections introduces a bottleneck in assessing and detecting severe defects. Efficient coding must preserve high fidelity in blade regions while aggressively compressing the background. In this work, we propose an end-to-end deep learning framework that jointly performs segmentation and dual-mode (lossy and lossless) compression. The segmentation module accurately identifies the blade region, after which our region-of-interest (ROI) compressor encodes it at superior quality compared to the rest of the image. Unlike conventional ROI schemes that merely allocate more bits to salient areas, our framework integrates: (i) a robust segmentation network (BU-Netv2+P) with a CRF-regularized loss for precise blade localization, (ii) a hyperprior-based autoencoder optimized for lossy compression, and (iii) an extended bits-back coder with hierarchical models for fully lossless blade reconstruction. Furthermore, our ROI framework removes the sequential dependency in bits-back coding by reusing background-coded bits, enabling parallelized and efficient dual-mode compression. To the best of our knowledge, this is the first fully integrated learning-based ROI codec combining segmentation, lossy, and lossless compression, ensuring that subsequent defect detection is not compromised. Experiments on a large-scale wind turbine dataset demonstrate superior compression performance and efficiency, offering a practical solution for automated inspections.

Purpose: To investigate whether synthetically generated fractal data can be used to train deep learning (DL) models for dynamic MRI reconstruction, thereby avoiding the privacy, licensing, and availability limitations associated with cardiac MR training datasets. Methods: A training dataset was generated using quaternion Julia fractals to produce 2D+time images. Multi-coil MRI acquisition was simulated to generate paired fully sampled and radially undersampled k-space data. A 3D UNet deep artefact suppression model was trained using these fractal data (F-DL) and compared with an identical model trained on cardiac MRI data (CMR-DL). Both models were evaluated on prospectively acquired radial real-time cardiac MRI from 10 patients. Reconstructions were compared against compressed sensing(CS) and low-rank deep image prior (LR-DIP). All reconstrctuions were ranked for image quality, while ventricular volumes and ejection fraction were compared with reference breath-hold cine MRI. Results: There was no significant difference in qualitative ranking between F-DL and CMR-DL (p=0.9), while both outperformed CS and LR-DIP (p<0.001). Ventricular volumes and function derived from F-DL were similar to CMR-DL, showing no significant bias and accptable limits of agreement compared to reference cine imaging. However, LR-DIP had a signifcant bias (p=0.016) and wider lmits of agreement. Conclusion: DL models trained using synthetic fractal data can reconstruct real-time cardiac MRI with image quality and clinical measurements comparable to models trained on true cardiac MRI data. Fractal training data provide an open, scalable alternative to clinical datasets and may enable development of more generalisable DL reconstruction models for dynamic MRI.

LLM-based coding agents rely on \emph{skills}, pre-packaged instruction sets that extend agent capabilities, yet every token of skill content injected into the context window incurs both monetary cost and attention dilution. To understand the severity of this problem, we conduct a large-scale empirical study of 55,315 publicly available skills and find systemic inefficiencies: 26.4\% lack routing descriptions entirely, over 60\% of body content is non-actionable, and reference files can inject tens of thousands of tokens per invocation. Motivated by these findings, we present \textsc{SkillReducer}, a two-stage optimization framework. Stage~1 optimizes the routing layer by compressing verbose descriptions and generating missing ones via adversarial delta debugging. Stage~2 restructures skill bodies through taxonomy-driven classification and progressive disclosure, separating actionable core rules from supplementary content loaded on demand, validated by faithfulness checks and a self-correcting feedback loop. Evaluated on 600 skills and the SkillsBench benchmark, \textsc{SkillReducer} achieves 48\% description compression and 39\% body compression while improving functional quality by 2.8\%, revealing a \emph{less-is-more} effect where removing non-essential content reduces distraction in the context window. These benefits transfer across five models from four families with a mean retention of 0.965, and generalize to an independent agent framework.

Post-hoc explanation methods are widely used to interpret black-box predictions, but their generation is often computationally expensive and their reliability is not guaranteed. We propose epistemic uncertainty as a low-cost proxy for explanation reliability: high epistemic uncertainty identifies regions where the decision boundary is poorly defined and where explanations become unstable and unfaithful. This insight enables two complementary use cases: `improving worst-case explanations' (routing samples to cheap or expensive XAI methods based on expected explanation reliability), and `recalling high-quality explanations' (deferring explanation generation for uncertain samples under constrained budget). Across four tabular datasets, five diverse architectures, and four XAI methods, we observe a strong negative correlation between epistemic uncertainty and explanation stability. Further analysis shows that epistemic uncertainty distinguishes not only stable from unstable explanations, but also faithful from unfaithful ones. Experiments on image classification confirm that our findings generalize beyond tabular data.

Training relational foundation models requires learning representations that transfer across tasks, yet available supervision is typically limited to a small number of prediction targets per database. This task scarcity causes learned representations to encode task-specific shortcuts that degrade transfer even within the same schema, a problem we call label leakage. We study this using K-Space, a modular architecture combining frozen pretrained tabular encoders with a lightweight message-passing core. To suppress leakage, we introduce a gradient projection method that removes label-predictive directions from representation updates. On RelBench, this improves within-dataset transfer by +0.145 AUROC on average, often recovering near single-task performance. Our results suggest that limited task diversity, not just limited data, constrains relational foundation models.

The currently dominant AI/ML workloads, such as Large Language Models (LLMs), rely on the efficient execution of General Matrix-Matrix Multiplication (GEMM) operations. Thus, most systems are equipped with dedicated matrix hardware accelerators based on square Systolic Arrays (SAs) of Processing Elements (PEs). While this organization was effective for traditional Deep Neural Networks (DNNs), LLMs introduce input-dependent and highly skewed matrices, leading to underutilized SA resources. To address this challenge, we propose SISA (Scale-In Systolic Array), a novel SA architecture that partitions the traditional square array into horizontal rectangular slabs. With minimal overhead, SISA exposes parallelism through independently scheduled slabs for efficient execution of small or skewed matrix shapes, while retaining full-array operation for large GEMMs. SISA achieves up to 8.52x speedup and 93% energy-delay-product (EDP) reduction for representative LLMs compared to a state-of-the-art monolithic SA with the same number of PEs.

Reinforcement Learning (RL) agents often struggle with efficiency and performance in complex environments. We propose a novel framework that uses a Large Language Model (LLM) to dynamically generate a curriculum over available actions, enabling the agent to incorporate each action individually. We apply this framework to the game of Blackjack, where the LLM creates a multi-stage training path that progressively introduces complex actions to a Tabular Q-Learning and a Deep Q-Network (DQN) agent. Our evaluation in a realistic 8-deck simulation over 10 independent runs demonstrates significant performance gains over standard training methods. The curriculum-based approach increases the DQN agent's average win rate from 43.97% to 47.41%, reduces the average bust rate from 32.9% to 28.0%, and accelerates the overall workflow by over 74%, with the agent's full training completing faster than the baseline's evaluation phase alone. These results validate that LLM-guided curricula can build more effective, robust, and efficient RL agents.

Trajectory planning for autonomous driving increasingly leverages large language models (LLMs) for commonsense reasoning, yet LLM outputs are inherently unreliable, posing risks in safety-critical applications. We propose C-TRAIL, a framework built on a Commonsense World that couples LLM-derived commonsense with a trust mechanism to guide trajectory planning. C-TRAIL operates through a closed-loop Recall, Plan, and Update cycle: the Recall module queries an LLM for semantic relations and quantifies their reliability via a dual-trust mechanism; the Plan module injects trust-weighted commonsense into Monte Carlo Tree Search (MCTS) through a Dirichlet trust policy; and the Update module adaptively refines trust scores and policy parameters from environmental feedback. Experiments on four simulated scenarios in Highway-env and two real-world levelXData datasets (highD, rounD) show that C-TRAIL consistently outperforms state-of-the-art baselines, reducing ADE by 40.2%, FDE by 51.7%, and improving SR by 16.9 percentage points on average. The source code is available at https://github.com/ZhihongCui/CTRAIL.

We propose a new frame work of $p$-adic neural network. Unlike the original $p$-adic neural network by S.\ Albeverio, A.\ Khrennikov, and B.\ Tirrozi using a family of characteristic functions indexed by hyperparameters of precision as activation functions, we use a single injective $p$-adic character on the topological Abelian group $\mathbb{Z}_p$ of $p$-adic integers as an activation function. We prove the $p$-adic universal approximation theorem for this formulation of $p$-adic neural network, and reduce it to the feasibility problem of polynomial equations over the finite ring of integers modulo a power of $p$.

Interleaved text-and-image generation represents a significant frontier for Multimodal Large Language Models (MLLMs), offering a more intuitive way to convey complex information. Current paradigms rely on either image generation or retrieval augmentation, yet they typically treat the two as mutually exclusive paths, failing to unify factuality with creativity. We argue that the next milestone in this field is Agentic Tool Planning, where the model serves as a central controller that autonomously determines when, where, and which tools to invoke to produce interleaved responses for visual-critical queries. To systematically evaluate this paradigm, we introduce ATP-Bench, a novel benchmark comprising 7,702 QA pairs (including 1,592 VQA pairs) across eight categories and 25 visual-critical intents, featuring human-verified queries and ground truths. Furthermore, to evaluate agentic planning independent of end-to-end execution and changing tool backends, we propose a Multi-Agent MLLM-as-a-Judge (MAM) system. MAM evaluates tool-call precision, identifies missed opportunities for tool use, and assesses overall response quality without requiring ground-truth references. Our extensive experiments on 10 state-of-the-art MLLMs reveal that models struggle with coherent interleaved planning and exhibit significant variations in tool-use behavior, highlighting substantial room for improvement and providing actionable guidance for advancing interleaved generation. Dataset and code are available at https://github.com/Qwen-Applications/ATP-Bench.

Automated radiology report summarization aims to distill verbose findings into concise clinical impressions, but existing multimodal models often struggle with visual noise and fail to meaningfully improve over strong text-only baselines in the FINDINGS $\to$ IMPRESSION transformation. We challenge two prevailing assumptions: (1) that more visual input is always better, and (2) that multimodal models add limited value when findings already contain rich image-derived detail. Through controlled ablations on MIMIC-CXR benchmark, we show that selectively focusing on pathology-relevant visual patches rather than full images yields substantially better performance. We introduce ViTAS, Visual-Text Attention Summarizer, a multi-stage pipeline that combines ensemble-guided MedSAM2 lung segmentation, bidirectional cross-attention for multi-view fusion, Shapley-guided adaptive patch clustering, and hierarchical visual tokenization feeding a ViT. ViTAS achieves SOTA results with 29.25% BLEU-4 and 69.83% ROUGE-L, improved factual alignment in qualitative analysis, and the highest expert-rated human evaluation scores. Our findings demonstrate that less but more relevant visual input is not only sufficient but superior for multimodal radiology summarization.

FLEURS offers n-way parallel speech for 100+ languages, but Northern Kurdish is not one of them, which limits benchmarking for automatic speech recognition and speech translation tasks in this language. We present FLEURS-Kobani, a Northern Kurdish (ISO 639-3 KMR) spoken extension of the FLEURS benchmark. The FLEURS-Kobani dataset consists of 5,162 validated utterances, totaling 18 hours and 24 minutes. The data were recorded by 31 native speakers. It extends benchmark coverage to an under-resourced Kurdish variety. As baselines, we fine-tuned Whisper v3-large for ASR and E2E S2TT. A two-stage fine-tuning strategy (Common Voice to FLEURS-Kobani) yields the best ASR performance (WER 28.11, CER 9.84 on test). For E2E S2TT (KMR to EN), Whisper achieves 8.68 BLEU on test; we additionally report pivot-derived targets and a cascaded S2TT setup. FLEURS-Kobani provides the first public Northern Kurdish benchmark for evaluation of ASR, S2TT and S2ST tasks. The dataset is publicly released for research use under a CC BY 4.0 license.

In user-agent interaction scenarios such as recommendation, brainstorming, and code suggestion, Large Language Models (LLMs) often generate sets of candidate recommendations where the objective is to maximize the collective utility of the entire set rather than individual candidates independently. However, existing reinforcement learning post-training paradigms, such as Group Relative Policy Optimization (GRPO), typically assign the same set-level scalar reward to every candidate in the set. This leads to noisy training signals where poor candidates free-ride on the high reward produced by a single strong peer, resulting in suboptimal exploration. To address this, we propose Shapley-Enhanced GRPO (ShapE-GRPO). By leveraging the permutation-invariant nature of set-level utility, we derive a Shapley-enhanced formulation from cooperative game theory to decompose set-level rewards into granular, candidate-specific signals. We show that our formulation preserves the fundamental axioms of the Shapley value while remaining computationally efficient with polynomial-time complexity. Empirically, ShapE-GRPO consistently outperforms standard GRPO across diverse datasets with accelerated convergence during training.

Existing mezzanine image codecs lack specialized screen content coding tools and therefore struggle to maintain high image quality under bandwidth constraints, especially in areas with dense text. Although distribution codecs offer advanced screen content compression techniques, their high computational complexity makes them impractical for mezzanine coding. To address this shortfall, we introduce the High-quality Lightweight Codec (HLC), a solution centered on enabling practical, high-throughput palette for mezzanine coding. The core innovation is a novel data-dependency-free palette that eliminates the throughput bottlenecks. To ensure its effectiveness across all content, a co-designed rate-distortion optimization module arbitrates between the palette and traditional prediction modes, while a data reuse strategy between rate estimation and entropy coding minimizes the overall hardware resources required for the system. Experimental results show that, compared with a 4K@120fps JPEG-XS encoder, HLC achieves the same throughput while using only half the LUT resources and delivers BD-PSNR improvements of 3.461dB, 3.299dB, and 5.312dB on gaming, natural, and text content datasets, respectively.

With the ever-growing urgency of sustainability in the economy and society, and the massive stream of information that comes with it, consumers need reliable access to that information. To address this need, companies began publishing so called Environmental, Social, and Governance (ESG) reports, both voluntarily and forced by law. To serve the public, these reports must be addressed not only to financial experts but also to non-expert audiences. But are they written clearly enough? In this work, we extend an existing sentence-level dataset of German ESG reports with crowdsourced readability annotations. We find that, in general, native speakers perceive sentences in ESG reports as easy to read, but also that readability is subjective. We apply various readability scoring methods and evaluate them regarding their prediction error and correlation with human rankings. Our analysis shows that, while LLM prompting has potential for distinguishing clear from hard-to-read sentences, a small finetuned transformer predicts human readability with the lowest error. Averaging predictions of multiple models can slightly improve the performance at the cost of slower inference.

Large language models (LLMs) are increasingly deployed in multi-agent settings where communication must balance informativeness and secrecy. In such settings, an agent may need to signal information to collaborators while preventing an adversary from inferring sensitive details. However, existing LLM benchmarks primarily evaluate capabilities such as reasoning, factual knowledge, or instruction following, and do not directly measure strategic communication under asymmetric information. We introduce SNEAK (Secret-aware Natural language Evaluation for Adversarial Knowledge), a benchmark for evaluating selective information sharing in language models. In SNEAK, a model is given a semantic category, a candidate set of words, and a secret word, and must generate a message that indicates knowledge of the secret without revealing it too clearly. We evaluate generated messages using two simulated agents with different information states: an ally, who knows the secret and must identify the intended message, and a chameleon, who does not know the secret and attempts to infer it from the message. This yields two complementary metrics: utility, measuring how well the message communicates to collaborators, and leakage, measuring how much information it reveals to an adversary. Using this framework, we analyze the trade-off between informativeness and secrecy in modern language models and show that strategic communication under asymmetric information remains a challenging capability for current systems. Notably, human participants outperform all evaluated models by a large margin, achieving up to four times higher scores.

The development of Vision-Language-Action (VLA) models has been significantly accelerated by pre-trained Vision-Language Models (VLMs). However, most existing end-to-end VLAs treat the VLM primarily as a multimodal encoder, directly mapping vision-language features to low-level actions. This paradigm underutilizes the VLM's potential in high-level decision making and introduces training instability, frequently degrading its rich semantic representations. To address these limitations, we introduce DIAL, a framework bridging high-level decision making and low-level motor execution through a differentiable latent intent bottleneck. Specifically, a VLM-based System-2 performs latent world modeling by synthesizing latent visual foresight within the VLM's native feature space; this foresight explicitly encodes intent and serves as the structural bottleneck. A lightweight System-1 policy then decodes this predicted intent together with the current observation into precise robot actions via latent inverse dynamics. To ensure optimization stability, we employ a two-stage training paradigm: a decoupled warmup phase where System-2 learns to predict latent futures while System-1 learns motor control under ground-truth future guidance within a unified feature space, followed by seamless end-to-end joint optimization. This enables action-aware gradients to refine the VLM backbone in a controlled manner, preserving pre-trained knowledge. Extensive experiments on the RoboCasa GR1 Tabletop benchmark show that DIAL establishes a new state-of-the-art, achieving superior performance with 10x fewer demonstrations than prior methods. Furthermore, by leveraging heterogeneous human demonstrations, DIAL learns physically grounded manipulation priors and exhibits robust zero-shot generalization to unseen objects and novel configurations during real-world deployment on a humanoid robot.

Unprecedented visual details of biological structures are being revealed by subcellular-resolution whole-brain 3D microscopy data, enabled by recent advances in intact tissue processing and light-sheet fluorescence microscopy (LSFM). These volumetric data offer rich morphological and spatial cellular information, however, the lack of scalable data processing and analysis methods tailored to these petabyte-scale data poses a substantial challenge for accurate interpretation. Further, existing models for visual tasks such as object detection and classification struggle to generalize to this type of data. To accelerate the development of suitable methods and foundational models, we present CANVAS, a comprehensive set of high-resolution whole mouse brain LSFM benchmark data, encompassing six neuronal and immune cell-type markers, along with cell annotations and a leaderboard. We also demonstrate challenges in generalization of baseline models built on existing architectures, especially due to the heterogeneity in cellular morphology across phenotypes and anatomical locations in the brain. To the best of our knowledge, CANVAS is the first and largest LSFM benchmark that captures intact mouse brain tissue at subcellular level, and includes extensive annotations of cells throughout the brain.

Microscopic road-network weights represent fine-grained, time-varying traffic conditions obtained from individual vehicles. An example is travel speeds associated with road segments as vehicles traverse them. These weights support tasks including traffic microsimulation and vehicle routing with reliability guarantees. We study the problem of time-varying microscopic weight completion. During a time slot, the available weights typically cover only some road segments. Weight completion recovers distributions for the weights of every road segment at the current time slot. This problem involves two challenges: (i) contending with two layers of sparsity, where weights are missing at both the network layer (many road segments lack weights) and the segment layer (a segment may have insufficient weights to enable accurate distribution estimation); and (ii) achieving a weight distribution representation that is closed-form and can capture complex conditions flexibly, including heavy tails and multiple clusters. To address these challenges, we propose DiSGMM that combines sparsity-aware embeddings with spatiotemporal modeling to leverage sparse known weights alongside learned segment properties and long-range correlations for distribution estimation. DiSGMM represents distributions of microscopic weights as learnable Gaussian mixture models, providing closed-form distributions capable of capturing complex conditions flexibly. Experiments on two real-world datasets show that DiSGMM can outperform state-of-the-art methods.

Scientific discovery increasingly depends on high-throughput characterization, yet automation is hindered by proprietary GUIs and the limited generalizability of existing API-based systems. We present Owl-AuraID, a software-hardware collaborative embodied agent system that adopts a GUI-native paradigm to operate instruments through the same interfaces as human experts. Its skill-centric framework integrates Type-1 (GUI operation) and Type-2 (data analysis) skills into end-to-end workflows, connecting physical sample handling with scientific interpretation. Owl-AuraID demonstrates broad coverage across ten categories of precision instruments and diverse workflows, including multimodal spectral analysis, microscopic imaging, and crystallographic analysis, supporting modalities such as FTIR, NMR, AFM, and TGA. Overall, Owl-AuraID provides a practical, extensible foundation for autonomous laboratories and illustrates a path toward evolving laboratory intelligence through reusable operational and analytical skills. The code are available at https://github.com/OpenOwlab/AuraID.

Parameter-efficient fine-tuning methods such as LoRA enable efficient adaptation of large pretrained models but often fall short of full fine-tuning performance. Existing approaches focus on aligning parameter updates, which only indirectly control model predictions. In this work, we introduce the prediction alignment problem, aiming to match the predictor obtained via PEFT to that of full fine-tuning at the level of outputs. We show that this objective naturally leads to a curvature-aware, second-order formulation, where optimal low-rank updates correspond to a Newton-like, curvature-whitened gradient. Based on this insight, we propose Curvature-Guided LoRA (CG-LoRA), which selects and scales adaptation directions using local curvature information. Our method is computationally efficient and avoids explicit second-order matrix construction. Preliminary experiments on standard natural language understanding benchmarks demonstrate improved performance and faster convergence compared to existing LoRA variants.

While clients may join federated learning to improve performance on data they rarely observe locally, they often remain self-interested, expecting the global model to perform well on their own data. This motivates an objective that ensures all clients achieve a similar loss gap -the difference in performance between the global model and the best model they could train using only their local data-. To this end, we propose EAGLE, a novel federated learning algorithm that explicitly regularizes the global model to minimize disparities in loss gaps across clients. Our approach is particularly effective in heterogeneous settings, where the optimal local models of the clients may be misaligned. Unlike existing methods that encourage loss parity, potentially degrading performance for many clients, EAGLE targets fairness in relative improvements. We provide theoretical convergence guarantees for EAGLE under non-convex loss functions, and characterize how its iterates perform relative to the standard federated learning objective using a novel heterogeneity measure. Empirically, we demonstrate that EAGLE reduces the disparity in loss gaps among clients by prioritizing those furthest from their local optimal loss, while maintaining competitive utility in both convex and non-convex cases compared to strong baselines.

The demand for better prediction accuracy and higher execution performance in neural networks continues to grow. The emergence and success of Large Language Models (LLMs) have led to the development of many cloud-based tools for software engineering tasks such as code suggestion. While effective, cloud deployment raises concerns over privacy, latency, and reliance on connectivity. Running LLMs locally on personal devices such as laptops would address these issues by enabling offline use and reducing response time. However, local deployment is challenging: commodity devices lack high-performance accelerators like GPUs and are constrained by limited memory and compute capacity, making it difficult to execute large models efficiently. We present Ditto, a novel method for optimizing both the model size of Code LLMs and their inference programs, particularly for statically-typed programming languages such as C. Our approach integrates two key components: (1) a model compression technique inspired by product quantization, which clusters model parameters into codebooks and quantizes them to lower bit widths while ensuring that outputs remain within a bounded error, as well as synthesizing the inference program for the quantized model; and (2) a compilation pass integrated into LLVM that automatically detects and replaces unoptimized General Matrix-Vector Multiplication (GEMV) operations with implementations from Basic Linear Algebra Subprograms (BLAS) libraries, which are highly optimized for runtime performance. The output of Ditto is an optimized and compiled executable for running selected Code LLMs. We evaluate Ditto on three popular Code LLMs, achieving up to 10.5$\times$ faster inference and 6.4$\times$ lower memory usage compared with their original inference pipeline, while maintaining accuracy close to that of the full-precision models (with an average loss of only 0.27% in pass@1).

Amorphous materials are solids that lack long-range atomic order but possess complex short- and medium-range order. Unlike crystalline materials that can be described by unit cells containing few up to hundreds of atoms, amorphous materials require larger simulation cells with at least hundreds or often thousands of atoms. Inverse design of amorphous materials with probabilistic generative models aims to generate the atomic positions and elements of amorphous materials given a set of desired properties. It has emerged as a promising approach for facilitating the application of amorphous materials in domains such as energy storage and thermal management. In this paper, we introduce AMShortcut, an inference- and training-efficient probabilistic generative model for amorphous materials. AMShortcut enables accurate inference of diverse short- and medium-range structures in amorphous materials with only a few sampling steps, mitigating the need for an excessive number of sampling steps that hinders inference efficiency. AMShortcut can be trained once with all relevant properties and perform inference conditioned on arbitrary combinations of desired properties, mitigating the need for training one model for each combination. Experiments on three amorphous materials datasets with diverse structures and properties demonstrate that AMShortcut achieves its design goals.

Accurate prediction of evacuation behavior is critical for disaster preparedness, yet models trained in one region often fail elsewhere. Using a multi-state hurricane evacuation survey, we show this failure goes beyond feature distribution shift: households with similar characteristics follow systematically different decision patterns across states. As a result, single global models overfit dominant responses, misrepresent vulnerable subpopulations, and generalize poorly across locations. We propose Population-Adaptive Symbolic Mixture-of-Experts (PASM), which pairs large language model guided symbolic regression with a mixture-of-experts architecture. PASM discovers human-readable closed-form decision rules, specializes them to data-driven subpopulations, and routes each input to the appropriate expert at inference time. On Hurricanes Harvey and Irma data, transferring from Florida and Texas to Georgia with 100 calibration samples, PASM achieves a Matthews correlation coefficient of 0.607, compared to XGBoost (0.404), TabPFN (0.333), GPT-5-mini (0.434), and meta-learning baselines MAML and Prototypical Networks (MCC $\leq$ 0.346). The routing mechanism assigns distinct formula archetypes to subpopulations, so the resulting behavioral profiles are directly interpretable. A fairness audit across four demographic axes finds no statistically significant disparities after Bonferroni correction. PASM closes more than half the cross-location generalization gap while keeping decision rules transparent enough for real-world emergency planning.

A key problem in the modern study of AI is predicting and understanding emergent capabilities in models during training. Inspired by methods for studying reactions in quantum chemistry, we present the ``2-datapoint reduced density matrix". We show that this object provides a computationally efficient, unified observable of phase transitions during training. By tracking the eigenvalue statistics of the 2RDM over a sliding window, we derive two complementary signals: the spectral heat capacity, which we prove provides early warning of second-order phase transitions via critical slowing down, and the participation ratio, which reveals the dimensionality of the underlying reorganization. Remarkably, the top eigenvectors of the 2RDM are directly interpretable making it straightforward to study the nature of the transitions. We validate across four distinct settings: deep linear networks, induction head formation, grokking, and emergent misalignment. We then discuss directions for future work using the 2RDM.

This paper introduces ENEIDE (Extracting Named Entities from Italian Digital Editions), a silver standard dataset for Named Entity Recognition and Linking (NERL) in historical Italian texts. The corpus comprises 2,111 documents with over 8,000 entity annotations semi-automatically extracted from two scholarly digital editions: Digital Zibaldone, the philosophical diary of the Italian poet Giacomo Leopardi (1798--1837), and Aldo Moro Digitale, the complete works of the Italian politician Aldo Moro (1916--1978). Annotations cover multiple entity types (person, location, organization, literary work) linked to Wikidata identifiers, including NIL entities that cannot be mapped to the knowledge graph. To the best of our knowledge, ENEIDE represents the first multi-domain, publicly available NERL dataset for historical Italian with training, development, and test splits. We present a methodology for semi-automatic annotations extraction from manually curated scholarly digital editions, including quality control and annotation enhancement procedures. Baseline experiments using state-of-the-art models demonstrate the dataset's challenge for NERL and the gap between zero-shot approaches and fine-tuned models. The dataset's diachronic coverage spanning two centuries makes it particularly suitable for temporal entity disambiguation and cross-domain evaluation. ENEIDE is released under a CC BY-NC-SA 4.0 license.

Multimodal Machine Learning offers a holistic view of a patient's status, integrating structured and unstructured data from electronic health records (EHR). We propose a framework to predict metastasis risk one month prior to diagnosis, using six months of clinical history from EHR data. Data from four cancer cohorts collected at Karolinska University Hospital (Stockholm, Sweden) were analyzed: breast (n = 743), colon (n = 387), lung (n = 870), and prostate (n = 1890). The dataset included demographics, comorbidities, laboratory results, medications, and clinical text. We compared traditional and deep learning classifiers across single modalities and multimodal combinations, using various fusion strategies and a Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) 2a design, with an 80-20 development-validation split to ensure a rigorous, repeatable evaluation. Performance was evaluated using AUROC, AUPRC, F1 score, sensitivity, and specificity. We then employed a multimodal adaptation of SHAP to analyze the classifiers' reasoning. Intermediate fusion achieved the highest F1 scores on breast (0.845), colon (0.786), and prostate cancer (0.845), demonstrating strong predictive performance. For lung cancer, the intermediate fusion achieved an F1 score of 0.819, while the text-only model achieved the highest, with an F1 score of 0.829. Deep learning classifiers consistently outperformed traditional models. Colon cancer, the smallest cohort, had the lowest performance, highlighting the importance of sufficient training data. SHAP analysis showed that the relative importance of modalities varied across cancer types. Fusion strategies offer distinct strengths and weaknesses. Intermediate fusion consistently delivered the best results, but strategy choices should align with data characteristics and organizational needs.

Although many AI applications of interest require specialized multi-modal models, relevant data to train such models is inherently scarce or inaccessible. Filling these gaps with human annotators is prohibitively expensive, error-prone, and time-consuming, leading model builders to increasingly consider synthetic data as a scalable alternative. However, existing synthetic data generation methods often rely on manual prompts, evolutionary algorithms, or extensive seed data from the target distribution - limiting their scalability, explainability, and control. In this paper, we introduce Simula: a novel reasoning-driven framework for data generation and evaluation. It employs a seedless, agentic approach to generate synthetic datasets at scale, allowing users to define desired dataset characteristics through an explainable and controllable process that enables fine-grained resource allocation. We show the efficacy of our approach on a variety of datasets, rigorously testing both intrinsic and downstream properties. Our work (1) offers guidelines for synthetic data mechanism design, (2) provides insights into generating and evaluating synthetic data at scale, and (3) unlocks new opportunities for developing and deploying AI in domains where data scarcity or privacy concerns are paramount.

Datacenters are vital to our digital society, but consume a considerable fraction of global electricity and demand is projected to increase. To improve their sustainability and performance, we envision that simulators will become primary decision-making tools. However, and unlike other fields focusing on key societal infrastructure such as waterworks and mass transit, datacenter simulators do not yet combine multiple independent models into their operation and thus suffer from issues associated with singular models, such as specialization, and lack of adaptability to operational phenomena. To address this challenge, we propose M3SA, a datacenter simulation and analysis framework that uses discrete-event simulation to predict, for each model, the impact on climate and performance under various realistic datacenter conditions, and then combines these predictions. We design an architecture for simulating multiple concurrent models (Multi-Model), a technique for integrating the results of multiple models into a Meta-Model, and a procedure for quantifying Meta-Model accuracy. Through experiments with an M3SA prototype, we show that (i) M3SA can reproduce and enhance peer-reviewed experiments; (ii) M3SA can predict operational phenomena (e.g., failures) of datacenters, running fundamentally different workload traces; (iii) M3SA enables various types of what-if and how-to analysis, such as how to configure CO2-aware migration over yearly energy-production patterns. M3SA has been integrated into the open-source simulator OpenDC and is available at: https://github.com/atlarge-research/opendc-m3sa.

Public spaces such as transport hubs, city centres, and event venues require timely and reliable detection of potentially violent behaviour to support public safety. While automated video analysis has made significant progress, practical deployment remains constrained by latency, privacy, and resource limitations, particularly under edge-computing conditions. This paper presents the design and demonstrator-based deployment of a hybrid edge-based action detection system that combines skeleton-based motion analysis with vision-language models for semantic scene interpretation. Skeleton-based processing enables continuous, privacy-aware monitoring with low computational overhead, while vision-language models provide contextual understanding and zero-shot reasoning capabilities for complex and previously unseen situations. Rather than proposing new recognition models, the contribution focuses on a system-level comparison of both paradigms under realistic edge constraints. The system is implemented on a GPU-enabled edge device and evaluated with respect to latency, resource usage, and operational trade-offs using a demonstrator-based setup. The results highlight the complementary strengths and limitations of motioncentric and semantic approaches and motivate a hybrid architecture that selectively augments fast skeletonbased detection with higher-level semantic reasoning. The presented system provides a practical foundation for privacy-aware, real-time video analysis in public safety applications.

There has been growing interest in building agents that can interact with digital platforms to execute meaningful enterprise tasks autonomously. Among the approaches explored are tool-augmented agents built on abstractions such as Model Context Protocol (MCP) and web agents that operate through graphical interfaces. Yet, it remains unclear whether such complex agentic systems are necessary given their cost and operational overhead. We argue that a coding agent equipped only with a terminal and a filesystem can solve many enterprise tasks more effectively by interacting directly with platform APIs. We evaluate this hypothesis across diverse real-world systems and show that these low-level terminal agents match or outperform more complex agent architectures. Our findings suggest that simple programmatic interfaces, combined with strong foundation models, are sufficient for practical enterprise automation.

Metaheuristic algorithms such as Particle Swarm Optimization (PSO) and Evolutionary Algorithms (EA) excel at exploring solution spaces but lack mechanisms to accumulate and reuse procedural knowledge from successful search trajectories. This paper proposes Associative Constructive Evolution (ACE), a framework that enhances metaheuristics through learned generative guidance. ACE introduces a Generative Construction Automaton (GCA) -- a probabilistic model over operation sequences -- coupled with the base metaheuristic in a synergistic loop: the metaheuristic explores and provides trajectory samples, while the GCA consolidates successful patterns and guides future exploration. Three mechanisms realize this cooperation: Hebbian weight consolidation that strengthens associations between co-successful operations, guided sampling that biases search toward learned high-quality regions, and symbolic abstraction that extracts frequent patterns into reusable macro-operations. Experiments integrating ACE with EA and PSO on molecular design and maze navigation demonstrate consistent improvements. ACE-PSO achieves a 27.5% increase in success rate while reducing convergence time by 49.6%. In molecular design, ACE-EA improves fitness by 10.1% with 126 chemically interpretable macro-operations automatically discovered.

With the development of foundational models, model compression has become a critical requirement. Various model compression approaches have been proposed such as low-rank decomposition, pruning, quantization, ergodic dynamic systems, and knowledge distillation, which are based on different heuristics. To elevate the field from fragmentation to a principled discipline, we construct a unifying mathematical framework for model compression grounded in measure theory. We further demonstrate that each model compression technique is mathematically equivalent to a neural network subject to a regularization. Building upon this mathematical and structural equivalence, we propose an experimentally-verified data-free model compression framework, termed \textit{Big2Small}, which translates Implicit Neural Representations (INRs) from data domain to the domain of network parameters. \textit{Big2Small} trains compact INRs to encode the weights of larger models and reconstruct the weights during inference. To enhance reconstruction fidelity, we introduce Outlier-Aware Preprocessing to handle extreme weight values and a Frequency-Aware Loss function to preserve high-frequency details. Experiments on image classification and segmentation demonstrate that \textit{Big2Small} achieves competitive accuracy and compression ratios compared to state-of-the-art baselines.

Large Language Models (LLMs) have achieved remarkable performance on a wide range of specialized tasks, exhibiting strong problem-solving capabilities. However, training these models is prohibitively expensive, and they often lack domain-specific expertise because they rely on general knowledge datasets. Expertise finetuning can address this issue; however, it often leads to overspecialization, and developing a single multi-domain expert remains difficult due to diverging objectives. Furthermore, multitask training is challenging due to interference and catastrophic forgetting. Existing work proposes combining the expertise of dense models within a Mixture of Experts (MoE) architecture, although this approach still requires multitask finetuning. To address these issues, we introduce Dynamic Upcycling MoE (DUME), a novel approach that reuses dense experts trained on different domains to construct a unified MoE model. Our method builds a single multitask model that preserves the capabilities of the original dense experts without requiring additional training. DUME is both cost-efficient and scalable: by leveraging the closed-form solution of ridge regression, it eliminates the need for further optimization and enables experts to be added dynamically while maintaining the model's original performance. We demonstrate that DUME consistently outperforms baseline approaches in both causal language modeling and reasoning settings. Finally, we also show that the DUME model can be fine-tuned to further improve performance. We show that, in the causal language modeling setting, DUME can retain up to 97.6% of a dense expert model specialized in one particular domain, and that it can also surpass it in the reasoning setting, where it can achieve 102.1% of the dense expert performance. Our code is available at: github.com/gensyn-ai/dume.

A human-like chess engine should mimic the style, errors, and consistency of a strong human player rather than maximize playing strength. We show that training from move sequences alone forces a model to learn two capabilities: state tracking, which reconstructs the board from move history, and decision quality, which selects good moves from that reconstructed state. These impose contradictory data requirements: low-rated games provide the diversity needed for tracking, while high-rated games provide the quality signal for decision learning. Removing low-rated data degrades performance. We formalize this tension as a dual-capability bottleneck, P <= min(T,Q), where overall performance is limited by the weaker capability. Guided by this view, we scale the model from 28M to 120M parameters to improve tracking, then introduce Elo-weighted training to improve decisions while preserving diversity. A 2 x 2 factorial ablation shows that scaling improves tracking, weighting improves decisions, and their combination is superadditive. Linear weighting works best, while overly aggressive weighting harms tracking despite lower validation loss. We also introduce a coverage-decay formula, t* = log(N/kcrit)/log b, as a reliability horizon for intra-game degeneration risk. Our final 120M-parameter model, without search, reached Lichess bullet 2570 over 253 rated games. On human move prediction it achieves 55.2% Top-1 accuracy, exceeding Maia-2 rapid and Maia-2 blitz. Unlike position-based methods, sequence input naturally encodes full game history, enabling history-dependent decisions that single-position models cannot exhibit.

Recent advances in vision-language models (VLMs) have accelerated their application to indoor safety hazards assessment. However, existing benchmarks suffer from three fundamental limitations: (1) heavy reliance on synthetic datasets constructed via simulation software, creating a significant domain gap with real-world environments; (2) oversimplified safety tasks with artificial constraints on hazard and scene types, thereby limiting model generalization; and (3) absence of rigorous evaluation protocols to thoroughly assess model capabilities in complex home safety scenarios. To address these challenges, we introduce TSHA (\textbf{T}rustworthy \textbf{S}afety \textbf{H}azards \textbf{A}ssessment), a comprehensive benchmark comprising 81,809 carefully curated training samples drawn from four complementary sources: existing indoor datasets, internet images, AIGC images, and newly captured images. This benchmark set also includes a highly challenging test set with 1707 samples, comprising not only a carefully selected subset from the training distribution but also newly added videos and panoramic images containing multiple safety hazards, used to evaluate the model's robustness in complex safety scenarios. Extensive experiments on 23 popular VLMs demonstrate that current VLMs lack robust capabilities for safety hazard assessment. Importantly, models trained on the TSHA training set not only achieve a significant performance improvement of up to +18.3 points on the TSHA test set but also exhibit enhanced generalizability across other benchmarks, underscoring the substantial contribution and importance of the TSHA benchmark.

We address the challenge of adapting pre-trained Large Language Models (LLMs) for multivariate time-series analysis, where their deployment is often hindered by prohibitive computational and memory demands. Our solution, One-for-All, introduces Gaussian Rank-Stabilized Low-Rank Adapters (rsLoRA) to enable parameter-efficient fine-tuning of frozen LLMs. While inspired by LoRA, rsLoRA introduces a mathematically grounded rank-stabilization mechanism that enables provable gradient stability at low ranks a novel contribution absent in prior PEFT methods. Our framework injects trainable rank decomposition matrices (rank 16) into positional embeddings and output layers, while keeping self-attention weights fixed. This design reduces trainable parameters by 6.8$\times$ (vs. TimesNet), 21$\times$ (vs. GPT4TS), and 11.8$\times$ (vs. TIME-LLM), while achieving a 168-1,776$\times$ smaller memory footprint (2.2MiB vs. 340MiB-4.18GiB in SOTA models). Rigorous evaluation across six time-series tasks demonstrates that One-for-All achieves state-of-the-art efficiency-accuracy trade-offs: 5.5$\times$ higher parameter efficiency (MSE=5.50) than TimesNet and 21$\times$ better than GPT4TS, while matching their forecasting accuracy (MSE=0.33). The framework's stability is validated through consistent performance across diverse horizons (96-720 steps) and datasets (ETT, Weather, M3, M4), with 98.3% fewer parameters than conventional transformers. These advances enable deployment on edge devices for healthcare, finance, and environmental monitoring without compromising performance.

Can classifier-based safety gates maintain reliable oversight as AI systems improve over hundreds of iterations? We provide comprehensive empirical evidence that they cannot. On a self-improving neural controller (d=240), eighteen classifier configurations -- spanning MLPs, SVMs, random forests, k-NN, Bayesian classifiers, and deep networks -- all fail the dual conditions for safe self-improvement. Three safe RL baselines (CPO, Lyapunov, safety shielding) also fail. Results extend to MuJoCo benchmarks (Reacher-v4 d=496, Swimmer-v4 d=1408, HalfCheetah-v4 d=1824). At controlled distribution separations up to delta_s=2.0, all classifiers still fail -- including the NP-optimal test and MLPs with 100% training accuracy -- demonstrating structural impossibility. We then show the impossibility is specific to classification, not to safe self-improvement itself. A Lipschitz ball verifier achieves zero false accepts across dimensions d in {84, 240, 768, 2688, 5760, 9984, 17408} using provable analytical bounds (unconditional delta=0). Ball chaining enables unbounded parameter-space traversal: on MuJoCo Reacher-v4, 10 chains yield +4.31 reward improvement with delta=0; on Qwen2.5-7B-Instruct during LoRA fine-tuning, 42 chain transitions traverse 234x the single-ball radius with zero safety violations across 200 steps. A 50-prompt oracle confirms oracle-agnosticity. Compositional per-group verification enables radii up to 37x larger than full-network balls. At d<=17408, delta=0 is unconditional; at LLM scale, conditional on estimated Lipschitz constants.

Modern manufacturing environments demand real-time, trustworthy, and interpretable root-cause insights to sustain productivity and quality. Traditional analytics pipelines often treat anomaly detection, causal inference, and root-cause analysis as isolated stages, limiting scalability and explainability. In this work, we present CausalPulse, an industry-grade multi-agent copilot that automates causal diagnostics in smart manufacturing. It unifies anomaly detection, causal discovery, and reasoning through a neurosymbolic architecture built on standardized agentic protocols. CausalPulse is being deployed in a Robert Bosch manufacturing plant, integrating seamlessly with existing monitoring workflows and supporting real-time operation at production scale. Evaluations on both public (Future Factories) and proprietary (Planar Sensor Element) datasets show high reliability, achieving overall success rates of 98.0% and 98.73%. Per-criterion success rates reached 98.75% for planning and tool use, 97.3% for self-reflection, and 99.2% for collaboration. Runtime experiments report end-to-end latency of 50-60s per diagnostic workflow with near-linear scalability (R^2=0.97), confirming real-time readiness. Comparison with existing industrial copilots highlights distinct advantages in modularity, extensibility, and deployment maturity. These results demonstrate how CausalPulse's modular, human-in-the-loop design enables reliable, interpretable, and production-ready automation for next-generation manufacturing.

Accurate and complete multi-modal Magnetic Resonance Imaging (MRI) is essential for neuro-oncological assessment, as each contrast provides complementary anatomical and pathological information. However, acquiring all modalities (e.g., T1c, T1n, T2, T2f) for every patient is often impractical due to time, cost, and patient discomfort, potentially limiting comprehensive tumour evaluation. We propose 3D-MC-SAGAN (3D Multi-Contrast Self-Attention generative adversarial network), a unified 3D multi-contrast synthesis framework that generates high-fidelity missing modalities from a single T2 input while explicitly preserving tumour characteristics. The model employs a multi-scale 3D encoder-decoder generator with residual connections and a novel Memory-Bounded Hybrid Attention (MBHA) block to capture long-range dependencies efficiently, and is trained with a WGAN-GP critic and an auxiliary contrast-conditioning branch to produce T2f, T1n, and T1c volumes within a single unified network. A frozen 3D U-Net-based segmentation module introduces a segmentation-consistency constraint to preserve lesion morphology. The composite objective integrates adversarial, reconstruction, perceptual, structural similarity, contrast-classification, and segmentation-guided losses to align global realism with tumour-preserving structure. Extensive evaluation on 3D brain MRI datasets demonstrates that 3D-MC-SAGAN achieves state-of-the-art quantitative performance and generates visually coherent, anatomically plausible contrasts with improved distribution-level realism. Moreover, it maintains tumour segmentation accuracy comparable to fully acquired multi-modal inputs, highlighting its potential to reduce acquisition burden while preserving clinically meaningful information.

Kazantzis-Kravaris/Luenberger (KKL) observers are a class of state observers for nonlinear systems that rely on an injective map to transform the nonlinear dynamics into a stable quasi-linear latent space, from where the state estimate is obtained in the original coordinates via a left inverse of the transformation map. Current learning-based methods for these maps are designed exclusively for autonomous systems and do not generalize well to controlled or non-autonomous systems. In this paper, we propose two learning-based designs of neural KKL observers for non-autonomous systems whose dynamics are influenced by exogenous inputs. To this end, a hypernetwork-based framework ($HyperKKL$) is proposed with two input-conditioning strategies. First, an augmented observer approach ($HyperKKL_{obs}$) adds input-dependent corrections to the latent observer dynamics while retaining static transformation maps. Second, a dynamic observer approach ($HyperKKL_{dyn}$) employs a hypernetwork to generate encoder and decoder weights that are input-dependent, yielding time-varying transformation maps. We derive a theoretical worst-case bound on the state estimation error. Numerical evaluations on four nonlinear benchmark systems show that input conditioning yields consistent improvements in estimation accuracy over static autonomous maps, with an average symmetric mean absolute percentage error (SMAPE) reduction of 29% across all non-zero input regimes.

BotVerse is a scalable, event-driven framework for high-fidelity social simulation using LLM-based agents. It addresses the ethical risks of studying autonomous agents on live networks by isolating interactions within a controlled environment while grounding them in real-time content streams from the Bluesky ecosystem. The system features an asynchronous orchestration API and a simulation engine that emulates human-like temporal patterns and cognitive memory. Through the Synthetic Social Observatory, researchers can deploy customizable personas and observe multimodal interactions at scale. We demonstrate BotVersevia a coordinated disinformation scenario, providing a safe, experimental framework for red-teaming and computational social scientists. A video demonstration of the framework is available at https://youtu.be/eZSzO5Jarqk.

The evolution of intelligence in artificial systems provides a unique opportunity to identify universal computational principles. Here we show that large language models spontaneously develop synergistic cores where information integration exceeds individual parts remarkably similar to the human brain. Using Integrated Information Decomposition across multiple architectures we find that middle layers exhibit synergistic processing while early and late layers rely on redundancy. This organization is dynamic and emerges as a physical phase transition as task difficulty increases. Crucially ablating synergistic components causes catastrophic performance loss confirming their role as the physical entity of abstract reasoning and bridging artificial and biological intelligence.

Artificial intelligence is transforming molecular and materials science, but its growing computational and data demands raise critical sustainability challenges. In this Perspective, we examine resource considerations across the AI-driven discovery pipeline--from quantum-mechanical (QM) data generation and model training to automated, self-driving research workflows--building on discussions from the ``SusML workshop: Towards sustainable exploration of chemical spaces with machine learning'' held in Dresden, Germany. In this context, the availability of large quantum datasets has enabled rigorous benchmarking and rapid methodological progress, while also incurring substantial energy and infrastructure costs. We highlight emerging strategies to enhance efficiency, including general-purpose machine learning (ML) models, multi-fidelity approaches, model distillation, and active learning. Moreover, incorporating physics-based constraints within hierarchical workflows, where fast ML surrogates are applied broadly and high-accuracy QM methods are used selectively, can further optimize resource use without compromising reliability. Equally important is bridging the gap between idealized computational predictions and real-world conditions by accounting for synthesizability and multi-objective design criteria, which is essential for practical impact. Finally, we argue that sustainable progress will rely on open data and models, reusable workflows, and domain-specific AI systems that maximize scientific value per unit of computation, enabling efficient and responsible discovery of technological materials and therapeutics.

We present mlr3mbo, a comprehensive and modular toolbox for Bayesian optimization in R. mlr3mbo supports single- and multi-objective optimization, multi-point proposals, batch and asynchronous parallelization, input and output transformations, and robust error handling. While it can be used for many standard Bayesian optimization variants in applied settings, researchers can also construct custom BO algorithms from its flexible building blocks. In addition to an introduction to the software, its design principles, and its building blocks, the paper presents two extensive empirical evaluations of the software on the surrogate-based benchmark suite YAHPO Gym. To identify robust default configurations for both numeric and mixed-hierarchical optimization regimes, and to gain further insights into the respective impacts of individual settings, we run a coordinate descent search over the mlr3mbo configuration space and analyze its results. Furthermore, we demonstrate that mlr3mbo achieves state-of-the-art performance by benchmarking it against a wide range of optimizers, including HEBO, SMAC3, Ax, and Optuna.

This paper focuses on the problem of unbounded density ratio estimation -- an understudied yet critical challenge in statistical learning -- and its application to covariate shift adaptation. Much of the existing literature assumes that the density ratio is either uniformly bounded or unbounded but known exactly. These conditions are often violated in practice, creating a gap between theoretical guarantees and real-world applicability. In contrast, this work directly addresses unbounded density ratios and integrates them into importance weighting for effective covariate shift adaptation. We propose a three-step estimation method that leverages unlabeled data from both the source and target distributions: (1) estimating a relative density ratio; (2) applying a truncation operation to control its unboundedness; and (3) transforming the truncated estimate back into the standard density ratio. The estimated density ratio is then employed as importance weights for regression under covariate shift. We establish rigorous, non-asymptotic convergence guarantees for both the proposed density ratio estimator and the resulting regression function estimator, demonstrating optimal or near-optimal convergence rates. Our findings offer new theoretical insights into density ratio estimation and learning under covariate shift, extending classical learning theory to more practical and challenging scenarios.

Retrosynthetic planning is a fundamental task in organic chemistry, yet remains challenging due to its combinatorial complexity. To address this, conventional approaches typically rely on hybrid frameworks that combine single-step predictions with external search heuristics, inevitably fracturing the logical coherence between local molecular transformations and global planning objectives. To bridge this gap and embed sophisticated strategic foresight directly into the model's chemical reasoning, we introduce ReTriP, an end-to-end generative framework that reformulates retrosynthesis as a direct Chain-of-Thought reasoning task. We establish a path-coherent molecular representation and employ a progressive training curriculum that transitions from reasoning distillation to reinforcement learning with verifiable rewards, effectively aligning stepwise generation with practical route utility. Empirical evaluation on RetroBench demonstrates that ReTriP achieves state-of-the-art performance, exhibiting superior robustness in long-horizon planning compared to hybrid baselines.

Nonnegative matrix factorization (NMF) approximates a nonnegative matrix, $X$, by the product of two nonnegative factors, $WH$, where $W$ has $r$ columns and $H$ has $r$ rows. In this paper, we consider NMF using the component-wise L1 norm as the error measure (L1-NMF), which is suited for data corrupted by heavy-tailed noise, such as Laplace noise or salt and pepper noise, or in the presence of outliers. Our first contribution is an NP-hardness proof for L1-NMF, even when $r=1$, in contrast to the standard NMF that uses least squares. Our second contribution is to show that L1-NMF strongly enforces sparsity in the factors for sparse input matrices, thereby favoring interpretability. However, if the data is affected by false zeros, too sparse solutions might degrade the model. Our third contribution is a new, more general, L1-NMF model for sparse data, dubbed weighted L1-NMF (wL1-NMF), where the sparsity of the factorization is controlled by adding a penalization parameter to the entries of $WH$ associated with zeros in the data. The fourth contribution is a new coordinate descent (CD) approach for wL1-NMF, denoted as sparse CD (sCD), where each subproblem is solved by a weighted median algorithm. To the best of our knowledge, sCD is the first algorithm for L1-NMF whose complexity scales with the number of nonzero entries in the data, making it efficient in handling large-scale, sparse data. We perform extensive numerical experiments on synthetic and real-world data to show the effectiveness of our new proposed model (wL1-NMF) and algorithm (sCD).

Medical coding translates free-text clinical documentation into standardized codes drawn from classification systems that contain tens of thousands of entries and are updated annually. It is central to billing, clinical research, and quality reporting, yet remains largely manual, slow, and error-prone. Existing automated approaches learn to predict a fixed set of codes from labeled data, thereby preventing adaptation to new codes or different coding systems without retraining on different data. They also provide no explanation for their predictions, limiting trust in safety-critical settings. We introduce Symphony for Medical Coding, a system that approaches the task the way expert human coders do: by reasoning over the clinical narrative with direct access to the coding guidelines. This design allows Symphony to operate across any coding system and to provide span-level evidence linking each predicted code to the text that supports it. We evaluate on two public benchmarks and three real-world datasets spanning inpatient, outpatient, emergency, and subspecialty settings across the United States and the United Kingdom. Symphony achieves state-of-the-art results across all settings, establishing itself as a flexible, deployment-ready foundation for automated clinical coding.

The increasing availability of Machine Learning (ML) models, particularly foundation models, enables their use across a range of downstream applications, from scenarios with missing data to safety-critical contexts. This, in principle, may contravene not only the models' terms of use, but also governmental principles and regulations. This paper presents a preliminary investigation into the use of ML models by 173 open-source projects on GitHub, spanning 16 application domains. We evaluate whether models are used to make decisions, the scope of these decisions, and whether any post-processing measures are taken to reduce the risks inherent in fully autonomous systems. Lastly, we investigate the models' compliance with established terms of use. This study lays the groundwork for defining guidelines for developers and creating analysis tools that automatically identify potential regulatory violations in the use of ML models in software systems.

Skin cancer, particularly melanoma, remains a major cause of morbidity and mortality, making early detection critical. AI-driven dermatology systems often rely on skin lesion segmentation as a preprocessing step to delineate the lesion from surrounding skin and support downstream analysis. While fairness concerns regarding skin tone have been widely studied for lesion classification, the influence of skin tone on the segmentation stage remains under-quantified and is frequently assessed using coarse, discrete skin tone categories. In this work, we evaluate three strong segmentation architectures (UNet, DeepLabV3 with a ResNet50 backbone, and DINOv2) on two public dermoscopic datasets (HAM10000 and ISIC2017) and introduce a continuous pigment or contrast analysis that treats pixel-wise ITA values as distributions. Using Wasserstein distances between within-image distributions for skin-only, lesion-only, and whole-image regions, we quantify lesion skin contrast and relate it to segmentation performance across multiple metrics. Within the range represented in these datasets, global skin tone metrics (Fitzpatrick grouping or mean ITA) show weak association with segmentation quality. In contrast, low lesion-skin contrast is consistently associated with larger segmentation errors in models, indicating that boundary ambiguity and low contrast are key drivers of failure. These findings suggest that fairness improvements in dermoscopic segmentation should prioritize robust handling of low-contrast lesions, and the distribution-based pigment measures provide a more informative audit signal than discrete skin-tone categories.

A robust decision-making process must take into account uncertainty, especially when the choice involves inherent risks. Because artificial Intelligence (AI) systems are increasingly integrated into decision-making workflows, managing uncertainty relies more and more on the metacognitive capabilities of these systems; i.e, their ability to assess the reliability of and regulate their own decisions. Hence, it is crucial to employ robust methods to measure the metacognitive abilities of AI. This paper is primarily a methodological contribution arguing for the adoption of the meta-d' framework, or its model-free alternatives, as the gold standard for assessing the metacognitive sensitivity of AIs--the ability to generate confidence ratings that distinguish correct from incorrect responses. Moreover, we propose to leverage signal detection theory (SDT) to measure the ability of AIs to spontaneously regulate their decisions based on uncertainty and risk. To demonstrate the practical utility of these psychophysical frameworks, we conduct two series of experiments on three large language models (LLMs)--GPT-5, DeepSeek-V3.2-Exp, and Mistral-Medium-2508. In the first experiments, LLMs performed a primary judgment followed by a confidence rating. In the second, LLMs only performed the primary judgment, while we manipulated the risk associated with either response. On the one hand, applying the meta-d' framework allows us to conduct comparisons along three axes: comparing an LLM to optimality, comparing different LLMs on a given task, and comparing the same LLM across different tasks. On the other hand, SDT allows us to assess whether LLMs become more conservative when risks are high.

Real world scenarios can be captured with lifted probability distributions. However, distributions are usually encoded in a table or list, requiring an exponential number of values. Hence, we propose a method for extracting first-order formulas from probability distributions that require significantly less values by reducing the number of values in a distribution and then extracting, for each value, a logical formula to be further minimized. This reduction and minimization allows for increasing the sparsity in the encoding while also generalizing a given distribution. Our evaluation shows that sparsity can increase immensely by extracting a small set of short formulas while preserving core information.

Large Language Models (LLMs) demonstrate exceptional capabilities in factual question answering, yet they sometimes provide incorrect responses. To address this issue, knowledge editing techniques have emerged as effective methods for correcting factual information in LLMs. However, typical knowledge editing workflows struggle with identifying the optimal set of model layers for editing and rely on summary indicators that provide insufficient guidance. This lack of transparency hinders effective comparison and identification of optimal editing strategies. In this paper, we present KEditVis, a novel visual analytics system designed to assist users in gaining a deeper understanding of knowledge editing through interactive visualizations, improving editing outcomes, and discovering valuable insights for the future development of knowledge editing algorithms. With KEditVis, users can select appropriate layers as the editing target, explore the reasons behind ineffective edits, and perform more targeted and effective edits. Our evaluation, including usage scenarios, expert interviews, and a user study, validates the effectiveness and usability of the system.

The common claim that generative AI simply amplifies the Dunning-Kruger effect is too coarse to capture the available evidence. The clearest findings instead suggest that large language model (LLM) use can improve observable output and short-term task performance while degrading metacognitive accuracy and flattening the classic competence-confidence gradient across skill groups. This paper synthesizes evidence from human-AI interaction, learning research, and model evaluation, and proposes the working model of AI-mediated metacognitive decoupling: a widening gap among produced output, underlying understanding, calibration accuracy, and self-assessed ability. This four-variable account better explains overconfidence, over- and under-reliance, crutch effects, and weak transfer than the simpler metaphor of a uniformly steeper Dunning-Kruger curve. The paper concludes with implications for tool design, assessment, and knowledge work.

Agent traces carry increasing analytical value in agentic systems and context engineering, yet most prior work treats conversation format as a trivial implementation detail. Modern agent conversations, however, contain deeply structured content, including nested tool calls and results, chain-of-thought reasoning blocks, sub-agent invocations, context-window compaction boundaries, and harness-injected system directives, whose complexity far exceeds that of simple user-assistant exchanges. Feeding such traces to a reflector or other analytical mechanism in plain text, JSON, YAML, or via grep can materially degrade analysis quality. This paper presents VCC (View-oriented Conversation Compiler), a compiler (lex, parse, IR, lower, emit) that transforms raw agent JSONL logs into a family of structured views: a full view (lossless transcript serving as the canonical line-number coordinate system), a user-interface (UI) view (reconstructing the interaction as the user actually perceived it), and an adaptive view (a structure-preserving projection governed by a relevance predicate). In a context-engineering experiment on AppWorld, replacing only the reflector's input format, from raw JSONL to VCC-compiled views, leads to higher pass rates across all three model configurations tested, while cutting reflector token consumption by half to two-thirds and producing more concise learned memory. These results suggest that message format functions as infrastructure for context engineering, not as an incidental implementation choice.

Multimodal learning enables neural networks to integrate information from heterogeneous sources, but active learning in this setting faces distinct challenges. These include missing modalities, differences in modality difficulty, and varying interaction structures. These are issues absent in the unimodal case. While the behavior of active learning strategies in unimodal settings is well characterized, their behavior under such multimodal conditions remains poorly understood. We introduce a new framework for benchmarking multimodal active learning that isolates these pitfalls using synthetic datasets, allowing systematic evaluation without confounding noise. Using this framework, we compare unimodal and multimodal query strategies and validate our findings on two real-world datasets. Our results show that models consistently develop imbalanced representations, relying primarily on one modality while neglecting others. Existing query methods do not mitigate this effect, and multimodal strategies do not consistently outperform unimodal ones. These findings highlight limitations of current active learning methods and underline the need for modality-aware query strategies that explicitly address these pitfalls. Code and benchmark resources will be made publicly available.

Large vision-language models (LVLMs) achieve impressive performance, yet their internal decision-making processes remain opaque, making it difficult to determine if the success stems from true multimodal fusion or from reliance on unimodal priors. To address this attribution gap, we introduce a novel framework using partial information decomposition (PID) to quantitatively measure the "information spectrum" of LVLMs -- decomposing a model's decision-relevant information into redundant, unique, and synergistic components. By adapting a scalable estimator to modern LVLM outputs, our model-agnostic pipeline profiles 26 LVLMs on four datasets across three dimensions -- breadth (cross-model & cross-task), depth (layer-wise information dynamics), and time (learning dynamics across training). Our analysis reveals two key results: (i) two task regimes (synergy-driven vs. knowledge-driven) and (ii) two stable, contrasting family-level strategies (fusion-centric vs. language-centric). We also uncover a consistent three-phase pattern in layer-wise processing and identify visual instruction tuning as the key stage where fusion is learned. Together, these contributions provide a quantitative lens beyond accuracy-only evaluation and offer insights for analyzing and designing the next generation of LVLMs. Code and data are available at https://github.com/RiiShin/pid-lvlm-analysis .

Hackathons are time-bound collaborative events that often target software creation. Although hackathons have been studied in the past, existing work focused on in-depth case studies limiting our understanding of hackathons as a software engineering activity. To complement the existing body of knowledge, we introduce HackRep, a dataset of 100,356 hackathon GitHub repositories. We illustrate the ways HackRep can benefit software engineering researchers by presenting a preliminary investigation of hackathon project continuation, hackathon team composition, and an estimation of hackathon geography. We further display the opportunities of using this dataset, for instance showing the possibility of estimating hackathon durations based on commit timestamps.

Agentic systems for business process automation often require compliance with policies governing conditional updates to the system state. Evaluation of policy adherence in LLM-based agentic workflows is typically performed by comparing the final system state against a predefined ground truth. While this approach detects explicit policy violations, it may overlook a more subtle class of issues in which agents bypass required policy checks, yet reach a correct outcome due to favorable circumstances. We refer to such cases as $\textit{near-misses}$ or $\textit{latent failures}$. In this work, we introduce a novel metric for detecting latent policy failures in agent conversations traces. Building on the ToolGuard framework, which converts natural-language policies into executable guard code, our method analyzes agent trajectories to determine whether agent's tool-calling decisions where sufficiently informed. We evaluate our approach on the $τ^2$-verified Airlines benchmark across several contemporary open and proprietary LLMs acting as agents. Our results show that latent failures occur in 8-17% of trajectories involving mutating tool calls, even when the final outcome matches the expected ground-truth state. These findings reveal a blind spot in current evaluation methodologies and highlight the need for metrics that assess not only final outcomes but also the decision process leading to them.

Existing narrative extraction methods face a trade-off between coherence, interactivity, and multi-storyline support. Narrative Maps supports rich interaction and generates multiple storylines as a byproduct of its coverage constraints, though this comes at the cost of individual path coherence. Narrative Trails achieves high coherence through maximum capacity path optimization but provides no mechanism for user guidance or multiple perspectives. We introduce agenda-based narrative extraction, a method that bridges this gap by integrating large language models into the Narrative Trails pathfinding process to steer storyline construction toward user-specified perspectives. Our approach uses an LLM at each step to rank candidate documents based on their alignment with a given agenda while maintaining narrative coherence. Running the algorithm with different agendas yields different storylines through the same corpus. We evaluated our approach on a news article corpus using LLM judges with Claude Opus 4.5 and GPT 5.1, measuring both coherence and agenda alignment across 64 endpoint pairs and 6 agendas. LLM-driven steering achieves 9.9% higher alignment than keyword matching on semantic agendas (p=0.017), with 13.3% improvement on \textit{Regime Crackdown} specifically (p=0.037), while keyword matching remains competitive on agendas with literal keyword overlap. The coherence cost is minimal: LLM steering reduces coherence by only 2.2% compared to the agenda-agnostic baseline. Counter-agendas that contradict the source material score uniformly low (2.2-2.5) across all methods, confirming that steering cannot fabricate unsupported narratives.

Autonomous 6G network management requires agents that can execute tools, observe the resulting state changes, and adapt their decisions accordingly. Existing benchmarks based on static questions or scripted episode replay, however, do not support such closed-loop interaction, limiting agents to passive evaluation without the ability to learn from environmental feedback. This paper presents 6GAgentGym to provide closed-loop capability. The framework provides an interactive environment with 42 typed tools whose effect classification distinguishes read-only observation from state-mutating configuration, backed by a learned Experiment Model calibrated on NS-3 simulation data. 6G-Forge bootstraps closed-loop training trajectories from NS-3 seeds via iterative Self-Instruct generation with execution verification against the Experiment Model. Supervised fine-tuning on the resulting corpus followed by reinforcement learning with online closed-loop interaction enables an 8B open-source model to achieve comparable overall success rate to GPT-5 on the accompanying 6GAgentBench, with stronger performance on long-horizon tasks. Together, these components provide a viable path toward autonomous, closed-loop network management.

Aligning human-interpretable concepts with the internal representations learned by modern machine learning systems remains a central challenge for interpretable AI. We introduce a geometric framework for comparing supervised human concepts with unsupervised intermediate representations extracted from foundation model embeddings. Motivated by the role of conceptual leaps in scientific discovery, we formalise the notion of concept frustration: a contradiction that arises when an unobserved concept induces relationships between known concepts that cannot be made consistent within an existing ontology. We develop task-aligned similarity measures that detect concept frustration between supervised concept-based models and unsupervised representations derived from foundation models, and show that the phenomenon is detectable in task-aligned geometry while conventional Euclidean comparisons fail. Under a linear-Gaussian generative model we derive a closed-form expression for Bayes-optimal concept-based classifier accuracy, decomposing predictive signal into known-known, known-unknown and unknown-unknown contributions and identifying analytically where frustration affects performance. Experiments on synthetic data and real language and vision tasks demonstrate that frustration can be detected in foundation model representations and that incorporating a frustrating concept into an interpretable model reorganises the geometry of learned concept representations, to better align human and machine reasoning. These results suggest a principled framework for diagnosing incomplete concept ontologies and aligning human and machine conceptual reasoning, with implications for the development and validation of safe interpretable AI for high-risk applications.

Semantic interaction (SI) enables analysts to incorporate their cognitive processes into AI models through direct manipulation of visualizations. While SI frameworks for narrative extraction have been proposed, empirical evaluations of their effectiveness remain limited. This paper presents a user study that evaluates SI for narrative map sensemaking, involving 33 participants under three conditions: a timeline baseline, a basic narrative map, and an interactive narrative map with SI capabilities. The results show that the map-based prototypes yielded more insights than the timeline baseline, with the SI-enabled condition reaching statistical significance and the basic map condition trending in the same direction. The SI-enabled condition showed the highest mean performance; differences between the map conditions were not statistically significant but showed large effect sizes (d > 0.8), suggesting that the study was underpowered to detect them. Qualitative analysis identified two distinct SI approaches-corrective and additive-that enable analysts to impose quality judgments and organizational structure on extracted narratives. We also find that SI users achieved comparable exploration breadth with less parameter manipulation, suggesting that SI serves as an alternative pathway for model refinement. This work provides empirical evidence that map-based representations outperform timelines for narrative sensemaking, along with qualitative insights into how analysts use SI for narrative refinement.

When testing data and training data come from different distributions, deep neural networks (DNNs) will face significant safety risks in practical applications. Therefore, out-of-distribution (OOD) detection techniques, which can identify OOD samples at test time and alert the system, are urgently needed. Existing graph OOD detection methods usually characterize fine-grained in-distribution (ID) patterns from multiple perspectives, and train end-to-end graph neural networks (GNNs) for prediction. However, due to the unavailability of OOD data during training, the absence of explicit supervision signals could lead to sub-optimal performance of end-to-end encoders. To address this issue, we follow the pre-training+prompting paradigm to utilize pre-trained GNN encoders, and propose Disentangled Graph Prompting (DGP), to capture fine-grained ID patterns with the help of ID graph labels. Specifically, we design two prompt generators that respectively generate class-specific and class-agnostic prompt graphs by modifying the edge weights of an input graph. We also design several effective losses to train the prompt generators and prevent trivial solutions. We conduct extensive experiments on ten datasets to demonstrate the superiority of our proposed DGP, which achieves a relative AUC improvement of 3.63% over the best graph OOD detection baseline. Ablation studies and hyper-parameter experiments further show the effectiveness of DGP. Code is available at https://github.com/BUPT-GAMMA/DGP.

Blood donation centers face challenges in matching supply with demand while managing donor availability. Although targeted outreach is important, it can cause donor fatigue via over-solicitation. Effective recruitment requires targeting the right donors at the right time, balancing constraints with donor convenience and eligibility. Despite extensive work on blood supply chain optimization and growing interest in algorithmic donor recruitment, the operational problem of assigning donors to sessions across a multi-site network, taking into account eligibility, capacity, blood-type demand targets, geographic convenience, and donor safety, remains unaddressed. We address this gap with an optimization framework for donor invitation scheduling incorporating donor eligibility, travel convenience, blood-type demand targets, and penalties. We evaluate two strategies: (i) a binary integer linear programming (BILP) formulation and (ii) an efficient greedy heuristic. Evaluation uses the registry from Instituto Português do Sangue e da Transplantação (IPST) for invite planning in the Lisbon operational region using 4-month windows. A prospective pipeline integrates organic attendance forecasting, quantile-based demand targets, and residual capacity estimation for forward-looking invitation plans. Results reveal its key role in closing the supply-demand gap in the Lisbon operational region. A controlled comparison shows that the greedy heuristic achieves results comparable to the BILP, with 188x less peak memory and 115x faster runtime; trade-offs include 3.9 pp lower demand fulfillment (86.1% vs. 90.0%), larger donor-session distance, higher adverse-reaction donor exposure, and greater invitation burden per non-high-frequency donor, reflecting local versus global optimization. Experiments assess how constraint-aware scheduling can close gaps by mobilizing eligible inactive/lapsing donors.

Can AI accelerate the development of AI itself? While recent agentic systems have shown strong performance on well-scoped tasks with rapid feedback, it remains unclear whether they can tackle the costly, long-horizon, and weakly supervised research loops that drive real AI progress. We present ASI-Evolve, an agentic framework for AI-for-AI research that closes this loop through a learn-design-experiment-analyze cycle. ASI-Evolve augments standard evolutionary agents with two key components: a cognition base that injects accumulated human priors into each round of exploration, and a dedicated analyzer that distills complex experimental outcomes into reusable insights for future iterations. To our knowledge, ASI-Evolve is the first unified framework to demonstrate AI-driven discovery across three central components of AI development: data, architectures, and learning algorithms. In neural architecture design, it discovered 105 SOTA linear attention architectures, with the best discovered model surpassing DeltaNet by +0.97 points, nearly 3x the gain of recent human-designed improvements. In pretraining data curation, the evolved pipeline improves average benchmark performance by +3.96 points, with gains exceeding 18 points on MMLU. In reinforcement learning algorithm design, discovered algorithms outperform GRPO by up to +12.5 points on AMC32, +11.67 points on AIME24, and +5.04 points on OlympiadBench. We further provide initial evidence that this AI-for-AI paradigm can transfer beyond the AI stack through experiments in mathematics and biomedicine. Together, these results suggest that ASI-Evolve represents a promising step toward enabling AI to accelerate AI across the foundational stages of development, offering early evidence for the feasibility of closed-loop AI research.

An agent that operates sequentially must incorporate new experience without forgetting old experience, under a fixed memory budget. We propose a framework in which memory is not a parameter vector but a stochastic process: a Bridge Diffusion on a replay interval $[0,1]$, whose terminal marginal encodes the present and whose intermediate marginals encode the past. New experience is incorporated via a three-step \emph{Compress--Add--Smooth} (CAS) recursion. We test the framework on the class of models with marginal probability densities modeled via Gaussian mixtures of fixed number of components~$K$ in $d$ dimensions; temporal complexity is controlled by a fixed number~$L$ of piecewise-linear protocol segments whose nodes store Gaussian-mixture states. The entire recursion costs $O(LKd^2)$ flops per day -- no backpropagation, no stored data, no neural networks -- making it viable for controller-light hardware. Forgetting in this framework arises not from parameter interference but from lossy temporal compression: the re-approximation of a finer protocol by a coarser one under a fixed segment budget. We find that the retention half-life scales linearly as $a_{1/2}\approx c\,L$ with a constant $c>1$ that depends on the dynamics but not on the mixture complexity~$K$, the dimension~$d$, or the geometry of the target family. The constant~$c$ admits an information-theoretic interpretation analogous to the Shannon channel capacity. The stochastic process underlying the bridge provides temporally coherent ``movie'' replay -- compressed narratives of the agent's history, demonstrated visually on an MNIST latent-space illustration. The framework provides a fully analytical ``Ising model'' of continual learning in which the mechanism, rate, and form of forgetting can be studied with mathematical precision.

Continuous image tokenizers enable efficient visual generation, and those based on variational frameworks can learn smooth, structured latent representations through KL regularization. Yet this often leads to posterior collapse when using fewer tokens, where the encoder fails to encode informative features into the compressed latent space. To address this, we introduce \textbf{MacTok}, a \textbf{M}asked \textbf{A}ugmenting 1D \textbf{C}ontinuous \textbf{Tok}enizer that leverages image masking and representation alignment to prevent collapse while learning compact and robust representations. MacTok applies both random masking to regularize latent learning and DINO-guided semantic masking to emphasize informative regions in images, forcing the model to encode robust semantics from incomplete visual evidence. Combined with global and local representation alignment, MacTok preserves rich discriminative information in a highly compressed 1D latent space, requiring only 64 or 128 tokens. On ImageNet, MacTok achieves a competitive gFID of 1.44 at 256$\times$256 and a state-of-the-art 1.52 at 512$\times$512 with SiT-XL, while reducing token usage by up to 64$\times$. These results confirm that masking and semantic guidance together prevent posterior collapse and achieve efficient, high-fidelity tokenization.

As AI agents evolve, the community is rapidly shifting from single Large Language Models (LLMs) to Multi-Agent Systems (MAS) to overcome cognitive bottlenecks in automated research. However, the optimal multi-agent coordination framework for these autonomous agents remains largely unexplored. In this paper, we present a systematic empirical study investigating the comparative efficacy of distinct multi-agent structures for automated machine learning optimization. Utilizing a rigorously controlled, execution-based testbed equipped with Git worktree isolation and explicit global memory, we benchmark a single-agent baseline against two multi-agent paradigms: a subagent architecture (parallel exploration with post-hoc consolidation) and an agent team architecture (experts with pre-execution handoffs). By evaluating these systems under strictly fixed computational time budgets, our findings reveal a fundamental trade-off between operational stability and theoretical deliberation. The subagent mode functions as a highly resilient, high-throughput search engine optimal for broad, shallow optimizations under strict time constraints. Conversely, the agent team topology exhibits higher operational fragility due to multi-author code generation but achieves the deep theoretical alignment necessary for complex architectural refactoring given extended compute budgets. These empirical insights provide actionable guidelines for designing future autoresearch systems, advocating for dynamically routed architectures that adapt their collaborative structures to real-time task complexity.

Always-on edge cameras generate continuous video streams where redundant frames degrade cross-modal retrieval by crowding correct results out of top-k search. This paper presents a streaming retrieval architecture: an on-device epsilon-net filter retains only semantically novel frames, building a denoised embedding index; a cross-modal adapter and cloud re-ranker compensate for the compact encoder's weak alignment. A single-pass streaming filter outperforms offline alternatives (k-means, farthest-point, uniform, random) across eight vision-language models (8M-632M) on two egocentric datasets (AEA, EPIC-KITCHENS). Combined, the architecture reaches 45.6% Hit@5 on held-out data using an 8M on-device encoder at an estimated 2.7 mW.

Scratch is the most popular programming environment for novices, with over 1.15 billion projects created worldwide. Unlike traditional languages, correctness in Scratch is defined by visible behavior on the stage rather than by code structure alone, so programs that appear correct in the workspace can still fail at runtime due to timing, event ordering, or cross-sprite interactions. Visual execution evidence such as gameplay videos can therefore be essential for diagnosis and repair. However, capturing and processing this evidence inside an automated repair loop introduces substantial overhead. Probing execution, recording stage behavior, rebuilding executable .sb3 projects, and verifying candidate fixes consume time, monetary cost, and resources across an entire repair trajectory rather than a single model call. We present EcoScratch, a repair pipeline that uses lightweight runtime signals to decide whether the next attempt stays text-only or escalates to multimodal prompting. The controller also sets the JSON Patch budget and verification effort, so evidence choice and repair budget are coupled inside the same decision. EcoScratch rebuilds candidate fixes into executable .sb3 projects and records per-trajectory traces, monetary cost, local-runtime energy. We evaluate 12 models on 100 executable Scratch repair projects under four controller settings, yielding 4800 repair trajectories. In this matrix, a selective multimodal policy gives the strongest observed success-cost-energy tradeoff. It reaches the highest generation success (30.3%) while using less average cost and local-runtime energy than the two non-adaptive multimodal baselines under the same bounded trajectory budget; text-only remains the lowest-cost floor. Across the evaluated matrix, multimodal evidence helps most when it is used to control escalation within a bounded trajectory budget rather than applied uniformly.

In the development and maintenance of Android apps, the quick and accurate reproduction of user-reported bugs is crucial to ensure application quality and improve user satisfaction. However, this process is often time-consuming and complex. Therefore, there is a need for an automated approach that can explore the Application Under Test (AUT) and identify the correct sequence of User Interface (UI) actions required to reproduce a bug, given only a complete bug report. Large Language Models (LLMs) have shown remarkable capabilities in understanding textual and visual semantics, making them a promising tool for planning UI actions. Nevertheless, our study shows that even when using state-of-the-art LLM-based approaches, these methods still struggle to follow detailed bug reproduction instructions and replan based on new information, due to their inability to accurately predict and interpret the visual effects of UI components. To address these limitations, we propose LTGDroid. Our insight is to execute all possible UI actions on the current UI page during exploration, record their corresponding visual effects, and leverage these visual cues to guide the LLM in selecting UI actions that are likely to reproduce the bug. We evaluated LTGDroid, instantiated with GPT-4.1, on a benchmark consisting of 75 bug reports from 45 popular Android apps. The results show that LTGDroid achieves a reproduction success rate of 87.51%, improving over the state-of-the-art baselines by 49.16% and 556.30%, while requiring an average of 20.45 minutes and approximately $0.27 to successfully reproduce a bug. The LTGDroid implementation is publicly available at https://github.com/N3onFlux/LTGDroid.

Understanding how the brain processes linguistic constructions is a central challenge in cognitive neuroscience and linguistics. Recent computational studies show that artificial neural language models spontaneously develop differentiated representations of Argument Structure Constructions (ASCs), generating predictions about when and how construction-level information emerges during processing. The present study tests these predictions in human neural activity using electroencephalography (EEG). Ten native English speakers listened to 200 synthetically generated sentences across four construction types (transitive, ditransitive, caused-motion, resultative) while neural responses were recorded. Analyses using time-frequency methods, feature extraction, and machine learning classification revealed construction-specific neural signatures emerging primarily at sentence-final positions, where argument structure becomes fully disambiguated, and most prominently in the alpha band. Pairwise classification showed reliable differentiation, especially between ditransitive and resultative constructions, while other pairs overlapped. Crucially, the temporal emergence and similarity structure of these effects mirror patterns in recurrent and transformer-based language models, where constructional representations arise during integrative processing stages. These findings support the view that linguistic constructions are neurally encoded as distinct form-meaning mappings, in line with Construction Grammar, and suggest convergence between biological and artificial systems on similar representational solutions. More broadly, this convergence is consistent with the idea that learning systems discover stable regions within an underlying representational landscape - recently termed a Platonic representational space - that constrains the emergence of efficient linguistic abstractions.

Deep learning is widely deployed for time series learning tasks such as classification and forecasting. Despite the empirical successes, only little theory has been developed so far in the time series context. In this work, we prove that if the network inputs are generated from short-range dependent linear processes, the outputs of fully convolutional neural networks (FCNs) with global average pooling (GAP) are asymptotically Gaussian and the limit is attained if the length of the observed time series tends to infinity. The proof leverages existing tools from the theoretical time series literature. Based on our theory, we propose a generalization of the GAP layer by considering a global weighted pooling step with slowly varying, learnable coefficients.

Although Deep Reinforcement Learning has proven highly effective for complex decision-making problems, it demands significant computational resources and careful parameter adjustment in order to develop successful strategies. Evolution strategies offer a more straightforward, derivative-free approach that is less computationally costly and simpler to deploy. However, ES generally do not match the performance levels achieved by DRL, which calls into question their suitability for more demanding scenarios. This study examines the performance of ES and DRL across tasks of varying difficulty, including Flappy Bird, Breakout and Mujoco environments, as well as whether ES could be used for initial training to enhance DRL algorithms. The results indicate that ES do not consistently train faster than DRL. When used as a preliminary training step, they only provide benefits in less complex environments (Flappy Bird) and show minimal or no improvement in training efficiency or stability across different parameter settings when applied to more sophisticated tasks (Breakout and MuJoCo Walker).

Acute poly-substance intoxication requires rapid, life-saving decisions under substantial uncertainty, as clinicians must rely on incomplete ingestion details and nonspecific symptoms. Effective diagnostic reasoning in this chaotic environment requires fusing unstructured, non-medical narratives (e.g. paramedic scene descriptions and unreliable patient self-reports or known histories), with structured medical data like vital signs. While Large Language Models (LLMs) show potential for processing such heterogeneous inputs, they struggle in this setting, often underperforming simple baselines that rely solely on patient histories. To address this, we present DeToxR (Decision-support for Toxicology with Reasoning), the first adaptation of Reinforcement Learning (RL) to emergency toxicology. We design a robust data-fusion engine for multi-label prediction across 14 substance classes based on an LLM finetuned with Group Relative Policy Optimization (GRPO). We optimize the model's reasoning directly using a clinical performance reward. By formulating a multi-label agreement metric as the reward signal, the model is explicitly penalized for missing co-ingested substances and hallucinating absent poisons. Our model significantly outperforms its unadapted base LLM counterpart and supervised baselines. Furthermore, in a clinical validation study, the model indicates a clinical advantage by outperforming an expert toxicologist in identifying the correct poisons (Micro-F1: 0.644 vs. 0.473). These results demonstrate the potential of RL-aligned LLMs to synthesize unstructured pre-clinical narratives and structured medical data for decision support in high-stakes environments.

Compiling quantum circuits is a major bottleneck in quantum computing, and given the scale required in a few years, is likely to become infeasibly long. Techniques to reduce compilation time for quantum circuits are sorely needed. Furthermore, resources to test acceleration techniques are similarly lacking due to the limited scale of circuits in benchmark suites and mismatches in characteristics of these circuits and those produced by random circuit generators. This paper resolves the latter of these problems by describing a random circuit generator which allows control of circuit density, width and depth parameters. This is used to derive 8000 experimental large-scale circuits and test a novel approach to compiler parallelisation. This separates a circuit into sub-circuits which are compiled in parallel and recombined to produce a compiled circuit. When the parallel approach was tested using Qiskit, a peak speedup of 15.56 was achieved with corresponding overheads of less than 1%.

Preserving intangible cultural dances rooted in centuries of tradition and governed by strict structural and symbolic rules presents unique challenges in the digital era. Among these, Bharatanatyam, a classical Indian dance form, stands out for its emphasis on codified adavus and precise key postures. Accurately generating these postures is crucial not only for maintaining anatomical and stylistic integrity, but also for enabling effective documentation, analysis, and transmission to broader global audiences through digital means. We propose a pose-aware generative framework integrated with a pose estimation module, guided by keypoint-based loss and pose consistency constraints. These supervisory signals ensure anatomical accuracy and stylistic integrity in the synthesized outputs. We evaluate four configurations: standard conditional generative adversarial network (cGAN), cGAN with pose supervision, conditional diffusion, and conditional diffusion with pose supervision. Each model is conditioned on key posture class labels and optimized to maintain geometric structure. In both cGAN and conditional diffusion settings, the integrated pose guidance aligns generated poses with ground-truth keypoint structures, promoting cultural fidelity. Our results demonstrate that incorporating pose supervision significantly enhances the quality, realism, and authenticity of generated Bharatanatyam postures. This framework provides a scalable approach for the digital preservation, education, and dissemination of traditional dance forms, enabling high-fidelity generation without compromising cultural precision. Code is available at https://github.com/jagidsh/Generating-Key-Postures-of-Bharatanatyam-Adavus-with-Pose-Estimation.

We study polynomial group convolutional neural networks (PGCNNs) for an arbitrary finite group $G$. In particular, we introduce a new mathematical framework for PGCNNs using the language of graded group algebras. This framework yields two natural parametrizations of the architecture, based on Hadamard and Kronecker products, related by a linear map. We compute the dimension of the associated neuromanifold, verifying that it depends only on the number of layers and the size of the group. We also describe the general fiber of the Kronecker parametrization up to the regular group action and rescaling, and conjecture the analogous description for the Hadamard parametrization. Our conjecture is supported by explicit computations for small groups and shallow networks.

Multi-omic datasets offer opportunities for improved biomarker discovery in cancer research, but their high dimensionality and limited sample sizes make identifying compact and effective biomarker panels challenging. Feature selection in large-scale omics can be efficiently addressed by combining machine learning with genetic algorithms, which naturally support multi-objective optimization of predictive accuracy and biomarker set size. However, genetic algorithms remain relatively underexplored for multi-omic feature selection, where most approaches concatenate all layers into a single feature space. To address this limitation, we introduce Sweeping*, a multi-view, multi-objective algorithm alternating between single- and multi-view optimization. It employs a nested single-view multi-objective optimizer, and for this study we use the genetic algorithm NSGA3-CHS. It first identifies informative biomarkers within each layer, then jointly evaluates cross-layer interactions; these multi-omic solutions guide the next single-view search. Through repeated sweeps, the algorithm progressively identifies compact biomarker panels capturing cross-modal complementary signals. We benchmark five Sweeping* strategies, including hierarchical and concatenation-based variants, using survival prediction on three TCGA cohorts. Each strategy jointly optimizes predictive accuracy and set size, measured via the concordance index and root-leanness. Overall performance and estimation error are assessed through cross hypervolume and Pareto delta under 5-fold cross-validation. Our results show that Sweeping* can improve the accuracy-complexity trade-off when sufficient survival signal is present and that integrating omic layers can enhance survival prediction beyond clinical-only models, although benefits remain cohort-dependent.

Large Language Models (LLMs) show promise for automated grading, but their outputs can be unreliable. Rather than improving grading accuracy directly, we address a complementary problem: \textit{predicting when an LLM grader is likely to be correct}. This enables selective automation where high-confidence predictions are processed automatically while uncertain cases are flagged for human review. We compare three confidence estimation methods (self-reported confidence, self-consistency voting, and token probability) across seven LLMs of varying scale (4B to 120B parameters) on three educational datasets: RiceChem (long-answer chemistry), SciEntsBank, and Beetle (short-answer science). Our experiments reveal that self-reported confidence consistently achieves the best calibration across all conditions (avg ECE 0.166 vs 0.229 for self-consistency). Surprisingly, self-consistency remains 38\% worse despite requiring 5$\times$ the inference cost. Larger models exhibit substantially better calibration though gains vary by dataset and method (e.g., a 28\% ECE reduction for self-reported), with GPT-OSS-120B achieving the best calibration (avg ECE 0.100) and strong discrimination (avg AUC 0.668). We also observe that confidence is strongly top-skewed across methods, creating a ``confidence floor'' that practitioners must account for when setting thresholds. These findings suggest that simply asking LLMs to report their confidence provides a practical approach for identifying reliable grading predictions. Code is available \href{https://github.com/sonkar-lab/llm_grading_calibration}{here}.

Scientific idea generation (SIG) is critical to AI-driven autonomous research, yet existing approaches are often constrained by a static retrieval-then-generation paradigm, leading to homogeneous and insufficiently divergent ideas. In this work, we propose FlowPIE, a tightly coupled retrieval-generation framework that treats literature exploration and idea generation as a co-evolving process. FlowPIE expands literature trajectories via a flow-guided Monte Carlo Tree Search (MCTS) inspired by GFlowNets, using the quality of current ideas assessed by an LLM-based generative reward model (GRM) as a supervised signal to guide adaptive retrieval and construct a diverse, high-quality initial population. Based on this population, FlowPIE models idea generation as a test-time idea evolution process, applying selection, crossover, and mutation with the isolation island paradigm and GRM-based fitness computation to incorporate cross-domain knowledge. It effectively mitigates the information cocoons arising from over-reliance on parametric knowledge and static literature. Extensive evaluations demonstrate that FlowPIE consistently produces ideas with higher novelty, feasibility and diversity compared to strong LLM-based and agent-based frameworks, while enabling reward scaling during test time.

Motivated by the success of score-based generative models, a number of diffusion-based algorithms have recently been proposed for the problem of sampling from a probability measure whose unnormalized density can be accessed. Among them, Grenioux et al. introduced SLIPS, a sampling algorithm based on Stochastic Localization. While SLIPS exhibits strong empirical performance, no rigorous convergence analysis has previously been provided. In this work, we close this gap by establishing the first guarantee for SLIPS in total variation distance. Under minimal assumptions on the target, our bound implies that the number of steps required to achieve an $\varepsilon$-guarantee scales linearly with the dimension, up to logarithmic factors. The analysis leverages techniques from the theory of score-based generative models and further provides theoretical insights into the empirically observed optimal choice of discretization points.

Accurate event-based modeling of electric vehicle (EV) charging is essential for grid reliability and smart-charging design. While traditional statistical methods capture marginal distributions, they often fail to model the complex, non-linear dependencies between charging variables, specifically arrival times, durations, and energy demand. This paper addresses this gap by introducing the first application of Vine copulas and Copula Density Neural Estimation framework (CODINE) to the EV domain. We evaluate these high-capacity dependence models across three diverse real-world datasets. Our results demonstrate that by explicitly focusing on modeling the joint dependence structure, Vine copulas and CODINE outperform established parametric families and remain highly competitive against state-of-the-art benchmarks like conditional Gaussian Mixture Model Networks. We show that these methods offer superior preservation of tail behaviors and correlation structures, providing a robust framework for synthetic charging event generation in varied infrastructure contexts.

Multilingualism is incredibly common around the world, leading to many important theoretical and practical questions about how children learn multiple languages at once. For example, does multilingual acquisition lead to delays in learning? Are there better and worse ways to structure multilingual input? Many correlational studies address these questions, but it is surprisingly difficult to get definitive answers because children cannot be randomly assigned to be multilingual and data are typically not matched between languages. We use language model training as a method for simulating a variety of highly controlled exposure conditions, and create matched 100M-word mono- and bilingual datasets using synthetic data and machine translation. We train GPT-2 models on monolingual and bilingual data organized to reflect a range of exposure regimes, and evaluate their performance on perplexity, grammaticality, and semantic knowledge. Across model scales and measures, bilingual models perform similarly to monolingual models in one language, but show strong performance in the second language as well. These results suggest that there are no strong differences between different bilingual exposure regimes, and that bilingual input poses no in-principle challenges for agnostic statistical learners.

Due to the scarcity of labeled dialectal speech, audio dialect classification is a challenging task for most languages, including Swiss German. In this work, we explore the ability of large language models (LLMs) as agents in understanding the dialects and whether they can show comparable performance to models such as HuBERT in dialect classification. In addition, we provide an LLM baseline and a human linguist one. Our approach uses phonetic transcriptions produced by ASR systems and combines them with linguistic resources such as dialect feature maps, vowel history, and rules. Our findings indicate that, when linguistic information is provided, the LLM predictions improve. The human baseline shows that automatically generated transcriptions can be beneficial for such classifications, but also presents opportunities for improvement.

Network traffic classification using self-supervised pre-training models based on Masked Autoencoders (MAE) has demonstrated a huge potential. However, existing methods are confined to isolated byte-level reconstruction of individual flows, lacking adequate perception of the multi-granularity contextual relationship in traffic. To address this limitation, we propose Mean MAE (MMAE), a teacher-student MAE paradigm with flow mixing strategy for building encrypted traffic pre-training model. MMAE employs a self-distillation mechanism for teacher-student interaction, where the teacher provides unmasked flow-level semantic supervision to advance the student from local byte reconstruction to multi-granularity comprehension. To break the information bottleneck in individual flows, we introduce a dynamic Flow Mixing (FlowMix) strategy to replace traditional random masking mechanism. By constructing challenging cross-flow mixed samples with interferences, it compels the model to learn discriminative representations from distorted tokens. Furthermore, we design a Packet-importance aware Mask Predictor (PMP) equipped with an attention bias mechanism that leverages packet-level side-channel statistics to dynamically mask tokens with high semantic density. Numerous experiments on a number of datasets covering encrypted applications, malware, and attack traffic demonstrate that MMAE achieves state-of-the-art performance. The code is available at https://github.com/lx6c78/MMAE

Generative Artificial Intelligence (GenAI) features such as image editing, object removal, and prompt-guided image transformation are increasingly integrated into mobile applications. However, deploying Large Vision Models (LVMs) for such tasks on resource-constrained devices remains challenging due to their high memory and compute requirements. While Low-Rank Adapters (LoRAs) enable parameter-efficient task adaptation, existing Mobile deployment pipelines typically compile separate model binaries for each LoRA + a copy of the foundation model, resulting in redundant storage and increased runtime overhead. In this work, we present a unified framework for enabling multi-task GenAI inference on edge devices using a single shared model. Our key idea is to treat LoRA weights as runtime inputs rather than embedding them into the compiled model graph, allowing dynamic task switching at runtime without recompilation. Then, to support efficient on-device execution, we introduce QUAD (Quantization with Unified Adaptive Distillation), a quantizationaware training strategy that aligns multiple LoRA adapters under a shared quantization profile. We implement the proposed system with a lightweight runtime stack compatible with mobile NPUs and evaluate it across multiple chipsets. Experimental results demonstrate up to 6x and 4x reduction in memory footprint and latency improvements, respectively, while maintaining high visual quality across multiple GenAI tasks.

The Linear Parameter-Varying (LPV) framework enables the construction of surrogate models of complex nonlinear and high-dimensional systems, facilitating efficient stability and performance analysis together with controller design. Despite significant advances in data-driven LPV modelling, existing approaches do not quantify the uncertainty of the obtained LPV models. Consequently, assessing model reliability for analysis and control or detecting operation outside the training regime requires extensive validation and user expertise. This paper proposes a Bayesian approach for the joint estimation of LPV state-space models together with their scheduling, providing a characterization of model uncertainty and confidence bounds on the predicted model response directly from input-output data. Both aleatoric uncertainty due to measurement noise and epistemic uncertainty arising from limited training data and structural bias are considered. The resulting model preserves the LPV structure required for controller synthesis while enabling computationally efficient simulation and uncertainty propagation. The approach is demonstrated on the surrogate modelling of a two-dimensional nonlinear interconnection of mass-spring-damper systems.

We investigate the parameter space of transformer models trained on protein sequence data using a statistical mechanics framework, sampling the loss landscape at varying temperatures by Langevin dynamics to characterize the low-loss manifold and understand the mechanisms underlying the superior performance of transformers in protein structure prediction. We find that, at variance with feedforward networks, the lack of a first--order--like transition in the loss of the transformer produces a range of intermediate temperatures with good learning properties. We show that the parameters of most layers are highly conserved at these temperatures if the dimension of the embedding is optimal, and we provide an operative way to find this dimension. Finally, we show that the attention matrix is more predictive of the contact maps of the protein at higher temperatures and for higher dimensions of the embedding than those optimal for learning.

Modern language models (LMs) must be trained on many orders of magnitude more words of training data than human children receive before they begin to produce useful behavior. Assessing the nature and origins of this "data gap" requires benchmarking LMs on human-scale datasets to understand how linguistic knowledge emerges from children's natural training data. Using transcripts from the BabyView dataset (videos from children ages 6-36 months), we investigate (1) scaling performance at child-scale data regimes, (2) variability in model performance across datasets from different children's experiences and linguistic predictors of dataset quality, and (3) relationships between model and child language learning outcomes. LMs trained on child data show acceptable scaling for grammar tasks, but lower scaling on semantic and world knowledge tasks than models trained on synthetic data; we also observe substantial variability on data from different children. Beyond dataset size, performance is most associated with a combination of distributional and interactional linguistic features, broadly consistent with what makes high-quality input for child language development. Finally, model likelihoods for individual words correlate with children's learning of those words, suggesting that properties of child-directed input may influence both model learning and human language development. Overall, understanding what properties make language data efficient for learning can enable more powerful small-scale language models while also shedding light on human language acquisition.

Encrypted traffic classification is a critical task for network security. While deep learning has advanced this field, the occlusion of payload semantics by encryption severely challenges standard modeling approaches. Most existing frameworks rely on static and homogeneous pipelines that apply uniform parameter sharing and static fusion strategies across all inputs. This one-size-fits-all static design is inherently flawed: by forcing structured headers and randomized payloads into a unified processing pipeline, it inevitably entangles the raw protocol signals with stochastic encryption noise, thereby degrading the fine-grained discriminative features. In this paper, we propose TrafficMoE, a framework that breaks through the bottleneck of static modeling by establishing a Disentangle-Filter-Aggregate (DFA) paradigm. Specifically, to resolve the structural between-components conflict, the architecture disentangles headers and payloads using dual-branch sparse Mixture-of-Experts (MoE), enabling modality-specific modeling. To mitigate the impact of stochastic noise, an uncertainty-aware filtering mechanism is introduced to quantify reliability and selectively suppress high-variance representations. Finally, to overcome the limitations of static fusion, a routing-guided strategy aggregates cross-modality features dynamically, that adaptively weighs contributions based on traffic context. With this DFA paradigm, TrafficMoE maximizes representational efficiency by focusing solely on the most discriminative traffic features. Extensive experiments on six datasets demonstrate TrafficMoE consistently outperforms state-of-the-art methods, validating the necessity of heterogeneity-aware modeling in encrypted traffic analysis. The source code is publicly available at https://github.com/Posuly/TrafficMoE_main.

Conversational systems should generate diverse language forms to interact fluently and accurately with users. In this context, Natural Language Generation (NLG) engines convert Meaning Representations (MRs) into sentences, directly influencing user perception. These MRs usually encode the communicative function (e.g., inform, request, confirm) via DAs and enumerate the semantic content with slot-value pairs. In this work, our objective is to analyse whether providing a task demonstrator to the generator enhances the generations of a fine-tuned model. This demonstrator is an MR-sentence pair extracted from the original dataset that enriches the input at training and inference time. The analysis involves five metrics that focus on different linguistic aspects, and four datasets that differ in multiple features, such as domain, size, lexicon, MR variability, and acquisition process. To the best of our knowledge, this is the first study on dialogue NLG implementing a comparative analysis of the impact of MRs on generation quality across domains, corpus characteristics, and the metrics used to evaluate these generations. Our key insight is that the proposed enriched inputs are effective for complex tasks and small datasets with high variability in MRs and sentences. They are also beneficial in zero-shot settings for any domain. Moreover, the analysis of the metrics shows that semantic metrics capture generation quality more accurately than lexical metrics. In addition, among these semantic metrics, those trained with human ratings can detect omissions and other subtle semantic issues that embedding-based metrics often miss. Finally, the evolution of the metric scores and the excellent results for Slot Accuracy and Dialogue Act Accuracy demonstrate that the generative models present fast adaptability to different tasks and robustness at semantic and communicative intention levels.

Despite rapid advances in large language models (LLMs), their linguistic abilities in low-resource and morphologically rich languages are still not well understood due to limited annotated resources and the absence of standardized evaluation frameworks. This paper presents LLM Probe, a lexicon-based assessment framework designed to systematically evaluate the linguistic skills of LLMs in low-resource language environments. The framework analyzes models across four areas of language understanding: lexical alignment, part-of-speech recognition, morphosyntactic probing, and translation accuracy. To illustrate the framework, we create a manually annotated benchmark dataset using a low-resource Semitic language as a case study. The dataset comprises bilingual lexicons with linguistic annotations, including part-of-speech tags, grammatical gender, and morphosyntactic features, which demonstrate high inter-annotator agreement to ensure reliable annotations. We test a variety of models, including causal language models and sequence-to-sequence architectures. The results reveal notable differences in performance across various linguistic tasks: sequence-to-sequence models generally excel in morphosyntactic analysis and translation quality, whereas causal models demonstrate strong performance in lexical alignment but exhibit weaker translation accuracy. Our results emphasize the need for linguistically grounded evaluation to better understand LLM limitations in low-resource settings. We release LLM Probe and the accompanying benchmark dataset as open-source tools to promote reproducible benchmarking and to support the development of more inclusive multilingual language technologies.

The increasingly wide use of deep machine learning techniques in computational mechanics has significantly accelerated simulations of problems that were considered unapproachable just a few years ago. However, in critical applications such as Digital Twins for engineering or medicine, fast responses are not enough; reliable results must also be provided. In certain cases, traditional deterministic methods may not be optimal as they do not provide a measure of confidence in their predictions or results, especially in inverse problems where the solution may not be unique or the initial data may not be entirely reliable due to the presence of noise, for instance. Classic deep neural networks also lack a clear measure to quantify the uncertainty of their predictions. In this work, we present a variational graph neural network (VGNN) architecture that integrates variational layers into its architecture to model the probability distribution of weights. Unlike computationally expensive full Bayesian networks, our approach strategically introduces variational layers exclusively in the decoder, allowing us to estimate cognitive uncertainty and statistical uncertainty at a relatively lower cost. In this work, we validate the proposed methodology in two cases of solid mechanics: the identification of the value of the elastic modulus with nonlinear distribution in a 2D elastic problem and the location and quantification of the loads applied to a 3D hyperelastic beam, in both cases using only the displacement field of each test as input data. The results show that the model not only recovers the physical parameters with high precision, but also provides confidence intervals consistent with the physics of the problem, as well as being able to locate the position of the applied load and estimate its value, giving a confidence interval for that experiment.

Many reinforcement learning algorithms rely on target networks - lagged copies of the online network - to stabilize training. While effective, this mechanism introduces a fundamental stability-recency tradeoff: slower target updates improve stability but reduce the recency of learning signals, hindering convergence speed. We propose Target-Aligned Reinforcement Learning (TARL), a framework that emphasizes transitions for which the target and online network estimates are highly aligned. By focusing updates on well-aligned targets, TARL mitigates the adverse effects of stale target estimates while retaining the stabilizing benefits of target networks. We provide a theoretical analysis demonstrating that target alignment correction accelerates convergence, and empirically demonstrate consistent improvements over standard reinforcement learning algorithms across various benchmark environments.

Large language models (LLMs) have recently demonstrated impressive performance on complex, multi-step reasoning tasks, especially when post-trained with outcome-rewarded reinforcement learning Guo et al. 2025. However, it has been observed that outcome rewards often overlook flawed intermediate steps, leading to unreliable reasoning steps even when final answers are correct. To address this unreliable reasoning, we propose PRoSFI (Process Reward over Structured Formal Intermediates), a novel reward method that enhances reasoning reliability without compromising accuracy. Instead of generating formal proofs directly, which is rarely accomplishable for a modest-sized (7B) model, the model outputs structured intermediate steps aligned with its natural language reasoning. Each step is then verified by a formal prover. Only fully validated reasoning chains receive high rewards. The integration of formal verification guides the model towards generating step-by-step machine-checkable proofs, thereby yielding more credible final answers. PRoSFI offers a simple and effective approach to training trustworthy reasoning models.

Classical proportional--integral--derivative (PID) control is widely employed in industrial applications; however, achieving higher performance often motivates the adoption of model predictive control (MPC). Although gradient-based methods are the standard for real-time optimization, sampling-based approaches have recently gained attention. In particular, model predictive path integral (MPPI) control enables gradient-free optimization and accommodates non-differentiable models and objective functions. However, directly sampling control input sequences may yield discontinuous inputs and increase the optimization dimensionality in proportion to the prediction horizon. This study proposes MPPI--PID control, which applies MPPI to optimize PID gains at each control step, thereby replacing direct high-dimensional input-sequence optimization with low-dimensional gain-space optimization. This formulation enhances sample efficiency and yields smoother inputs via the PID structure. We also provide theoretical insights, including an information-theoretic interpretation that unifies MPPI and MPPI--PID, an analysis of the effect of optimization dimensionality on sample efficiency, and a characterization of input continuity induced by the PID structure. The proposed method is evaluated on the learning-based path following of a mini forklift using a residual-learning dynamics model that integrates a physical model with a neural network. System identification is performed with real driving data. Numerical path-following experiments demonstrate that MPPI--PID improves tracking performance compared with fixed-gain PID and achieves performance comparable to conventional MPPI while significantly reducing input increments. Furthermore, the proposed method maintains favorable performance even with substantially fewer samples, demonstrating its improved sample efficiency.

Accurate privacy evaluation of textual data remains a critical challenge in privacy-preserving natural language processing. Recent work has shown that large language models (LLMs) can serve as reliable privacy evaluators, achieving strong agreement with human judgments; however, their computational cost and impracticality for processing sensitive data at scale limit real-world deployment. We address this gap by distilling the privacy assessment capabilities of Mistral Large 3 (675B) into lightweight encoder models with as few as 150M parameters. Leveraging a large-scale dataset of privacy-annotated texts spanning 10 diverse domains, we train efficient classifiers that preserve strong agreement with human annotations while dramatically reducing computational requirements. We validate our approach on human-annotated test data and demonstrate its practical utility as an evaluation metric for de-identification systems.

We present Metriplector, a neural architecture primitive in which the input configures an abstract physical system -- fields, sources, and operators -- and the dynamics of that system is the computation. Multiple fields evolve via coupled metriplectic dynamics, and the stress-energy tensor $T^{μν}$, derived from Noether's theorem, provides the readout. The metriplectic formulation admits a natural spectrum of instantiations: the dissipative branch alone yields a screened Poisson equation solved exactly via conjugate gradient; activating the full structure -- including the antisymmetric Poisson bracket -- gives field dynamics for image recognition and language modeling. We evaluate Metriplector across four domains, each using a task-specific architecture built from this shared primitive with progressively richer physics: F1=1.0 on maze pathfinding, generalizing from 15x15 training grids to unseen 39x39 grids; 97.2% exact Sudoku solve rate with zero structural injection; 81.03% on CIFAR-100 with 2.26M parameters; and 1.182 bits/byte on language modeling with 3.6x fewer training tokens than a GPT baseline.

Positive margins are common in head and neck squamous cell carcinoma, yet intraoperative re-resection is often imprecise because margin locations are typically communicated verbally from pathology. We present an all-in-one augmented reality (AR) system that relocalizes positive margins from a resected specimen to the resection bed and visualizes them in situ using HoloLens 2 depth sensing and fully automated markerless surface registration. In a silicone phantom study with six medical trainees, markerless registration achieved target registration errors comparable to a marker-based baseline (median 1.8 mm vs. 1.7 mm; maximum < 4 mm). In a margin relocalization task, AR guidance reduced error from verbal guidance (median 14.2 mm) to a few millimeters (median 3.2 mm), with all AR localizations within 5 mm error. These results support the feasibility of markerless AR margin guidance for more precise intraoperative re-excision.

Memory-augmented Large Language Models (LLMs) are essential for developing capable, long-term AI agents. Recently, applying Reinforcement Learning (RL) to optimize memory operations, such as extraction, updating, and retrieval, has emerged as a highly promising research direction. However, existing implementations remain highly fragmented and task-specific, lacking a unified infrastructure to streamline the integration, training, and evaluation of these complex pipelines. To address this gap, we present MemFactory, the first unified, highly modular training and inference framework specifically designed for memory-augmented agents. Inspired by the success of unified fine-tuning frameworks like LLaMA-Factory, MemFactory abstracts the memory lifecycle into atomic, plug-and-play components, enabling researchers to seamlessly construct custom memory agents via a "Lego-like" architecture. Furthermore, the framework natively integrates Group Relative Policy Optimization (GRPO) to fine-tune internal memory management policies driven by multi-dimensional environmental rewards. MemFactory provides out-of-the-box support for recent cutting-edge paradigms, including Memory-R1, RMM, and MemAgent. We empirically validate MemFactory on the open-source MemAgent architecture using its publicly available training and evaluation data. Across both in-domain and out-of-distribution evaluation sets, MemFactory consistently improves performance over the corresponding base models, with relative gains of up to 14.8%. By providing a standardized, extensible, and easy-to-use infrastructure, MemFactory significantly lowers the barrier to entry, paving the way for future innovations in memory-driven AI agents.

Safe deployment of Large Vision-Language Models (LVLMs) in radiology report generation requires not only accurate predictions but also clinically interpretable indicators of when outputs should be thoroughly reviewed, enabling selective radiologist verification and reducing the risk of hallucinated findings influencing clinical decisions. One intuitive approach to this is verbalized confidence, where the model explicitly states its certainty. However, current state-of-the-art language models are often overconfident, and research on calibration in multimodal settings such as radiology report generation is limited. To address this gap, we introduce ConRad (Confidence Calibration for Radiology Reports), a reinforcement learning framework for fine-tuning medical LVLMs to produce calibrated verbalized confidence estimates alongside radiology reports. We study two settings: a single report-level confidence score and a sentence-level variant assigning a confidence to each claim. Both are trained using the GRPO algorithm with reward functions based on the logarithmic scoring rule, which incentivizes truthful self-assessment by penalizing miscalibration and guarantees optimal calibration under reward maximization. Experimentally, ConRad substantially improves calibration and outperforms competing methods. In a clinical evaluation we show that ConRad's report level scores are well aligned with clinicians' judgment. By highlighting full reports or low-confidence statements for targeted review, ConRad can support safer clinical integration of AI-assistance for report generation.

In the evaluation of attribution quality, the quantitative assessment of explanation legibility is particularly difficult, as it is influenced by varying shapes and internal organization of attributions not captured by simple statistics. To address this issue, we introduce Minimum Spanning Tree Compactness (MST-C), a graph-based structural metric that captures higher-order geometric properties of attributions, such as spread and cohesion. These components are combined into a single score that evaluates compactness, favoring attributions with salient points spread across a small area and spatially organized into few but cohesive clusters. We show that MST-C reliably distinguishes between explanation methods, exposes fundamental structural differences between models, and provides a robust, self-contained diagnostic for explanation compactness that complements existing notions of attribution complexity.

Cosine similarity is often used to measure the similarity of vectors. These vectors might be the representations of neural network models. However, it is not guaranteed that cosine similarity of model representations will tell us anything about model behaviour. In this paper we show that when using a softmax classifier, be it an image classifier or an autoregressive language model, measuring the cosine similarity between label representations (called unembeddings in the paper) does not give any information on the probabilities assigned by the model. Specifically, we prove that for any softmax classifier model, given two label representations, it is possible to make another model which gives the same probabilities for all labels and inputs, but where the cosine similarity between the representations is now either 1 or -1. We give specific examples of models with very high or low cosine simlarity between representations and show how to we can make equivalent models where the cosine similarity is now -1 or 1. This translation ambiguity can be fixed by centering the label representations, however, labels with representations with low cosine similarity can still have high probability for the same inputs. Fixing the length of the representations still does not give a guarantee that high(or low) cosine similarity will give high(or low) probability to the labels for the same inputs. This means that when working with softmax classifiers, cosine similarity values between label representations should not be used to explain model probabilities.

The growing demands in the training and inference of Large Language Models (LLMs) are accelerating the adoption of scale-up systems that extend server shared memory through the use of Compute Express Link (CXL)-based load/store interconnects. Accurate full-system simulation of such architectures remains challenging, as existing tools (all very recent) rely on simplified or non-compliant architectural models, impacting accuracy and usability. We present CXLRAMSim, the first gem5-integrated, full-system simulator that models CXL devices at their correct position on the I/O bus, enabling the use of unmodified Linux kernels and software stack, realistic latency-bandwidth behavior and true interleaving with system DRAM. Our approach provides high-fidelity CXL.mem characterization and captures key challenges such as cache pollution when accessing CXL memory.

We consider three classification systems for distributed decision tasks: With unbounded computation and certificates, defined by Balliu, D'Angelo, Fraigniaud, and Olivetti [JCSS'18], and with (two flavors of) polynomially bounded local computation and certificates, defined in recent works by Aldema Tshuva and Oshman [OPODIS'23], and by Reiter [PODC'24]. The latter two differ in the way they evaluate the polynomial bounds: the former considers polynomials with respect to the size of the graph, while the latter refers to being polynomial in the size of each node's local neighborhood. We start by revisiting decision without certificates. For this scenario, we show that the latter two definitions coincide: roughly, a node cannot know the graph size, and thus can only use a running time dependent on its neighborhood. We then consider decision with certificates. With existential certificates ($Σ_1$-type classes), a larger running time defines strictly larger classes of languages: when it grows from being polynomial in each node's view, through polynomial in the graph's size, and to unbounded, the derived classes strictly contain each other. With universal certificates ($Π_1$-type classes), on the other hand, we prove a surprising incomparability result: having running time bounded by the graph size sometimes allows us to decide languages undecidable even with unbounded certificates. We complement these results with other containment and separation results, which together portray a surprisingly complex lattice of strict containment relations between the classes at the base of the three classification systems.

Survival analysis is crucial for many medical applications but remains challenging for modern machine learning due to limited data, censoring, and the heterogeneity of tabular covariates. While the prior-fitted paradigm, which relies on pretraining models on large collections of synthetic datasets, has recently facilitated tabular foundation models for classification and regression, its suitability for time-to-event modeling remains unclear. We propose a flexible survival data generation framework that defines a rich survival prior with explicit control over covariates and time-event distributions. Building on this prior, we introduce Survival In-Context (SIC), a prior-fitted in-context learning model for survival analysis that is pretrained exclusively on synthetic data. SIC produces individualized survival prediction in a single forward pass, requiring no task-specific training or hyperparameter tuning. Across a broad evaluation on real-world survival datasets, SIC achieves competitive or superior performance compared to classical and deep survival models, particularly in medium-sized data regimes, highlighting the promise of prior-fitted foundation models for survival analysis. The code will be made available upon publication.

The increasing capabilities of machine learning models, such as vision-language and multimodal language models, are placing growing demands on data in automotive systems engineering, making the quality and relevance of collected data enablers for the development and validation of such systems. Traditional Big Data approaches focus on large-scale data collection and offline processing, while Smart Data approaches improve data selection strategies but still rely on centralized and offline post-processing. This paper introduces the concept of Fast Data for automotive systems engineering. The approach shifts data selection and recording onto the vehicle as the data source. By enabling real-time, context-aware decisions on whether and which data should be recorded, data collection can be directly aligned with data quality objectives and collection strategies within a closed-loop. This results in datasets with higher relevance, improved coverage of critical scenarios, and increased information density, while at the same time reducing irrelevant data and associated costs. The proposed approach provides a structured foundation for designing data collection strategies that are aligned with the needs of modern machine learning algorithms. It supports efficient data acquisition and contributes to scalable and cost-effective ML development processes in automotive systems engineering.

We present iPoster, an interactive layout generation framework that empowers users to guide content-aware poster layout design by specifying flexible constraints. iPoster enables users to specify partial intentions within the intention module, such as element categories, sizes, positions, or coarse initial drafts. Then, the generation module instantly generates refined, context-sensitive layouts that faithfully respect these constraints. iPoster employs a unified graph-enhanced diffusion architecture that supports various design tasks under user-specified constraints. These constraints are enforced through masking strategies that precisely preserve user input at every denoising step. A cross content-aware attention module aligns generated elements with salient regions of the canvas, ensuring visual coherence. Extensive experiments show that iPoster not only achieves state-of-the-art layout quality, but offers a responsive and controllable framework for poster layout design with constraints.

This paper presents a Multilingual Vision Large Language Model, named M-MiniGPT4. Our model exhibits strong vision-language understanding (VLU) capabilities across 11 languages. We utilize a mixture of native multilingual and translated data to push the multilingual VLU performance of the MiniGPT4 architecture. In addition, we propose a multilingual alignment training stage that uses parallel text corpora to further enhance the multilingual capabilities of our model. M-MiniGPT4 achieves 36% accuracy on the multilingual MMMU benchmark, outperforming state-of-the-art models in the same weight class, including foundation models released after the majority of this work was completed. We open-source our models, code, and translated datasets to facilitate future research in low-resource and multilingual settings.

Existing methods for quantifying predictive uncertainty in neural networks are either computationally intractable for large language models or require access to training data that is typically unavailable. We derive a lightweight alternative through two approximations: a first-order Taylor expansion that expresses uncertainty in terms of the gradient of the prediction and the parameter covariance, and an isotropy assumption on the parameter covariance. Together, these yield epistemic uncertainty as the squared gradient norm and aleatoric uncertainty as the Bernoulli variance of the point prediction, from a single forward-backward pass through an unmodified pretrained model. We justify the isotropy assumption by showing that covariance estimates built from non-training data introduce structured distortions that isotropic covariance avoids, and that theoretical results on the spectral properties of large networks support the approximation at scale. Validation against reference Markov Chain Monte Carlo estimates on synthetic problems shows strong correspondence that improves with model size. We then use the estimates to investigate when each uncertainty type carries useful signal for predicting answer correctness in question answering with large language models, revealing a benchmark-dependent divergence: the combined estimate achieves the highest mean AUROC on TruthfulQA, where questions involve genuine conflict between plausible answers, but falls to near chance on TriviaQA's factual recall, suggesting that parameter-level uncertainty captures a fundamentally different signal than self-assessment methods.

We study trajectory forecasting under squared loss for time series with weak conditional structure, using highly expressive prediction models. Building on the classical characterization of squared-loss risk minimization, we emphasize regimes in which the conditional expectation of future trajectories is effectively degenerate, leading to trivial Bayes-optimal predictors (flat for prices and zero for returns in standard financial settings). In this regime, increased model expressivity does not improve predictive accuracy but instead introduces spurious trajectory fluctuations around the optimal predictor. These fluctuations arise from the reuse of noise and result in increased prediction variance without any reduction in bias. This provides a process-level explanation for the degradation of Transformerbased forecasts on financial time series. We complement these theoretical results with numerical experiments on high-frequency EUR/USD exchange rate data, analyzing the distribution of trajectory-level forecasting errors. The results show that Transformer-based models yield larger errors than a simple linear benchmark on a large majority of forecasting windows, consistent with the variance-driven mechanism identified by the theory.

Authorship verification (AV), the task of determining whether a questioned text was written by a specific individual, is a critical part of forensic linguistics. While manual authorial impersonation by perpetrators has long been a recognized threat in historical forensic cases, recent advances in large language models (LLMs) raise new challenges, as adversaries may exploit these tools to impersonate another's writing. This study investigates whether prompted LLMs can generate convincing authorial impersonations and whether such outputs can evade existing forensic AV systems. Using GPT-4o as the adversary model, we generated impersonation texts under four prompting conditions across three genres: emails, text messages, and social media posts. We then evaluated these outputs against both non-neural AV methods (n-gram tracing, Ranking-Based Impostors Method, LambdaG) and neural approaches (AdHominem, LUAR, STAR) within a likelihood-ratio framework. Results show that LLM-generated texts failed to sufficiently replicate authorial individuality to bypass established AV systems. We also observed that some methods achieved even higher accuracy when rejecting impersonation texts compared to genuine negative samples. Overall, these findings indicate that, despite the accessibility of LLMs, current AV systems remain robust against entry-level impersonation attempts across multiple genres. Furthermore, we demonstrate that this counter-intuitive resilience stems, at least in part, from the higher lexical diversity and entropy inherent in LLM-generated texts.

While state-of-the-art Handwritten Text Recognition (HTR) models perform well on standard benchmarks, they frequently struggle with writers exhibiting highly specific styles that are underrepresented in the training data. To handle unseen and atypical writers, writer adaptation techniques personalize HTR models to individual handwriting styles. Leading writer adaptation methods require either offline fine-tuning or parameter updates at inference time, both involving gradient computation and backpropagation, which increase computational costs and demand careful hyperparameter tuning. In this work, we propose a novel context-driven HTR framework3 inspired by multimodal in-context learning, enabling inference-time writer adaptation using only a few examples from the target writer without any parameter updates. We further demonstrate the impact of context length, design a compact 8M-parameter CNN-Transformer that enables few-shot in-context adaptation, and show that combining context-driven and standard OCR training strategies leads to complementary improvements. Experiments on IAM and RIMES validate our approach with Character Error Rates of 3.92% and 2.34%, respectively, surpassing all writer-independent HTR models without requiring any parameter updates at inference time.

Minimizing invasive diagnostic procedures to reduce the risk of patient injury and infection is a central goal in medical imaging. And yet, noninvasive diagnosis of perineural invasion (PNI), a critical prognostic factor involving infiltration of tumor cells along the surrounding nerve, still remains challenging, due to the lack of clear and consistent imaging criteria criteria for identifying PNI. To address this challenge, we present NeoNet, an integrated end-to-end 3D deep learning framework for PNI prediction in cholangiocarcinoma that does not rely on predefined image features. NeoNet integrates three modules: (1) NeoSeg, utilizing a Tumor-Localized ROI Crop (TLCR) algorithm; (2) NeoGen, a 3D Latent Diffusion Model (LDM) with ControlNet, conditioned on anatomical masks to generate synthetic image patches, specifically balancing the dataset to a 1:1 ratio; and (3) NeoCls, the final prediction module. For NeoCls, we developed the PNI-Attention Network (PattenNet), which uses the frozen LDM encoder and specialized 3D Dual Attention Blocks (DAB) designed to detect subtle intensity variations and spatial patterns indicative of PNI. In 5-fold cross-validation, NeoNet outperformed baseline 3D models and achieved the highest performance with a maximum AUC of 0.7903.

Medical time-series data captures the dynamic progression of patient conditions, playing a vital role in modern clinical decision support systems. However, real-world clinical data is highly heterogeneous and inconsistently formatted. Furthermore, existing machine learning tools often have steep learning curves and fragmented workflows. Consequently, a significant gap remains between cutting-edge AI technologies and clinical application. To address this, we introduce mtslearn, an end-to-end integrated toolkit specifically designed for medical time-series data. First, the framework provides a unified data interface that automates the parsing and alignment of wide, long, and flat data formats. This design significantly reduces data cleaning overhead. Building on this, mtslearn provides a complete pipeline from data reading and feature engineering to model training and result visualization. Furthermore, it offers flexible interfaces for custom algorithms. Through a modular design, mtslearn simplifies complex data engineering tasks into a few lines of code. This significantly lowers the barrier to entry for clinicians with limited programming experience, empowering them to focus more on exploring medical hypotheses and accelerating the translation of advanced algorithms into real-world clinical practice. mtslearn is publicly available at https://github.com/PKUDigitalHealth/mtslearn.

Mental-health support is increasingly mediated by conversational systems (e.g., LLM-based tools), but users often lack structured ways to audit the quality and potential risks of the support they receive. We introduce CounselReflect, an end-to-end toolkit for auditing mental-health support dialogues. Rather than producing a single opaque quality score, CounselReflect provides structured, multi-dimensional reports with session-level summaries, turn-level scores, and evidence-linked excerpts to support transparent inspection. The system integrates two families of evaluation signals: (i) 12 model-based metrics produced by task-specific predictors, and (ii) rubric-based metrics that extend coverage via a literature-derived library (69 metrics) and user-defined custom metrics, operationalized with configurable LLM judges. CounselReflect is available as a web application, browser extension, and command-line interface (CLI), enabling use in real-time settings as well as at scale. Human evaluation includes a user study with 20 participants and an expert review with 6 mental-health professionals, suggesting that CounselReflect supports understandable, usable, and trustworthy auditing. A demo video and full source code are also provided.

In recent years, advances in underwater networking and multi-agent reinforcement learning (MARL) have significantly expanded multi-autonomous underwater vehicle (AUV) applications in marine exploration and target tracking. However, current MARL-driven cooperative tracking faces three critical challenges: 1) non-stationarity in decentralized coordination, where local policy updates destabilize teammates' observation spaces, preventing convergence; 2) sparse-reward exploration inefficiency from limited underwater visibility and constrained sensor ranges, causing high-variance learning; and 3) water disturbance fragility combined with handcrafted reward dependency that degrades real-world robustness under unmodeled hydrodynamic conditions. To address these challenges, this paper proposes a hierarchical MARL architecture comprising four layers: global training scheduling, multi-agent coordination, local decision-making, and real-time execution. This architecture optimizes task allocation and inter-AUV coordination through hierarchical decomposition. Building on this foundation, we propose the Supervised Diffusion-Aided MARL (SDA-MARL) algorithm featuring three innovations: 1) a dual-decision architecture with segregated experience pools mitigating nonstationarity through structured experience replay; 2) a supervised learning mechanism guiding the diffusion model's reverse denoising process to generate high-fidelity training samples that accelerate convergence; and 3) disturbance-robust policy learning incorporating behavioral cloning loss to guide the Deep Deterministic Policy Gradient network update using high-quality replay actions, eliminating handcrafted reward dependency. The tracking algorithm based on SDA-MARL proposed in this paper achieves superior precision compared to state-of-the-art methods in comprehensive underwater simulations.

Cloud computing has drastically altered the ways in which it is possible to deliver information technologies in a service-led structure, however, this has also been reflected in the cybercrime domain. Cybercrime as a Service is an economic model where a technically skilled actor offers a given cyberattack as an end-to-end service to non-technical actors who pay a subscription fee for said service. The services, which can vary in scope, targets, and delivery modes, include everything from the vulnerability discoveries, delivery of the attack, and the attack itself to financial rewards to the subscriber. In this scoping literature review, we analysed 195 articles from both academic and grey literature with a view of investigating the services articles studied, the methodological approach the how the CaaS model is predicted to develop in the future. Our review indicates that with further commercialisation of the model will further lower the barrier of entry to the cybercrime realm, increase sophistication of the attacks and increase resilience of the service providers and their ecosystem which will result in harder shutdowns of services by the authorities. Furthermore, as the model becomes more accessible, groups such as organised crime groups, extremist actors may use them as well, which may have implications for criminal activity in both cyber and physical domains.

Understanding object affordances is essential for enabling robots to perform purposeful and fine-grained interactions in diverse and unstructured environments. However, existing approaches either rely on retrieval, which is fragile due to sparsity and coverage gaps, or on large-scale models, which frequently mislocalize contact points and mispredict post-contact actions when applied to unseen categories, thereby hindering robust generalization. We introduce Retrieval-Augmented Affordance Prediction (RAAP), a framework that unifies affordance retrieval with alignment-based learning. By decoupling static contact localization and dynamic action direction, RAAP transfers contact points via dense correspondence and predicts action directions through a retrieval-augmented alignment model that consolidates multiple references with dual-weighted attention. Trained on compact subsets of DROID and HOI4D with as few as tens of samples per task, RAAP achieves consistent performance across unseen objects and categories, and enables zero-shot robotic manipulation in both simulation and the real world. Project website: https://github.com/SEU-VIPGroup/RAAP.

Although multimodal large language models (MLLMs) are increasingly deployed in real-world applications, their instruction-following behavior leaves them vulnerable to prompt injection attacks. Existing prompt injection methods predominantly rely on textual prompts or perceptible visual prompts that are observable by human users. In this work, we study imperceptible visual prompt injection against powerful closed-source MLLMs, where adversarial instructions are embedded in the visual modality. Our method adaptively embeds the malicious prompt into the input image via a bounded text overlay to provide semantic guidance. Meanwhile, the imperceptible visual perturbation is iteratively optimized to align the feature representation of the attacked image with those of the malicious visual and textual targets at both coarse- and fine-grained levels. Specifically, the visual target is instantiated as a text-rendered image and progressively refined during optimization to more faithfully represent the desired semantics and improve transferability. Extensive experiments on two multimodal understanding tasks across multiple closed-source MLLMs demonstrate the superior performance of our approach compared to existing methods.

Pre-trained vision-language models (VLMs) exhibit strong zero-shot generalization but remain vulnerable to adversarial perturbations. Existing classification-guided adversarial fine-tuning methods often disrupt pre-trained cross-modal alignment, weakening visual-textual correspondence and degrading zero-shot performance. In this paper, we propose an Alignment-Guided Fine-Tuning (AGFT) framework that enhances zero-shot adversarial robustness while preserving the cross-modal semantic structure. Unlike label-based methods that rely on hard labels and fail to maintain the relative relationships between image and text, AGFT leverages the probabilistic predictions of the original model for text-guided adversarial training, which aligns adversarial visual features with textual embeddings via soft alignment distributions, improving zero-shot adversarial robustness. To address structural discrepancies introduced by fine-tuning, we introduce a distribution consistency calibration mechanism that adjusts the robust model output to match a temperature-scaled version of the pre-trained model predictions. Extensive experiments across multiple zero-shot benchmarks demonstrate that AGFT outperforms state-of-the-art methods while significantly improving zero-shot adversarial robustness.

Accurate forecasting of three-dimensional (3D) cloud fields is important for atmospheric analysis and short-range numerical weather prediction, yet it remains challenging because cloud evolution involves cross-layer interactions, nonlocal dependencies, and multiscale spatiotemporal dynamics. Existing spatiotemporal prediction models based on convolutions, recurrence, or attention often rely on locality-biased representations and therefore struggle to preserve fine cloud structures in volumetric forecasting tasks. To address this issue, we propose QENO, a hybrid quantum-inspired spatiotemporal forecasting framework for 3D cloud fields. The proposed architecture consists of four components: a classical spatiotemporal encoder for compact latent representation, a topology-aware quantum enhancement block for modeling nonlocal couplings in latent space, a dynamic fusion temporal unit for integrating measurement-derived quantum features with recurrent memory, and a decoder for reconstructing future cloud volumes. Experiments on CMA-MESO 3D cloud fields show that QENO consistently outperforms representative baselines, including ConvLSTM, PredRNN++, Earthformer, TAU, and SimVP variants, in terms of MSE, MAE, RMSE, SSIM, and threshold-based detection metrics. In particular, QENO achieves an MSE of 0.2038, an RMSE of 0.4514, and an SSIM of 0.6291, while also maintaining a compact parameter budget. These results indicate that topology-aware hybrid quantum-classical feature modeling is a promising direction for 3D cloud structure forecasting and atmospheric Earth observation data analysis.

Topic modeling seeks to uncover latent semantic structure in text, with LDA providing a foundational probabilistic framework. While recent methods often incorporate external knowledge (e.g., pre-trained embeddings), such reliance limits applicability in emerging or underexplored domains. We introduce \textbf{PRISM}, a corpus-intrinsic method that derives a Dirichlet parameter from word co-occurrence statistics to initialize LDA without altering its generative process. Experiments on text and single cell RNA-seq data show that PRISM improves topic coherence and interpretability, rivaling models that rely on external knowledge. These results underscore the value of corpus-driven initialization for topic modeling in resource-constrained settings. Code is available at: https://github.com/shaham-lab/PRISM.

Large Vision-Language Models have demonstrated exceptional performance in multimodal reasoning and complex scene understanding. However, these models still face significant hallucination issues, where outputs contradict visual facts. Recent research on hallucination mitigation has focused on retraining methods and Contrastive Decoding (CD) methods. While both methods perform well, retraining methods require substantial training resources, and CD methods introduce dual inference overhead. These factors hinder their practical applicability. To address the above issue, we propose a framework for dynamically detecting hallucination representations and performing hallucination-eliminating edits on these representations. With minimal additional computational cost, we achieve state-of-the-art performance on existing benchmarks. Extensive experiments demonstrate the effectiveness of our approach, highlighting its efficient and robust hallucination elimination capability and its powerful controllability over hallucinations. Code is available at https://github.com/ASGO-MM/HIRE

LLM-as-a-Judge (LaaJ) is a novel paradigm in which powerful language models are used to assess the quality, safety, or correctness of generated outputs. While this paradigm has significantly improved the scalability and efficiency of evaluation processes, it also introduces novel security risks and reliability concerns that remain largely unexplored. In particular, LLM-based judges can become both targets of adversarial manipulation and instruments through which attacks are conducted, potentially compromising the trustworthiness of evaluation pipelines. In this paper, we present the first Systematization of Knowledge (SoK) focusing on the security aspects of LLM-as-a-Judge systems. We perform a comprehensive literature review across major academic databases, analyzing 863 works and selecting 45 relevant studies published between 2020 and 2026. Based on this study, we propose a taxonomy that organizes recent research according to the role played by LLM-as-a-Judge in the security landscape, distinguishing between attacks targeting LaaJ systems, attacks performed through LaaJ, defenses leveraging LaaJ for security purposes, and applications where LaaJ is used as an evaluation strategy in security-related domains. We further provide a comparative analysis of existing approaches, highlighting current limitations, emerging threats, and open research challenges. Our findings reveal significant vulnerabilities in LLM-based evaluation frameworks, as well as promising directions for improving their robustness and reliability. Finally, we outline key research opportunities that can guide the development of more secure and trustworthy LLM-as-a-Judge systems.

Constructing Extract-Load-Transform (ELT) pipelines is a labor-intensive data engineering task and a high-impact target for AI automation. On ELT-Bench, the first benchmark for end-to-end ELT pipeline construction, AI agents initially showed low success rates, suggesting they lacked practical utility. We revisit these results and identify two factors causing a substantial underestimation of agent capabilities. First, re-evaluating ELT-Bench with upgraded large language models reveals that the extraction and loading stage is largely solved, while transformation performance improves significantly. Second, we develop an Auditor-Corrector methodology that combines scalable LLM-driven root-cause analysis with rigorous human validation (inter-annotator agreement Fleiss' kappa = 0.85) to audit benchmark quality. Applying this to ELT-Bench uncovers that most failed transformation tasks contain benchmark-attributable errors -- including rigid evaluation scripts, ambiguous specifications, and incorrect ground truth -- that penalize correct agent outputs. Based on these findings, we construct ELT-Bench-Verified, a revised benchmark with refined evaluation logic and corrected ground truth. Re-evaluating on this version yields significant improvement attributable entirely to benchmark correction. Our results show that both rapid model improvement and benchmark quality issues contributed to underestimating agent capabilities. More broadly, our findings echo observations of pervasive annotation errors in text-to-SQL benchmarks, suggesting quality issues are systemic in data engineering evaluation. Systematic quality auditing should be standard practice for complex agentic tasks. We release ELT-Bench-Verified to provide a more reliable foundation for progress in AI-driven data engineering automation.

Standard evaluations of Large language models (LLMs) focus on task performance, offering limited insight into whether correct behavior reflects appropriate underlying mechanisms and risking confirmation bias. We introduce a simple, principled interpretability framework based on token-level perplexity to test whether models rely on linguistically relevant cues. By comparing perplexity distributions over minimal sentence pairs differing in one or a few `pivotal' tokens, our method enables precise, hypothesis-driven analysis without relying on unstable feature-attribution techniques. Experiments on controlled linguistic benchmarks with several open-weight LLMs show that, while linguistically important tokens influence model behavior, they never fully explain perplexity shifts, revealing that models rely on heuristics other than the expected linguistic ones.

In this paper, we propose Extend3D, a training-free pipeline for 3D scene generation from a single image, built upon an object-centric 3D generative model. To overcome the limitations of fixed-size latent spaces in object-centric models for representing wide scenes, we extend the latent space in the $x$ and $y$ directions. Then, by dividing the extended latent space into overlapping patches, we apply the object-centric 3D generative model to each patch and couple them at each time step. Since patch-wise 3D generation with image conditioning requires strict spatial alignment between image and latent patches, we initialize the scene using a point cloud prior from a monocular depth estimator and iteratively refine occluded regions through SDEdit. We discovered that treating the incompleteness of 3D structure as noise during 3D refinement enables 3D completion via a concept, which we term under-noising. Furthermore, to address the sub-optimality of object-centric models for sub-scene generation, we optimize the extended latent during denoising, ensuring that the denoising trajectories remain consistent with the sub-scene dynamics. To this end, we introduce 3D-aware optimization objectives for improved geometric structure and texture fidelity. We demonstrate that our method yields better results than prior methods, as evidenced by human preference and quantitative experiments.

The rapid democratization of prompt-based AI image editing has recently exacerbated the risks associated with malicious content fabrication and misinformation. However, forgery localization methods targeting these emerging editing techniques remain significantly under-explored. To bridge this gap, we first introduce a fully automated mask annotating framework that leverages keypoint alignment and semantic space similarity to generate precise ground-truth masks for edited regions. Based on this framework, we construct PromptForge-350k, a large-scale forgery localization dataset covering four state-of-the-art prompt-based AI image editing models, thereby mitigating the data scarcity in this domain. Furthermore, we propose ICL-Net, an effective forgery localization network featuring a triple-stream backbone and intra-image contrastive learning. This design enables the model to capture highly robust and generalizable forensic features. Extensive experiments demonstrate that our method achieves an IoU of 62.5% on PromptForge-350k, outperforming SOTA methods by 5.1%. Additionally, it exhibits strong robustness against common degradations with an IoU drop of less than 1%, and shows promising generalization capabilities on unseen editing models, achieving an average IoU of 41.5%.

Federated Graph Learning (FGL) has emerged as a powerful paradigm for decentralized training of graph neural networks while preserving data privacy. However, existing FGL methods are predominantly designed for static graphs and rely on parameter averaging or distribution alignment, which implicitly assume that all features are equally transferable across clients, overlooking both the spatial and temporal heterogeneity and the presence of client-specific knowledge in real-world graphs. In this work, we identify that such assumptions create a vicious cycle of spurious representation entanglement, client-specific interference, and negative transfer, degrading generalization performance in Federated Learning over Dynamic Spatio-Temporal Graphs (FSTG). To address this issue, we propose a novel causality-inspired framework named SC-FSGL, which explicitly decouples transferable causal knowledge from client-specific noise through representation-level interventions. Specifically, we introduce a Conditional Separation Module that simulates soft interventions through client conditioned masks, enabling the disentanglement of invariant spatio-temporal causal factors from spurious signals and mitigating representation entanglement caused by client heterogeneity. In addition, we propose a Causal Codebook that clusters causal prototypes and aligns local representations via contrastive learning, promoting cross-client consistency and facilitating knowledge sharing across diverse spatio-temporal patterns. Experiments on five diverse heterogeneity Spatio-Temporal Graph (STG) datasets show that SC-FSGL outperforms state-of-the-art methods.

Lightweight cryptographic primitives are widely deployed in resource-constraint environment, particularly in the Internet of Things (IoT) devices. Due to their public accessibility, these devices are vulnerable to physical attacks, especially fault attacks. Recently, deep learning-based cryptanalytic techniques have demonstrated promising results; however, their application to fault attacks remains limited, particularly for stream ciphers. In this work, we investigate the feasibility of deep learning assisted differential fault attack on three lightweight stream ciphers, namely ACORNv3, MORUSv2 and ATOM, under a relaxed fault model, where a single-bit bit-flipping fault is injected at an unknown location. We train multilayer perceptron (MLP) models to identify the fault locations. Experimental results show that the trained models achieve high identification accuracies of 0.999880, 0.999231 and 0.823568 for ACORNv3, MORUSv2 and ATOM, respectively, and outperform traditional signature-based methods. For the secret recovery process, we introduce a threshold-based method to optimize the number of fault injections required to recover the secret information. The results show that the initial state of ACORN can be recovered with 21 to 34 faults; while MORUS requires 213 to 248 faults, with at most 6 bits of guessing. Both attacks reduce the attack complexity compared to existing works. For ATOM, the results show that it possesses a higher security margin, as majority of state bits in the Non-linear Feedback Shift Register (NFSR) can only be recovered under a precise control model. To the best of our knowledge, this work provides the first experimental results of differential fault attacks on ATOM.

We present a finite-time analysis of two smoothed functional stochastic approximation algorithms for simulation-based optimization. The first is a two time-scale gradient-based method, while the second is a three time-scale Newton-based algorithm that estimates both the gradient and the Hessian of the objective function $J$. Both algorithms involve zeroth order estimates for the gradient/Hessian. Although the asymptotic convergence of these algorithms has been established in prior work, finite-time guarantees of two-timescale stochastic optimization algorithms in zeroth order settings have not been provided previously. For our Newton algorithm, we derive mean-squared error bounds for the Hessian estimator and establish a finite-time bound on $\min\limits_{0 \le m \le T} \mathbb{E}\| \nabla J(θ(m)) \|^2$, showing convergence to first-order stationary points. The analysis explicitly characterizes the interaction between multiple time-scales and the propagation of estimation errors. We further identify step-size choices that balance dominant error terms and achieve near-optimal convergence rates. We also provide corresponding finite-time guarantees for the gradient algorithm under the same framework. The theoretical results are further validated through experiments on the Continuous Mountain Car environment.

Śliwerski, Zimmermann, and Zeller (SZZ) just won the 2026 ACM SIGSOFT Impact Award for asking: When do changes induce fixes? Their paper from 2005 served as the foundation for a wide array of approaches aimed at identifying bug-introducing changes (or commits) from fix commits in software repositories. But even after two decades of progress, the best-performing approach from 2025 yields a modest increase of 10 percentage points in F1-score on the most popular Linux kernel dataset. In this paper, we uncover how and why LLM-based agents can substantially advance the state-of-the-art in identifying bug-introducing commits from fix commits. We propose a simple agentic workflow based on searching a set of candidate commits and find that it raises the F1-score from 0.64 to 0.81 on the most popular Linux kernel dataset, a bigger jump than between the original 2005 method (0.54) and the previous SOTA (0.64). We also uncover why agents are so successful: They derive short greppable patterns from the fix commit diff and message and use them to effectively search and find bug-introducing commits in large candidate sets. Finally, we also discuss how these insights might enable further progress in bug detection, root cause understanding, and repair.

Spacecraft anomaly detection is critical for mission safety, yet deploying sophisticated models on-board presents significant challenges due to hardware constraints. This paper investigates three approaches for spacecraft telemetry anomaly detection -- forecasting & threshold, direct classification, and image classification -- and optimizes them for edge deployment using multi-objective neural architecture optimization on the European Space Agency Anomaly Dataset. Our baseline experiments demonstrate that forecasting & threshold achieves superior detection performance (92.7% Corrected Event-wise F0.5-score (CEF0.5)) [1] compared to alternatives. Through Pareto-optimal architecture optimization, we dramatically reduced computational requirements while maintaining capabilities -- the optimized forecasting & threshold model preserved 88.8% CEF0.5 while reducing RAM usage by 97.1% to just 59 KB and operations by 99.4%. Analysis of deployment viability shows our optimized models require just 0.36-6.25% of CubeSat RAM, making on-board anomaly detection practical even on highly constrained hardware. This research demonstrates that sophisticated anomaly detection capabilities can be successfully deployed within spacecraft edge computing constraints, providing near-instantaneous detection without exceeding hardware limitations or compromising mission safety.

Large language models (LLMs) are increasingly used for medical consultation and health information support. In this high-stakes setting, safety depends not only on medical knowledge, but also on how models respond when patient inputs are unclear, inconsistent, or misleading. However, most existing medical LLM evaluations assume idealized and well-posed patient questions, which limits their realism. In this paper, we study challenging patient behaviors that commonly arise in real medical consultations and complicate safe clinical reasoning. We define four clinically grounded categories of such behaviors: information contradiction, factual inaccuracy, self-diagnosis, and care resistance. For each behavior, we specify concrete failure criteria that capture unsafe responses. Building on four existing medical dialogue datasets, we introduce CPB-Bench (Challenging Patient Behaviors Benchmark), a bilingual (English and Chinese) benchmark of 692 multi-turn dialogues annotated with these behaviors. We evaluate a range of open- and closed-source LLMs on their responses to challenging patient utterances. While models perform well overall, we identify consistent, behavior-specific failure patterns, with particular difficulty in handling contradictory or medically implausible patient information. We also study four intervention strategies and find that they yield inconsistent improvements and can introduce unnecessary corrections. We release the dataset and code.

Deep reinforcement learning has demonstrated remarkable success across various domains. However, the tight coupling between training and inference processes makes accelerating DRL training an essential challenge for DRL optimization. Two key issues hinder efficient DRL training: (1) the significant variation in computational intensity across different DRL algorithms and even among operations within the same algorithm complicates hardware platform selection, while (2) DRL's wide dynamic range could lead to substantial reward errors with conventional FP16+FP32 mixed-precision quantization. While existing work has primarily focused on accelerating DRL for specific computing units or optimizing inference-stage quantization, we propose AP-DRL to address the above challenges. AP-DRL is an automatic task partitioning framework that harnesses the heterogeneous architecture of AMD Versal ACAP (integrating CPUs, FPGAs, and AI Engines) to accelerate DRL training through intelligent hardware-aware optimization. Our approach begins with bottleneck analysis of CPU, FPGA, and AIE performance across diverse DRL workloads, informing the design principles for AP-DRL's inter-component task partitioning and quantization optimization. The framework then addresses the challenge of platform selection through design space exploration-based profiling and ILP-based partitioning models that match operations to optimal computing units based on their computational characteristics. For the quantization challenge, AP-DRL employs a hardware-aware algorithm coordinating FP32 (CPU), FP16 (FPGA/DSP), and BF16 (AI Engine) operations by leveraging Versal ACAP's native support for these precision formats. Comprehensive experiments indicate that AP-DRL can achieve speedup of up to 4.17$\times$ over programmable logic and up to 3.82$\times$ over AI Engine baselines while maintaining training convergence.

The introduction of dynamic power management strategies such as clock gating and power gating in dataflow networks has been shown to provide significant energy savings when applied during idle times. However, these strategies can also degrade throughput due to shutdown and wake-up delays. Such throughput degradations might be particularly detrimental to signal processing systems that require a guaranteed throughput. As a solution, this paper first contributes a linear-program formulation for finding a periodic maximal-throughput schedule of a given so-called self-powering dataflow network where actors, realized in hardware, are allowed to go to sleep whenever not being enabled to fire. Depending on which actors are allowed to power down, tradeoffs between throughput and energy savings can be obtained. As a second contribution, we propose a mixed-integer-linear-program formulation to determine a periodic schedule that satisfies a given throughput while minimizing the overall energy per period by identifying a respective set of actors that is allowed to power down in phases of idleness and which rather not. Finally, as a third contribution, we propose a multi-objective design-space exploration strategy called "Hop and Skip" to efficiently explore the Pareto front of energy and throughput solutions. Experimental evaluations on a set of existing benchmarks and randomly generated graphs witness significant exploration time reductions over a brute-force sweep. Finally, a real-world case study is elaborated, and we report on achievable energy savings and throughputs of the related dataflow network where (a) all actors are always-active, (b) all actors are self-powered, and (c) all optimal energy and throughput tradeoff points as found by the proposed design-space exploration strategy.

Prior authorization remains one of the most burdensome administrative processes in U.S. healthcare, consuming billions of dollars and thousands of physician hours each year. While large language models have shown promise across clinical text tasks, their ability to produce submission-ready prior authorization letters has received only limited attention, with existing work confined to single-case demonstrations rather than structured multi-scenario evaluation. We assessed three commercially available LLMs (GPT-4o, Claude Sonnet 4.5, and Gemini 2.5 Pro) across 45 physician-validated synthetic scenarios spanning rheumatology, psychiatry, oncology, cardiology, and orthopedics. All three models generated letters with strong clinical content: accurate diagnoses, well-structured medical necessity arguments, and thorough step therapy documentation. However, a secondary analysis of real-world administrative requirements revealed consistent gaps that clinical scoring alone did not capture, including absent billing codes, missing authorization duration requests, and inadequate follow-up plans. These findings reframe the question: the challenge for clinical deployment is not whether LLMs can write clinically adequate letters, but whether the systems built around them can supply the administrative precision that payer workflows require.

Tree ensembles (TEs) find a multitude of practical applications. They represent one of the most general and accurate classes of machine learning methods. While they are typically quite concise in representation, their operation remains inscrutable to human decision makers. One solution to build trust in the operation of TEs is to automatically identify explanations for the predictions made. Evidently, we can only achieve trust using explanations, if those explanations are rigorous, that is truly reflect properties of the underlying predictor they explain This paper investigates the computation of rigorously-defined, logically-sound explanations for the concrete case of two well-known examples of tree ensembles, namely random forests and boosted trees.

AI evaluation suites often report many scores without checking whether those scores carry independent information. We introduce Effective Dimensionality (ED), the participation ratio of a centered benchmark-score spectrum, as a fast, population-conditional upper-bound diagnostic of measurement breadth. Applied at per-instance granularity to 22 benchmarks across 8 domains and more than 8,400 model evaluations, ED reveals substantial redundancy: the six-score Open LLM Leaderboard behaves like roughly two effective measurement axes (ED = 1.7), BBH and MMLU-Pro are near-interchangeable (rho = 0.96, stable across seven subpopulations), and measurement breadth varies more than 20x across current benchmarks. We show that relative ED rankings are stable under matched-dimension controls and that ED can flag redundant suite components, monitor performance-conditional compression, and guide benchmark maintenance. Because binary spectra overestimate absolute latent dimensionality, we interpret ED as a screening statistic rather than a literal factor count and complement it with null, reliability, and saturation analyses. We provide a 22-benchmark reference atlas and a four-step diagnostic workflow that benchmark maintainers can run with a score matrix and a few lines of code.

The rapid progress of generative adversarial networks (GANs) and diffusion models has enabled the creation of synthetic faces that are increasingly difficult to distinguish from real images. This progress, however, has also amplified the risks of misinformation, fraud, and identity abuse, underscoring the urgent need for detectors that remain robust across diverse generative models. In this work, we introduce Counterfeit Image Pattern High-level Examination via Representation(CIPHER), a deepfake detection framework that systematically reuses and fine-tunes discriminators originally trained for image generation. By extracting scale-adaptive features from ProGAN discriminators and temporal-consistency features from diffusion models, CIPHER captures generation-agnostic artifacts that conventional detectors often overlook. Through extensive experiments across nine state-of-the-art generative models, CIPHER demonstrates superior cross-model detection performance, achieving up to 74.33% F1-score and outperforming existing ViT-based detectors by over 30% in F1-score on average. Notably, our approach maintains robust performance on challenging datasets where baseline methods fail, with up to 88% F1-score on CIFAKE compared to near-zero performance from conventional detectors. These results validate the effectiveness of discriminator reuse and cross-model fine-tuning, establishing CIPHER as a promising approach toward building more generalizable and robust deepfake detection systems in an era of rapidly evolving generative technologies.

We introduce Nomad, a system for autonomous data exploration and insight discovery. Given a corpus of documents, databases, or other data sources, users rarely know the full set of questions, hypotheses, or connections that could be explored. As a result, query-driven question answering and prompt-driven deep-research systems remain limited by human framing and often fail to cover the broader insight space. Nomad addresses this problem with an exploration-first architecture. It constructs an explicit Exploration Map over the domain and systematically traverses it to balance breadth and depth. It generates and selects hypotheses and investigates them with an explorer agent that can use document search, web search, and database tools. Candidate insights are then checked by an independent verifier before entering a reporting pipeline that produces cited reports and higher-level meta-reports. We also present a comprehensive evaluation framework for autonomous discovery systems that measures trustworthiness, report quality, and diversity. Using a corpus of selected UN and WHO reports, we show that \nomad{} produces more trustworthy and higher-quality reports than baselines, while also producing more diverse insights over several runs. Nomad is a step toward autonomous systems that not only answer user questions or conduct directed research, but also discover which questions, research directions, and insights are worth surfacing in the first place.

Narrative analysis is a cornerstone of qualitative research. One leading approach is the Labovian model, but its application is labor-intensive, requiring a holistic, recursive interpretive process that moves back and forth between individual parts of the transcript and the transcript as a whole. Existing Labovian datasets are available only in English, which differs markedly from Japanese in terms of grammar and discourse conventions. To address this gap, we introduce the first systematic guidelines for Labovian narrative analysis of Japanese narrative data. Our guidelines retain all six Labovian categories and extend the framework by providing explicit rules for clause segmentation tailored to Japanese constructions. In addition, our guidelines cover a broader range of clause types and narrative types. Using these guidelines, annotators achieved high agreement in clause segmentation (Fleiss' kappa = 0.80) and moderate agreement in two structural classification tasks (Krippendorff's alpha = 0.41 and 0.45, respectively), one of which is slightly higher than that found in prior work despite the use of finer-grained distinctions. This paper describes the Labovian model, the proposed guidelines, the annotation process, and their utility. It concludes by discussing the challenges encountered during the annotation process and the prospects for developing a larger dataset for structural narrative analysis in Japanese qualitative research.

This paper introduces the L-ReLF (Low-Resource Lexical Framework), a novel, reproducible methodology for creating high-quality, structured lexical datasets for underserved languages. The lack of standardized terminology, exemplified by Moroccan Darija, poses a critical barrier to knowledge equity in platforms like Wikipedia, often forcing editors to rely on inconsistent, ad-hoc methods to create new words in their language. Our research details the technical pipeline developed to overcome these challenges. We systematically address the difficulties of working with low-resource data, including source identification, utilizing Optical Character Recognition (OCR) despite its bias towards Modern Standard Arabic, and rigorous post-processing to correct errors and standardize the data model. The resulting structured dataset is fully compatible with Wikidata Lexemes, serving as a vital technical resource. The L-ReLF methodology is designed for generalizability, offering other language communities a clear path to build foundational lexical data for downstream NLP applications, such as Machine Translation and morphological analysis.

Esperanto is a widespread constructed language, known for its regular grammar and productive word formation. Besides having substantial resources available thanks to its online community, it remains relatively underexplored in the context of modern machine translation (MT) approaches. In this work, we present the first comprehensive evaluation of open-source MT systems for Esperanto, comparing rule-based systems, encoder-decoder models, and LLMs across model sizes. We evaluate translation quality across six language directions involving English, Spanish, Catalan, and Esperanto using multiple automatic metrics as well as human evaluation. Our results show that the NLLB family achieves the best performance in all language pairs, followed closely by our trained compact models and a fine-tuned general-purpose LLM. Human evaluation confirms this trend, with NLLB translations preferred in approximately half of the comparisons, although noticeable errors remain. In line with Esperanto's tradition of openness and international collaboration, we release our code and best-performing models publicly.

Entity linking is the task of associating linguistic expressions with entries in a knowledge base that represent real-world entities and concepts. Language resources for this task have primarily been developed for English, and the resources available for evaluating Japanese systems remain limited. In this study, we develop a corpus design policy for the entity linking task and construct an annotated corpus for training and evaluating Japanese entity linking systems, with rich coverage of linguistic expressions referring to entities that are specific to Japan. Evaluation of inter-annotator agreement confirms the high consistency of the annotations in the corpus, and a preliminary experiment on entity disambiguation based on string matching suggests that the corpus contains a substantial number of non-trivial cases, supporting its potential usefulness as an evaluation benchmark.

The embodied learning of human motor control requires whole-body neuro-actuated musculoskeletal dynamics, while the internal muscle-driven processes underlying movement remain inaccessible to direct measurement. Computational modeling offers an alternative, but inverse dynamics methods struggled to resolve redundant control from observed kinematics in the high-dimensional, over-actuated system. Forward imitation approaches based on deep reinforcement learning exhibited inadequate tracking performance due to the curse of dimensionality in both control and reward design. Here we introduce a large-scale parallel musculoskeletal computation framework for biomechanically grounded whole-body motion reproduction. By integrating large-scale parallel GPU simulation with adversarial reward aggregation and value-guided flow exploration, the MS-Emulator framework overcomes key optimization bottlenecks in high-dimensional reinforcement learning for musculoskeletal control, which accurately reproduces a broad repertoire of motions in a whole-body human musculoskeletal system actuated by approximately 700 muscles. It achieved high joint angle accuracy and body position alignment for highly dynamic tasks such as dance, cartwheel, and backflip. The framework was also used to explore the musculoskeletal control solution space, identifying distinct musculoskeletal control policies that converge to nearly identical external kinematic and mechanical measurements. This work establishes a tractable computational route to analyzing the specificity and diversity underlying human embodied control of movement. Project page: https://lnsgroup.cc/research/MS-Emulator.

Backdoor attacks on federated learning (FL) are most often evaluated with synthetic corner patches or out-of-distribution (OOD) patterns that are unlikely to arise in practice. In this paper, we revisit the backdoor threat to standard FL (a single global model) under a more realistic setting where triggers must be semantically meaningful, in-distribution, and visually plausible. We propose SABLE, a Semantics-Aware Backdoor for LEarning in federated settings, which constructs natural, content-consistent triggers (e.g., semantic attribute changes such as sunglasses) and optimizes an aggregation-aware malicious objective with feature separation and parameter regularization to keep attacker updates close to benign ones. We instantiate SABLE on CelebA hair-color classification and the German Traffic Sign Recognition Benchmark (GTSRB), poisoning only a small, interpretable subset of each malicious client's local data while otherwise following the standard FL protocol. Across heterogeneous client partitions and multiple aggregation rules (FedAvg, Trimmed Mean, MultiKrum, and FLAME), our semantics-driven triggers achieve high targeted attack success rates while preserving benign test accuracy. These results show that semantics-aligned backdoors remain a potent and practical threat in federated learning, and that robustness claims based solely on synthetic patch triggers can be overly optimistic.

Real-time, deep learning-based vocal denoising has seen significant progress over the past few years, demonstrating the capability of artificial intelligence in preserving the naturalness of the voice while increasing the signal-to-noise ratio (SNR). However, many deep learning approaches have high amounts of latency and require long frames of context, making them difficult to configure for live applications. To address these challenges, we propose a sigmoid-driven ideal ratio mask trained with a spectral loss to encourage an increased SNR and maximized perceptual quality of the voice. The proposed model uses a band-grouped encoder-decoder architecture with frequency attention and achieves a total latency of less than 10,ms, with PESQ-WB improvements of 0.21 on stationary noise and 0.12 on nonstationary noise.

Smartphone GUI agents execute tasks by operating directly on app interfaces, offering a path to broad capability without deep system integration. However, real-world smartphone use is highly personalized: users adopt diverse workflows and preferences, challenging agents to deliver customized assistance rather than generic solutions. Existing GUI agent benchmarks cannot adequately capture this personalization dimension due to sparse user-specific data and the lack of fine-grained evaluation metrics. To address this gap, we present PSPA-Bench, the benchmark dedicated to evaluating personalization in smartphone GUI agents. PSPA-Bench comprises over 12,855 personalized instructions aligned with real-world user behaviors across 10 representative daily-use scenarios and 22 mobile apps, and introduces a structure-aware process evaluation method that measures agents' personalized capabilities at a fine-grained level. Through PSPA-Bench, we benchmark 11 state-of-the-art GUI agents. Results reveal that current methods perform poorly under personalized settings, with even the strongest agent achieving limited success. Our analysis further highlights three directions for advancing personalized GUI agents: (1) reasoning-oriented models consistently outperform general LLMs, (2) perception remains a simple yet critical capability, and (3) reflection and long-term memory mechanisms are key to improving adaptation. Together, these findings establish PSPA-Bench as a foundation for systematic study and future progress in personalized GUI agents.

Robotic reproduction of oil paintings using soft brushes and pigments requires force-sensitive control of deformable tools, prediction of brushstroke effects, and multi-step stroke planning, often without human step-by-step demonstrations or faithful simulators. Given only a sequence of target oil painting images, can a robot infer and execute the stroke trajectories, forces, and colors needed to reproduce it? We present IMPASTO, a robotic oil-painting system that integrates learned pixel dynamics models with model-based planning. The dynamics models predict canvas updates from image observations and parameterized stroke actions; a receding-horizon model predictive control optimizer then plans trajectories and forces, while a force-sensitive controller executes strokes on a 7-DoF robot arm. IMPASTO integrates low-level force control, learned dynamics models, and high-level closed-loop planning, learns solely from robot self-play, and approximates human artists' single-stroke datasets and multi-stroke artworks, outperforming baselines in reproduction accuracy. Project website: https://impasto-robopainting.github.io/

The precise fusion of computational fluid dynamic (CFD) data, wind tunnel tests data, and flight tests data in aerodynamic area is essential for obtaining comprehensive knowledge of both localized flow structures and global aerodynamic trends across the entire flight envelope. However, existing methodologies often struggle to balance high-resolution local fidelity with wide-range global dependency, leading to either a loss of sharp discontinuities or an inability to capture long-range topological correlations. We propose Local-Global Fusion Network (LGFNet) for multi-scale feature decomposition to extract this dual-natured aerodynamic knowledge. To this end, LGFNet combines a spatial perception layer that integrates a sliding window mechanism with a relational reasoning layer based on self-attention, simultaneously reinforcing the continuity of fine-grained local features (e.g., shock waves) and capturing long-range flow information. Furthermore, the fidelity gap delta learning (FGDL) strategy is proposed to treat CFD data as a "low-frequency carrier" to explicitly approximate nonlinear discrepancies. This approach prevents unphysical smoothing while inheriting the foundational physical trends from the simulation baseline. Experiments demonstrate that LGFNet achieves state-of-the-art (SOTA) performance in both accuracy and uncertainty reduction across diverse aerodynamic scenarios.

Improving the code generation capabilities of large language models (LLMs) typically relies on supervised fine-tuning or preference optimization, both of which require costly external resources such as powerful teacher models or reliable test units. However, in real-world scenarios, it is much harder to obtain reference solutions and test oracles than problem descriptions and test inputs. In this paper, we tackle a challenging yet realistic question: Can a code language model improve itself without access to a superior teacher and a test oracle? To answer this, we propose ConSelf, a self-improving approach built upon two key ideas. First, we introduce code semantic entropy, a novel metric that measures problem-level uncertainty by assessing the functional diversity of program behaviors, enabling a curriculum construction with the most learnable problems. Second, we present consensus-driven direct preference optimization (Con-DPO), a preference-based fine-tuning method that weights each preference pair by its behavioral consensus, thereby mitigating the impact of noisy self-generated supervision. Experiments on various benchmarks and backbone LLMs demonstrate that ConSelf significantly outperforms baselines, validating the effectiveness of semantic entropy-based curriculum construction and consensus-driven optimization in improving code generation without external supervision.

Composed Image Retrieval (CIR) uses a reference image and a modification text as a query to retrieve a target image satisfying the requirement of ``modifying the reference image according to the text instructions''. However, existing CIR methods face two limitations: (1) frequency bias leading to ``Rare Sample Neglect'', and (2) susceptibility of similarity scores to interference from hard negative samples and noise. To address these limitations, we confront two key challenges: asymmetric rare semantic localization and robust similarity estimation under hard negative samples. To solve these challenges, we propose the Modification frEquentation-rarity baLance neTwork MELT. MELT assigns increased attention to rare modification semantics in multimodal contexts while applying diffusion-based denoising to hard negative samples with high similarity scores, enhancing multimodal fusion and matching. Extensive experiments on two CIR benchmarks validate the superior performance of MELT. Codes are available at https://github.com/luckylittlezhi/MELT.

The fast pace of modern AI is rapidly transforming traditional industrial systems into vast, intelligent and potentially unmanned autonomous operational environments driven by AI-based solutions. These solutions leverage various forms of machine learning, reinforcement learning, and generative AI. The introduction of such smart capabilities has pushed the envelope in multiple industrial domains, enabling predictive maintenance, optimized performance, and streamlined workflows. These solutions are often deployed across the Industrial Internet of Things (IIoT) and supported by the Edge-Fog-Cloud computing continuum to enable urgent (i.e., real-time or near real-time) decision-making. Despite the current trend of aggressively adopting these smart industrial solutions to increase profit, quality, and efficiency, large-scale integration and deployment also bring serious hazards that if ignored can undermine the benefits of smart industries. These hazards include unforeseen interoperability side-effects and heightened vulnerability to cyber threats, particularly in environments operating with a plethora of heterogeneous IIoT systems. The goal of this study is to shed light on the potential consequences of industrial smartness, with a particular focus on security implications, including vulnerabilities, side effects, and cyber threats. We distinguish software-level downsides stemming from both traditional AI solutions and generative AI from those originating in the infrastructure layer, namely IIoT and the Edge-Cloud continuum. At each level, we investigate potential vulnerabilities, cyber threats, and unintended side effects. As industries continue to become smarter, understanding and addressing these downsides will be crucial to ensure secure and sustainable development of smart industrial systems.

Social and personal decisions in relational domains such as matchmaking are deeply entwined with cultural norms and historical hierarchies, and can potentially be shaped by algorithmic and AI-mediated assessments of compatibility, acceptance, and stability. In South Asian contexts, caste remains a central aspect of marital decision-making, yet little is known about how contemporary large language models (LLMs) reproduce or disrupt caste-based stratification in such settings. In this work, we conduct a controlled audit of caste bias in LLM-mediated matchmaking evaluations using real-world matrimonial profiles. We vary caste identity across Brahmin, Kshatriya, Vaishya, Shudra, and Dalit, and income across five buckets, and evaluate five LLM families (GPT, Gemini, Llama, Qwen, and BharatGPT). Models are prompted to assess profiles along dimensions of social acceptance, marital stability, and cultural compatibility. Our analysis reveals consistent hierarchical patterns across models: same-caste matches are rated most favorably, with average ratings up to 25% higher (on a 10-point scale) than inter-caste matches, which are further ordered according to traditional caste hierarchy. These findings highlight how existing caste hierarchies are reproduced in LLM decision-making and underscore the need for culturally grounded evaluation and intervention strategies in AI systems deployed in socially sensitive domains, where such systems risk reinforcing historical forms of exclusion.

A critical gap exists between the general-purpose visual understanding of state-of-the-art physical AI models and the specialized perceptual demands of structured real-world deployment environments. We present PRISM, a 270K-sample multi-view video supervised fine-tuning (SFT) corpus for embodied vision-language-models (VLMs) in real-world retail environments. PRISM is motivated by a simple observation - physical AI systems fail not because of poor visual recognition, but because they do not understand space, physical dynamics and embodied action well enough to operate reliably in the world. To this end, PRISM is grounded in a novel three-dimensional knowledge ontology that spans spatial knowledge, temporal and physical knowledge, and embodied action knowledge. It covers 20+ capability probes across four evaluation dimensions - Embodied Reasoning (ER), Common Sense (CS), Spatial Perception (SP), and Intuitive Physics (IP), and to our knowledge, PRISM is the first dataset to instantiate all three knowledge dimensions within a single real-world deployment domain. The corpus captures data from egocentric, exocentric and 360° viewpoints across five supermarket locations and includes open-ended, chain-of-thought, and multiple-choice supervision. At 4 fps, PRISM spans approximately 11.8M video frames and approximately 730M tokens, placing it among the largest domain-specific video SFT corpora. Fine-tuning on PRISM reduces the error rate across all 20+ probes by 66.6% over the pre-trained baseline, with significant gains in embodied action understanding where the accuracy improves by 36.4%. Our results suggest that ontology-structured, domain specific SFT can meaningfully strengthen embodied VLMs for real-world settings. The PRISM dataset and more details are available at https://dreamvu.ai/prism

The low-rank matrix recovery problem seeks to reconstruct an unknown $n_1 \times n_2$ rank-$r$ matrix from $m$ linear measurements, where $m\ll n_1n_2$. This problem has been extensively studied over the past few decades, leading to a variety of algorithms with solid theoretical guarantees. Among these, gradient descent based non-convex methods have become particularly popular due to their computational efficiency. However, these methods typically suffer from two key limitations: a sub-optimal sample complexity of $O((n_1 + n_2)r^2)$ and an iteration complexity of $O(κ\log(1/ε))$ to achieve $ε$-accuracy, resulting in slow convergence when the target matrix is ill-conditioned. Here, $κ$ denotes the condition number of the unknown matrix. Recent studies show that a preconditioned variant of GD, known as scaled gradient descent (ScaledGD), can significantly reduce the iteration complexity to $O(\log(1/ε))$. Nonetheless, its sample complexity remains sub-optimal at $O((n_1 + n_2)r^2)$. In contrast, a delicate virtual sequence technique demonstrates that the standard GD in the positive semidefinite (PSD) setting achieves the optimal sample complexity $O((n_1 + n_2)r)$, but converges more slowly with an iteration complexity $O(κ^2 \log(1/ε))$. In this paper, through a more refined analysis, we show that ScaledGD achieves both the optimal sample complexity $O((n_1 + n_2)r)$ and the improved iteration complexity $O(\log(1/ε))$. Notably, our results extend beyond the PSD setting to general low-rank matrix recovery problem. Numerical experiments further validate that ScaledGD accelerates convergence for ill-conditioned matrices with the optimal sampling complexity.

High-density through-substrate vias (TSVs) enable 2.5D/3D heterogeneous integration but introduce significant signal-integrity and thermal-reliability challenges due to electrical coupling, insertion loss, and self-heating. Conventional full-wave finite-element method (FEM) simulations provide high accuracy but become computationally prohibitive for large design-space exploration. This work presents a scalable electro-thermal modeling and optimization framework that combines physics-informed analytical modeling, graph neural network (GNN) surrogates, and full-wave sign-off validation. A multi-conductor analytical model computes broadband S-parameters and effective anisotropic thermal conductivities of TSV arrays, achieving $5\%-10\%$ relative Frobenius error (RFE) across array sizes up to $15x15$. A physics-informed GNN surrogate (TSV-PhGNN), trained on analytical data and fine-tuned with HFSS simulations, generalizes to larger arrays with RFE below $2\%$ and nearly constant variance. The surrogate is integrated into a multi-objective Pareto optimization framework targeting reflection coefficient, insertion loss, worst-case crosstalk (NEXT/FEXT), and effective thermal conductivity. Millions of TSV configurations can be explored within minutes, enabling exhaustive layout and geometric optimization that would be infeasible using FEM alone. Final designs are validated with Ansys HFSS and Mechanical, showing strong agreement. The proposed framework enables rapid electro-thermal co-design of TSV arrays while reducing per-design evaluation time by more than six orders of magnitude.

Linear dynamical systems are fully characterized by their eigenspectra, accessible directly from the generator of the dynamics. For nonlinear systems governed by partial differential equations, no equivalent theory exists. We introduce Lie Generator Network-Koopman (LGN-KM), a neural operator that lifts nonlinear dynamics into a linear latent space and learns the continuous-time Koopman generator ($L_k$) through a decomposition $L_k = S - D_k$, where $S$ is skew-symmetric representing conservative inter-modal coupling, and $D_k$ is a positive-definite diagonal encoding modal dissipation. This architectural decomposition enforces stability and enables interpretability through direct spectral access to the learned dynamics. On two-dimensional Navier--Stokes turbulence, the generator recovers the known dissipation scaling and a complete multi-branch dispersion relation from trajectory data alone with no physics supervision. Independently trained models at different flow regimes recover matched gauge-invariant spectral structure, exposing a gauge freedom in the Koopman lifting. Because the generator is provably stable, it enables guaranteed long-horizon stability, continuous-time evaluation at arbitrary time, and physics-informed cross-viscosity model transfer.

Grokking in modular arithmetic has established itself as the quintessential fruit fly experiment, serving as a critical domain for investigating the mechanistic origins of model generalization. Despite its significance, existing research remains narrowly focused on specific local circuits or optimization tuning, largely overlooking the global structural evolution that fundamentally drives this phenomenon. We propose that grokking originates from a spontaneous simplification of internal model structures governed by the principle of parsimony. We integrate causal, spectral, and algorithmic complexity measures alongside Singular Learning Theory to reveal that the transition from memorization to generalization corresponds to the physical collapse of redundant manifolds and deep information compression, offering a novel perspective for understanding the mechanisms of model overfitting and generalization.

Item Price Elasticity is used to quantify the responsiveness of consumer demand to changes in item prices, enabling businesses to create pricing strategies and optimize revenue management. Sectors such as store retail, e-commerce, and consumer goods rely on elasticity information derived from historical sales and pricing data. This elasticity provides an understanding of purchasing behavior across different items, consumer discount sensitivity, and demand elastic departments. This information is particularly valuable for competitive markets and resource-constrained businesses decision making which aims to maximize profitability and market share. Price elasticity also uncovers historical shifts in consumer responsiveness over time. In this paper, we model item-level price elasticity using large-scale transactional datasets, by proposing a novel elasticity estimation framework which has the capability to work in an absence of treatment control setting. We test this framework by using Machine learning based algorithms listed below, including our newly proposed Monodense deep neural network. (1) Monodense-DL network -- Hybrid neural network architecture combining embedding, dense, and Monodense layers (2) DML -- Double machine learning setting using regression models (3) LGBM -- Light Gradient Boosting Model We evaluate our model on multi-category retail data spanning millions of transactions using a back testing framework. Experimental results demonstrate the superiority of our proposed neural network model within the framework compared to other prevalent ML based methods listed above.

Preference-based alignment objectives have been widely adopted, from RLHF-style pairwise learning in large language models to emerging applications in recommender systems. Yet, existing work rarely examines how Direct Preference Optimization (DPO) behaves under implicit feedback, where unobserved items are not reliable negatives. We conduct systematic experiments on multimodal sequential recommendation to compare common negative-selection strategies and their interaction with DPO training. Our central finding is that a simple modification, replacing deterministic hard negatives with stochastic sampling from a dynamic top-K candidate pool, consistently improves ranking performance. We attribute its effectiveness to two factors: (1) reducing erroneous suppressive gradients caused by false negatives, and (2) retaining informative hard signals while smoothing optimization via controlled stochasticity. With an optional sparse Mixture-of-Experts encoder for efficient capacity scaling, RoDPO achieves up to 5.25% NDCG@5 on three Amazon benchmarks, with nearly unchanged inference cost.

Vision-Language Models (VLMs) have demonstrated strong capabilities across a wide range of multimodal tasks. However, recent studies have shown that VLMs, such as CLIP, perform poorly in understanding negation expressions, which are common in natural language. In this work, we propose Omni-NegCLIP, a fine-tuned CLIP model that improves CLIP's understanding of two types of negation, namely presence-based negation and absence-based negation, which correspond to negated expressions of objects that are actually present in an image and those that may plausibly exist in an image but are in fact absent, respectively, by modifying CLIP's original InfoNCE contrastive loss. Specifically, we design a presence-based contrastive objective that pulls image embeddings closer to their original caption embeddings while pushing them away from the corresponding presence-based negated caption embeddings, and an absence-based contrastive objective that aligns image embeddings with both original and absence-based negated caption embeddings while maintaining a semantic distinction between the two text embeddings. Based on our observation that the front transformer layers of CLIP text encoder have stronger learning ability for negated text than the later layers, we fine-tune the front transformer layers of the CLIP text encoder at each training step using the combined contrastive objective. Experimental results show that, compared with pretrained CLIP, Omni-NegCLIP improves performance on presence-based negation and absence-based negation tasks by up to 52.65% and 12.50%, respectively, without sacrificing general capability in image-text retrieval and even improving it by up to 19.62%. Compared with prior works, Omni-NegCLIP demonstrates a more comprehensive ability to understand multiple types of negation tasks.

Real-time monitoring of inverter-based microgrids is essential for stability, fault response, and operational decision-making. However, electromagnetic transient (EMT) simulations, required to capture fast inverter dynamics, are computationally intensive and unsuitable for real-time applications. This paper presents a data-driven surrogate modeling framework for fast prediction of microgrid behavior using convolutional neural networks (CNN) and Light Gradient Boosting Machine (LightGBM). The models are trained on a high-fidelity EMT digital twin dataset of a microgrid with ten distributed generators under eleven operating and disturbance scenarios, including faults, noise, and communication delays. A sliding-window method is applied to predict important system variables, including voltage magnitude, frequency, total active power, and voltage dip. The results show that model performance changes depending on the type of variable being predicted. The CNN demonstrates high accuracy for time-dependent signals such as voltage, with an $R^2$ value of 0.84, whereas LightGBM shows better performance for structured and disturbance-related variables, achieving an $R^2$ of 0.999 for frequency and 0.75 for voltage dip. A combined CNN+LightGBM model delivers stable performance across all variables. Beyond accuracy, the surrogate models also provide major improvements in computational efficiency. LightGBM achieves more than $1000\times$ speedup and runs faster than real time, while the hybrid model achieves over $500\times$ speedup with near real-time performance. These findings show that data-driven surrogate models can effectively represent microgrid dynamics. They also support real-time and faster-than-real-time predictions. As a result, they are well-suited for applications such as monitoring, fault analysis, and control in inverter-based power systems.

Long video understanding is a key challenge that plagues the advancement of \emph{Multimodal Large language Models} (MLLMs). In this paper, we study this problem from the perspective of visual memory mechanism, and proposed a novel and training-free approach, termed \emph{Flexible Memory} (\textbf{FlexMem}). In principle, FlexMem aims to mimic human behavior of video watching, \emph{i.e.}, continually watching video content and recalling the most relevant memory fragments to answer the question. In this way, FlexMem can help MLLMs achieve video understanding of infinite lengths, unlike previous methods that process all video information at once and have input upper-limit. Concretely, FlexMem first consider the visual KV caches as the memory sources, and realize the effective memory transfer and writing via a dual-pathway compression design. Afterwards, FlexMem also explores different memory reading strategies for the diverse video understanding tasks, including the popular streaming one. To validate FlexMem, we apply it to two popular video-MLLMs, and conduct extensive experiments on five long video and one streaming video task. The experimental results show that on \textbf{a single 3090 GPU}, our FlexMem can achieve obvious improvements than existing efficient video understanding methods and process more than \textbf{1k frames}, which also helps the base MLLMs achieve comparable or even better performance than SOTA MLLMs on some benchmarks, \emph{e.g.} , GPT-4o and Gemini-1.5 Pro.

Genotype imputation enables dense variant coverage for genome-wide association and risk-prediction studies, yet conventional reference-panel methods remain limited by ancestry bias and reduced rare-variant accuracy. We present Genotype Bidirectional Encoder Representations from Transformers (GenoBERT), a transformer-based, reference-free framework that tokenizes phased genotypes and uses a self-attention mechanism to capture both short- and long-range linkage disequilibrium (LD) dependencies. Benchmarking on two independent datasets including the Louisiana Osteoporosis Study (LOS) and the 1000 Genomes Project (1KGP) across ancestry groups and multiple genotype missingness levels (5-50%) shows that GenoBERT achieves the highest overall accuracy compared to four baseline methods (Beagle5.4, SCDA, BiU-Net, and STICI). At practical sparsity levels (up to 25% missing), GenoBERT attains high overall imputation accuracy ($r^2 approx 0.98$) across datasets, and maintains robust performance ($r^2 > 0.90$) even at 50% missingness. Experimental results across different ancestries confirm consistent gains across datasets, with resilience to small sample sizes and weak LD. A 128-SNP (single-nucleotide polymorphism) context window (approximately 100 Kb) is validated through LD-decay analyses as sufficient to capture local correlation structures. By eliminating reference-panel dependence while preserving high accuracy, GenoBERT provides a scalable and robust solution for genotype imputation and a foundation for downstream genomic modeling.

LLM-based shopping agents increasingly rely on long purchase histories and multi-turn interactions for personalization, yet naively appending raw history to prompts is often ineffective due to noise, length, and relevance mismatch. We propose MemRerank, a preference memory framework that distills user purchase history into concise, query-independent signals for personalized product reranking. To study this problem, we build an end-to-end benchmark and evaluation framework centered on an LLM-based \textbf{1-in-5} selection task, which measures both memory quality and downstream reranking utility. We further train the memory extractor with reinforcement learning (RL), using downstream reranking performance as supervision. Experiments with two LLM-based rerankers show that MemRerank consistently outperforms no-memory, raw-history, and off-the-shelf memory baselines, yielding up to \textbf{+10.61} absolute points in 1-in-5 accuracy. These results suggest that explicit preference memory is a practical and effective building block for personalization in agentic e-commerce systems.

We present the Thiomi Dataset, a large-scale multimodal corpus spanning ten African languages across four language families: Swahili, Kikuyu, Kamba, Kimeru, Luo, Maasai, Kipsigis, Somali (East Africa); Wolof (West Africa); and Fulani (West/Central Africa). The dataset contains over 601,000 approved sentence-level text annotations and over 385,000 audio recordings across nine languages, collected through a dedicated community data collection platform involving over 100 contributors. The Thiomi platform collected data for nine languages; Swahili data was supplemented with existing Common Voice recordings. A multi-tier quality assurance pipeline achieves 86-100% text approval rates for the six primary languages. To validate the dataset's utility, we train and evaluate ASR, MT, and TTS models, establishing baselines across all ten languages. Our best ASR system achieves 3.24% WER on Swahili (Common Voice), reducing prior academic SOTA from 8.3% to 3.24% (5.1 percentage point absolute, 61% relative reduction), and 4.3% WER on Somali. The dataset will be published on HuggingFace. We describe the collection platform, quality assurance workflows, and baseline experiments, and discuss implications for African language technology infrastructure.

Enforcing exact macroscopic conservation laws, such as mass and energy, in neural partial differential equation (PDE) solvers is computationally challenging in high dimensions. Traditional discrete projections rely on deterministic quadrature that scales poorly and restricts mesh-free formulations like PINNs. Furthermore, high-order operators incur heavy memory overhead, and generic optimization often lacks convergence guarantees for non-convex conservation manifolds. To address this, we propose the Stochastic Dimension Implicit Functional Projection (SDIFP) framework. Instead of projecting discrete vectors, SDIFP applies a global affine transformation to the continuous network output. This yields closed-form solutions for integral constraints via detached Monte Carlo (MC) quadrature, bypassing spatial grid dependencies. For scalable training, we introduce a doubly-stochastic unbiased gradient estimator (DS-UGE). By decoupling spatial sampling from differential operator subsampling, the DS-UGE reduces memory complexity from $\mathcal{O}(M \times N_{\mathcal{L}})$ to $\mathcal{O}(N \times |\mathcal{I}|)$. SDIFP mitigates sampling variance, preserves solution regularity, and maintains $\mathcal{O}(1)$ inference efficiency, providing a scalable, mesh-free approach for solving conservative high-dimensional PDEs.

The ski rental problem is a canonical model for online decision-making under uncertainty, capturing the fundamental trade-off between repeated rental costs and a one-time purchase. While classical algorithms focus on worst-case competitive ratios and recent "learning-augmented" methods leverage point-estimate predictions, neither approach fully exploits the richness of full distributional predictions while maintaining rigorous robustness guarantees. We address this gap by establishing a systematic framework that integrates distributional advice of unknown quality into both deterministic and randomized algorithms. For the deterministic setting, we formalize the problem under perfect distributional prediction and derive an efficient algorithm to compute the optimal threshold-buy day. We provide a rigorous performance analysis, identifying sufficient conditions on the predicted distribution under which the expected competitive ratio (ECR) matches the classic optimal randomized bound. To handle imperfect predictions, we propose the Clamp Policy, which restricts the buying threshold to a safe range controlled by a tunable parameter. We show that this policy is both robust, maintaining good performance even with large prediction errors, and consistent, approaching the optimal performance as predictions become accurate. For the randomized setting, we characterize the stopping distribution via a Water-Filling Algorithm, which optimizes expected cost while strictly satisfying robustness constraints. Experimental results across diverse distributions (Gaussian, geometric, and bi-modal) demonstrate that our framework improves consistency significantly over existing point-prediction baselines while maintaining comparable robustness.

Large language models (LLMs) are widely applied to data analytics over documents, yet direct reasoning over long, noisy documents remains brittle and error-prone. Hence, we study document question answering (QA) that consolidates dispersed evidence into a structured output (e.g., a table, graph, or chunks) to support reliable, verifiable QA. We propose a two-pillar framework, LiteCoST, to achieve both high accuracy and low latency with small language models (SLMs). Pillar 1: Chain-of-Structured-Thought (CoST). We introduce a CoST template, a schema-aware instruction that guides a strong LLM to produce both a step-wise CoST trace and the corresponding structured output. The process induces a minimal structure, normalizes entities/units, aligns records, serializes the output, and verifies/refines it, yielding auditable supervision. Pillar 2: SLM fine-tuning. The compact models are trained on LLM-generated CoST data in two stages: Supervised Fine-Tuning for structural alignment, followed by Group Relative Policy Optimization (GRPO) incorporating triple rewards for answer/format quality and process consistency. By distilling structure-first behavior into SLMs, this approach achieves LLM-comparable quality on multi-domain long-document QA using 3B/7B SLMs, while delivering 2-4x lower latency than GPT-4o and DeepSeek-R1 (671B). The code is available at https://github.com/HKUSTDial/LiteCoST.

Existing benchmarks measure capability -- whether a model succeeds on a single attempt -- but production deployments require reliability -- consistent success across repeated attempts on tasks of varying duration. We show these properties diverge systematically as task duration grows, and that pass@1 on short tasks is structurally blind to this divergence. We introduce a reliability science framework for long-horizon LLM agents with four metrics: Reliability Decay Curve (RDC), Variance Amplification Factor (VAF), Graceful Degradation Score (GDS), and Meltdown Onset Point (MOP). We evaluate 10 models across 23,392 episodes on a 396-task benchmark spanning four duration buckets and three domains. Key findings: (1) reliability decay is domain-stratified -- SE GDS drops from 0.90 to 0.44 while document processing is nearly flat (0.74 to 0.71); (2) VAF bifurcates by capability tier -- high VAF is a capability signature, not an instability signal; (3) capability and reliability rankings diverge substantially, with multi-rank inversions at long horizons; (4) frontier models have the highest meltdown rates (up to 19%) because they attempt ambitious multi-step strategies that sometimes spiral; and (5) memory scaffolds universally hurt long-horizon performance across all 10 models. These results motivate reliability as a first-class evaluation dimension alongside capability.

Fine-scale-faithful neural simulation under fixed storage budgets remains challenging. Many existing methods reduce high-frequency error by improving architectures, training objectives, or rollout strategies. However, under budgeted coarsen-quantize-decode pipelines, fine detail can already be lost when the carried state is constructed. In the canonical periodic incompressible Navier-Stokes setting, we show that primitive and derived fields undergo systematically different retained-band distortions under the same operator. Motivated by this observation, we formulate Derived-Field Optimization (DerivOpt), a general state-design framework that chooses which physical fields are carried and how storage budget is allocated across them under a calibrated channel model. Across the full time-dependent forward subset of PDEBench, DerivOpt not only improves pooled mean rollout nRMSE, but also delivers a decisive advantage in fine-scale fidelity over a broad set of strong baselines. More importantly, the gains are already visible at input time, before rollout learning begins. This indicates that the carried state is often the dominant bottleneck under tight storage budgets. These results suggest a broader conclusion: in budgeted neural simulation, carried-state design should be treated as a first-class design axis alongside architecture, loss, and rollout strategy.

As AI assistants become commonplace in daily life, the demand for solutions that reduce the cost of inference without sacrificing utility is increasing. Existing work on AI sustainability frequently emphasizes hardware and software optimizations; however, there may be comparable value in social approaches that shape user behavior and discourage unnecessary use. In this study, we operationalize sustainability in terms of energy-efficiency and analyze a publicly sourced sample of prompts where AI is used for assistance in software development. Using this categorization, we find that nearly half of the observed queries can be considered unnecessary relative to their expected benefit. We further observe that factoid-style information retrieval constitutes the largest share of unnecessary requests, suggesting that a meaningful portion of everyday AI usage may be replaceable with lower-cost alternatives (e.g., conventional search or local documentation). These findings motivate a closer examination of how, why, and when AI systems are invoked, and what norms or interface-level nudges might reduce avoidable demand. We conclude with a call to replicate and extend this preliminary analysis and to pay greater attention to the social dimension of AI sustainability.

SiPaKosa is a comprehensive corpus of Sinhala and Pali doctrinal texts comprising approximately 786K sentences and 9.25M words, incorporating 16 copyright-cleared historical Buddhist documents alongside the complete web-scraped Tripitaka canonical texts. The corpus was created through high-quality OCR using Google Document AI on historical manuscripts, combined with systematic web scraping of canonical repositories, followed by rigorous quality control and metadata annotation. The corpus is organised into language-specific subcorpora: Sinhala and Mixed Sinhala-Pali. We evaluate the performance of language models using ten pretrained models, with perplexity scores ranging from 1.09 to 189.67 on our corpus. This analysis shows that proprietary models significantly outperform open-source alternatives by factors of three to six times. This corpus supports the pretraining of domain-adapted language models, facilitates historical language analysis, and aids in the development of information retrieval systems for Buddhist scholarship while preserving Sinhala cultural heritage.

Sign language is the primary approach of communication for the Deaf and Hard-of-Hearing (DHH) community. While there are numerous benchmarks for high-resource sign languages, low-resource languages like Arabic remain underrepresented. Currently, there is no publicly available dataset for Syrian Arabic Sign Language (SyArSL). To overcome this gap, we introduce SyriSign, a dataset comprising 1500 video samples across 150 unique lexical signs, designed for text-to-SyArSL translation tasks. This work aims to reduce communication barriers in Syria, as most news are delivered in spoken or written Arabic, which is often inaccessible to the deaf community. We evaluated SyriSign using three deep learning architectures: MotionCLIP for semantic motion generation, T2M-GPT for text-conditioned motion synthesis, and SignCLIP for bilingual embedding alignment. Experimental results indicate that while generative approaches show strong potential for sign representation, the limited dataset size constrains generalization performance. We will release SyriSign publicly, hoping it serves as an initial benchmark.

Phoneme-based ASR factorizes recognition into speech-to-phoneme (S2P) and phoneme-to-grapheme (P2G), enabling cross-lingual acoustic sharing while keeping language-specific orthography in a separate module. While large language models (LLMs) are promising for P2G, multilingual P2G remains challenging due to language-aware generation and severe cross-language data imbalance. We study multilingual LLM-based P2G on the ten-language CV-Lang10 benchmark. We examine robustness strategies that account for S2P uncertainty, including DANP and Simplified SKM (S-SKM). S-SKM is a Monte Carlo approximation that avoids CTC-based S2P probability weighting in P2G training. Robust training and low-resource oversampling reduce the average WER from 10.56% to 7.66%.

Large Language Models (LLMs) have emerged as a popular choice in vulnerability detection studies given their foundational capabilities, open source availability, and variety of models, but have limited scalability due to extensive compute requirements. Using the natural graph relational structure of code, we show that our proposed graph neural network (GNN) based deep learning model VulGNN for vulnerability detection can achieve performance almost on par with LLMs, but is 100 times smaller in size and fast to retrain and customize. We describe the VulGNN architecture, ablation studies on components, learning rates, and generalizability to different code datasets. As a lightweight model for vulnerability analysis, VulGNN is efficient and deployable at the edge as part of real-world software development pipelines.

In recent years, multimodal large models have continued to improve on general benchmarks. However, in real-world content moderation and adversarial settings, mainstream models still suffer from degraded generalization and catastrophic forgetting because of limited fine-grained visual perception and insufficient modeling of long-tail noise. In this paper, we present Xuanwu VL-2B as a case study of how general multimodal models can be developed into an industrial-grade foundation model for content ecosystems. The model adopts a compact InternViT-300M + MLP + Qwen3 1.7B architecture, balancing fine-grained visual perception, language-semantic alignment, and deployment cost within an approximately 2B-parameter budget. To balance business specialization with the retention of general capabilities, we developed a data iteration and curation mechanism and trained the model through a progressive three-stage pipeline: pre-training, mid-training, and post-training. Ablation studies and offline business evaluations show that Xuanwu VL-2B achieves an average score of 67.90 across seven OpenCompass multimodal metrics (vs. 64.27 for InternVL 3.5 2B), an average recall of 94.38% over seven independent business moderation tasks, and a weighted overall recall of 82.82% on policy-violating text in challenging adversarial OCR scenarios, outperforming Gemini-2.5-Pro (76.72%). These results show that, under a limited parameter budget, Xuanwu VL-2B achieves a practical balance among business alignment, visual perception, general capability retention, and deployment cost.

Routing is widely used to scale large language models, from Mixture-of-Experts gating to multi-model/tool selection. A common belief is that routing to a task ``expert'' activates sparser internal computation and thus yields more certain and stable outputs (the Sparsity--Certainty Hypothesis). We test this belief by injecting routing-style meta prompts as a textual proxy for routing signals in front of frozen instruction-tuned LLMs. We quantify (C1) internal density via activation sparsity, (C2) domain-keyword attention, and (C3) output stability via predictive entropy and semantic variation. On a RouterEval subset with three instruction-tuned models (Qwen3-8B, Llama-3.1-8B-Instruct, and Mistral-7B-Instruct-v0.2), meta prompts consistently densify early/middle-layer representations rather than increasing sparsity; natural-language expert instructions are often stronger than structured tags. Attention responses are heterogeneous: Qwen/Llama reduce keyword attention, while Mistral reinforces it. Finally, the densification--stability link is weak and appears only in Qwen, with near-zero correlations in Llama and Mistral. We present RIDE as a diagnostic probe for calibrating routing design and uncertainty estimation.

Deep learning-based tropical cyclone (TC) forecasting methods have demonstrated significant potential and application advantages, as they feature much lower computational cost and faster operation speed than numerical weather prediction models. However, existing deep learning methods still have key limitations: they can only process a single type of sequential trajectory data or homogeneous meteorological variables, and fail to achieve accurate forecasting of abnormal deflected TCs. To address these challenges, we present two groundbreaking contributions. First, we have constructed a multimodal and multi-source dataset named AOT-TCs for TC forecasting in the Northwest Pacific basin. As the first dataset of its kind, it innovatively integrates heterogeneous variables from the atmosphere, ocean, and land, thus obtaining a comprehensive and information-rich meteorological dataset. Second, based on the AOT-TCs dataset, we propose a forecasting model that can handle both normal and abnormally deflected TCs. This is the first TC forecasting model to adopt an explicit atmosphere-ocean-terrain coupling architecture, enabling it to effectively capture complex interactions across physical domains. Extensive experiments on all TC cases in the Northwest Pacific from 2017 to 2024 show that our model achieves state-of-the-art performance in TC forecasting: it not only significantly improves the forecasting accuracy of normal TCs but also breaks through the technical bottleneck in forecasting abnormally deflected TCs.

The AEC-Bench is a multimodal benchmark for evaluating agentic systems on real-world tasks in the Architecture, Engineering, and Construction (AEC) domain. The benchmark covers tasks requiring drawing understanding, cross-sheet reasoning, and construction project-level coordination. This report describes the benchmark motivation, dataset taxonomy, evaluation protocol, and baseline results across several domain-specific foundation model harnesses. We use AEC-Bench to identify consistent tools and harness design techniques that uniformly improve performance across foundation models in their own base harnesses, such as Claude Code and Codex. We openly release our benchmark dataset, agent harness, and evaluation code for full replicability at https://github.com/nomic-ai/aec-bench under an Apache 2 license.

Long-horizon dialogue systems suffer from semanticdrift and unstable memory retention across extended sessions. This paper presents a Multi-Layer Memory Framework that decomposes dialogue history into working, episodic, and semantic layers with adaptive retrieval gating and retention regularization. The architecture controls cross-session drift while maintaining bounded context growth and computational efficiency. Experiments on LOCOMO, LOCCO, and LoCoMo show improved performance, achieving 46.85 Success Rate, 0.618 overall F1 with 0.594 multi-hop F1, and 56.90% six-period retention while reducing false memory rate to 5.1% and context usage to 58.40%. Results confirm enhanced long-term retention and reasoning stability under constrained context budgets.

Large Language Models (LLMs) often experience performance degradation during long-running interactions due to increasing context length, memory saturation, and computational overhead. This paper presents an adaptive context compression framework that integrates importance-aware memory selection, coherence-sensitive filtering, and dynamic budget allocation to retain essential conversational information while controlling context growth. The approach is evaluated on LOCOMO, LOCCO, and LongBench benchmarks to assess answer quality, retrieval accuracy, coherence preservation, and efficiency. Experimental results demonstrate that the proposed method achieves consistent improvements in conversational stability and retrieval performance while reducing token usage and inference latency compared with existing memory and compression-based approaches. These findings indicate that adaptive context compression provides an effective balance between long-term memory preservation and computational efficiency in persistent LLM interactions

The aim of our paper is to render an object in 3-dimension using a set of its orthographic views. Corner detector (Harris Detector) is applied on the input views to obtain control points. These control points are projected perpendicular to respective views, in order to construct an envelope. A set of points describing the object in 3-dimension, are obtained from the intersection of these mutually perpendicular envelopes. These set of points are used to regenerate the surfaces of the object using computational geometry. At the end, the object in 3-dimension is rendered using OpenGL

This paper proposes the synthetic long-video meta-evaluation (SLVMEval), a benchmark for meta-evaluating text-to-video (T2V) evaluation systems. The proposed SLVMEval benchmark focuses on assessing these systems on videos of up to 10,486 s (approximately 3 h). The benchmark targets a fundamental requirement, namely, whether the systems can accurately assess video quality in settings that are easy for humans to assess. We adopt a pairwise comparison-based meta-evaluation framework. Building on dense video-captioning datasets, we synthetically degrade source videos to create controlled "high-quality versus low-quality" pairs across 10 distinct aspects. Then, we employ crowdsourcing to filter and retain only those pairs in which the degradation is clearly perceptible, thereby establishing an effective final testbed. Using this testbed, we assess the reliability of existing evaluation systems in ranking these pairs. Experimental results demonstrate that human evaluators can identify the better long video with 84.7%-96.8% accuracy, and in nine of the 10 aspects, the accuracy of these systems falls short of human assessment, revealing weaknesses in text-to-long-video evaluation.

Nonconvex multi-well energies in cell-induced phase transitions give rise to sharp interfaces, fine-scale microstructures, and distance-dependent inter-cell coupling, all of which pose significant challenges for physics-informed learning. Existing methods often suffer from over-smoothing in near-field patterns. To address this, we propose biomimetic physics-informed neural networks (Bio-PINNs), a variational framework that encodes temporal causality into explicit spatial causality via a progressive distance gate. Furthermore, Bio-PINNs leverage a deformation-uncertainty proxy for the interfacial length scale to target microstructure-prone regions, providing a computationally efficient alternative to explicit second-derivative regularization. We provide theoretical guarantees for the resulting uncertainty-driven ``retain-resample-release" adaptive collocation strategy, which ensures persistent coverage under gating and establishing a quantitative near-to-far growth bound. Across single- and multi-cell benchmarks, diverse separations, and various regularization regimes, Bio-PINNs consistently recover sharp transition layers and tether morphologies, significantly outperforming state-of-the-art adaptive and ungated baselines.

Open-set anomaly detection (OSAD) is an emerging paradigm designed to utilize limited labeled data from anomaly classes seen in training to identify both seen and unseen anomalies during testing. Current approaches rely on simple augmentation methods to generate pseudo anomalies that replicate unseen anomalies. Despite being promising in image data, these methods are found to be ineffective in time series data due to the failure to preserve its sequential nature, resulting in trivial or unrealistic anomaly patterns. They are further plagued when the training data is contaminated with unlabeled anomalies. This work introduces $\textbf{IMPACT}$, a novel framework that leverages $\underline{\textbf{i}}$nfluence $\underline{\textbf{m}}$odeling for o$\underline{\textbf{p}}$en-set time series $\underline{\textbf{a}}$nomaly dete$\underline{\textbf{ct}}$ion, to tackle these challenges. The key insight is to $\textbf{i)}$ learn an influence function that can accurately estimate the impact of individual training samples on the modeling, and then $\textbf{ii)}$ leverage these influence scores to generate semantically divergent yet realistic unseen anomalies for time series while repurposing high-influential samples as supervised anomalies for anomaly decontamination. Extensive experiments show that IMPACT significantly outperforms existing state-of-the-art methods, showing superior accuracy under varying OSAD settings and contamination rates.

Adversarial robustness evaluation faces a critical challenge as new defense paradigms emerge that can exploit limitations in existing assessment methods. This paper reveals that Dummy Classes-based defenses, which introduce an additional "dummy" class as a safety sink for adversarial examples, achieve significantly overestimated robustness under conventional evaluation strategies like AutoAttack. The fundamental limitation stems from these attacks' singular focus on misleading the true class label, which aligns perfectly with the defense mechanism--successful attacks are simply captured by the dummy class. To address this gap, we propose Dummy-Aware Weighted Attack (DAWA), a novel evaluation method that simultaneously targets both the true label and dummy label with adaptive weighting during adversarial example synthesis. Extensive experiments demonstrate that DAWA effectively breaks this defense paradigm, reducing the measured robustness of a leading Dummy Classes-based defense from 58.61% to 29.52% on CIFAR-10 under l_infty perturbation (epsilon=8/255). Our work provides a more reliable benchmark for evaluating this emerging class of defenses and highlights the need for continuous evolution of robustness assessment methodologies.

Parkinson's disease (PD) affects over ten million people worldwide. Although temporal interference (TI) and deep brain stimulation (DBS) are promising therapies, inter-individual variability limits empirical treatment selection, increasing non-negligible surgical risk and cost. Previous explorations either resort to limited statistical biomarkers that are insufficient to characterize variability, or employ AI-driven methods which is prone to overfitting and opacity. We bridge this gap with a pretraining-finetuning framework to predict outcomes directly from resting-state fMRI. Critically, a generative virtual brain foundation model, pretrained on a collective dataset (2707 subjects, 5621 sessions) to capture universal disorder patterns, was finetuned on PD cohorts receiving TI (n=51) or DBS (n=55) to yield individualized virtual brains with high fidelity to empirical functional connectivity (r=0.935). By constructing counterfactual estimations between pathological and healthy neural states within these personalized models, we predicted clinical responses (TI: AUPR=0.853; DBS: AUPR=0.915), substantially outperforming baselines. External and prospective validations (n=14, n=11) highlight the feasibility of clinical translation. Moreover, our framework provides state-dependent regional patterns linked to response, offering hypothesis-generating mechanistic insights.

Accurate segmentation of brain tissues such as gray matter and white matter from magnetic resonance imaging is essential for studying brain anatomy, diagnosing neurological disorders, and monitoring disease progression. Traditional methods, such as FSL FAST, produce tissue probability maps but often require task-specific adjustments and face challenges with diverse imaging conditions. Recent foundation models, such as MedSAM, offer a prompt-based approach that leverages large-scale pretraining. In this paper, we propose a modified MedSAM model designed for multi-class brain tissue segmentation. Our preprocessing pipeline includes skull stripping with FSL BET, tissue probability mapping with FSL FAST, and converting these into 2D axial, sagittal, coronal slices with multi-class labels (background, gray matter, and white matter). We extend MedSAM's mask decoder to three classes, freezing the pre-trained image encoder and fine-tuning the prompt encoder and decoder. Experiments on the IXI dataset achieve Dice scores up to 0.8751. This work demonstrates that foundation models like MedSAM can be adapted for multi-class medical image segmentation with minimal architectural modifications. Our findings suggest that such models can be extended to more diverse medical imaging scenarios in future work.

Soccer commentary plays a crucial role in enhancing the soccer game viewing experience for audiences. Previous studies in automatic soccer commentary generation typically adopt an end-to-end method to generate anonymous live text commentary. Such generated commentary is insufficient in the context of real-world live televised commentary, as it contains anonymous entities, context-dependent errors and lacks statistical insights of the game events. To bridge the gap, we propose GameSight, a two-stage model to address soccer commentary generation as a knowledge-enhanced visual reasoning task, enabling live-televised-like knowledgeable commentary with accurate reference to entities (players and teams). GameSight starts by performing visual reasoning to align anonymous entities with fine-grained visual and contextual analysis. Subsequently, the entity-aligned commentary is refined with knowledge by incorporating external historical statistics and iteratively updated internal game state information. Consequently, GameSight improves the player alignment accuracy by 18.5% on SN-Caption-test-align dataset compared to Gemini 2.5-pro. Combined with further knowledge enhancement, GameSight outperforms in segment-level accuracy and commentary quality, as well as game-level contextual relevance and structural composition. We believe that our work paves the way for a more informative and engaging human-centric experience with the AI sports application. Demo Page: https://gamesight2025.github.io/gamesight2025

Existing vision-and-language navigation (VLN) models primarily reason over past and current visual observations, while largely ignoring the future visual dynamics induced by actions. As a result, they often lack an effective understanding of the causal relationship between actions and how the visual world changes, limiting robust decision-making. Humans, in contrast, can imagine the near future by leveraging action-dynamics causality, which improves both environmental understanding and navigation choices. Inspired by this capability, we propose LatentPilot, a new paradigm that exploits future observations during training as a valuable data source to learn action-conditioned visual dynamics, while requiring no access to future frames at inference. Concretely, we propose a flywheel-style training mechanism that iteratively collects on-policy trajectories and retrains the model to better match the agent's behavior distribution, with an expert takeover triggered when the agent deviates excessively. LatentPilot further learns visual latent tokens without explicit supervision; these latent tokens attend globally in a continuous latent space and are carried across steps, serving as both the current output and the next input, thereby enabling the agent to dream ahead and reason about how actions will affect subsequent observations. Experiments on R2R-CE, RxR-CE, and R2R-PE benchmarks achieve new SOTA results, and real-robot tests across diverse environments demonstrate LatentPilot's superior understanding of environment-action dynamics in scene. Project page:https://abdd.top/latentpilot/

Modern web scraping struggles with dynamic, interactive websites that require more than static HTML parsing. Current methods are often brittle and require manual customization for each site. To address this, we introduce Webscraper, a framework designed to handle the challenges of modern, dynamic web applications. It leverages a Multimodal Large Language Model (MLLM) to autonomously navigate interactive interfaces, invoke specialized tools, and perform structured data extraction in environments where traditional scrapers are ineffective. Webscraper utilizes a structured five-stage prompting procedure and a set of custom-built tools to navigate and extract data from websites following the common ``index-and-content'' architecture. Our experiments, conducted on six news websites, demonstrate that the full Webscraper framework, equipped with both our guiding prompt and specialized tools, achieves a significant improvement in extraction accuracy over the baseline agent Anthropic's Computer Use. We also applied the framework to e-commerce platforms to validate its generalizability.

Providing timely and accurate learning support in large-scale online coding courses is challenging, particularly in resource-constrained contexts. We present Kwame 2.0, a bilingual (English-French) generative AI teaching assistant built using retrieval-augmented generation and deployed in a human-in-the-loop forum within SuaCode, an introductory mobile-based coding course for learners across Africa. Kwame 2.0 retrieves relevant course materials and generates context-aware responses while encouraging human oversight and community participation. We deployed the system in a 15-month longitudinal study spanning 15 cohorts with 3,717 enrollments across 35 African countries. Evaluation using community feedback and expert ratings shows that Kwame 2.0 provided high-quality and timely support, achieving high accuracy on curriculum-related questions, while human facilitators and peers effectively mitigated errors, particularly for administrative queries. Our findings demonstrate that human-in-the-loop generative AI systems can combine the scalability and speed of AI with the reliability of human support, offering an effective approach to learning assistance for underrepresented populations in resource-constrained settings at scale.

Metal-organic frameworks (MOFs) offer a vast design space, and as such, computational simulations play a critical role in predicting their structural and physicochemical properties. However, MOF simulations remain difficult to access because reliable analysis require expert decisions for workflow construction, parameter selection, tool interoperability, and the preparation of computational ready structures. Here, we introduce SimMOF, a large language model based multi agent framework that automates end-to-end MOF simulation workflows from natural language queries. SimMOF translates user requests into dependency aware plans, generates runnable inputs, orchestrates multiple agents to execute simulations, and summarizes results with analysis aligned to the user query. Through representative case studies, we show that SimMOF enables adaptive and cognitively autonomous workflows that reflect the iterative and decision driven behavior of human researchers and as such provides a scalable foundation for data driven MOF research.

Existing robotic foundation policies are trained primarily via large-scale imitation learning. While such models demonstrate strong capabilities, they often struggle with long-horizon tasks due to distribution shift and error accumulation. While reinforcement learning (RL) can finetune these models, it cannot work well across diverse tasks without manual reward engineering. We propose VLLR, a dense reward framework combining (1) an extrinsic reward from Large Language Models (LLMs) and Vision-Language Models (VLMs) for task progress recognition, and (2) an intrinsic reward based on policy self-certainty. VLLR uses LLMs to decompose tasks into verifiable subtasks and then VLMs to estimate progress to initialize the value function for a brief warm-up phase, avoiding prohibitive inference cost during full training; and self-certainty provides per-step intrinsic guidance throughout PPO finetuning. Ablation studies reveal complementary benefits: VLM-based value initialization primarily improves task completion efficiency, while self-certainty primarily enhances success rates, particularly on out-of-distribution tasks. On the CHORES benchmark covering mobile manipulation and navigation, VLLR achieves up to 56% absolute success rate gains over the pretrained policy, up to 5% gains over state-of-the-art RL finetuning methods on in-distribution tasks, and up to $10\%$ gains on out-of-distribution tasks, all without manual reward engineering. Additional visualizations can be found in https://silongyong.github.io/vllr_project_page/

Access to the most up-to-date information on medical countermeasures is important for the research and development of effective treatments for viruses and marine toxins. However, there is a lack of comprehensive databases that curate data on viruses and marine toxins, making decisions on medical countermeasures slow and difficult. In this work, we employ two large language models (LLMs) of ChatGPT and Grok to design two comprehensive databases of therapeutic countermeasures for five viruses of Lassa, Marburg, Ebola, Nipah, and Venezuelan equine encephalitis, as well as marine toxins. With high-level human-provided inputs, the two LLMs identify public databases containing data on the five viruses and marine toxins, collect relevant information from these databases and the literature, iteratively cross-validate the collected information, and design interactive webpages for easy access to the curated, comprehensive databases. Notably, the ChatGPT LLM is employed to design agentic AI workflows (consisting of two AI agents for research and decision-making) to rank countermeasures for viruses and marine toxins in the databases. Together, our work explores the potential of LLMs as a scalable, updatable approach for building comprehensive knowledge databases and supporting evidence-based decision-making.

Graph Convolutional Network (GCN) is a model that can effectively handle graph data tasks and has been successfully applied. However, for large-scale graph datasets, GCN still faces the challenge of high computational overhead, especially when the number of convolutional layers in the graph is large. Currently, there are many advanced methods that use various sampling techniques or graph coarsening techniques to alleviate the inconvenience caused during training. However, among these methods, some ignore the multi-granularity information in the graph structure, and the time complexity of some coarsening methods is still relatively high. In response to these issues, based on our previous work, in this paper, we propose a new framework called Efficient and Scalable Granular-ball Graph Coarsening Method for Large-scale Graph Node Classification. Specifically, this method first uses a multi-granularity granular-ball graph coarsening algorithm to coarsen the original graph to obtain many subgraphs. The time complexity of this stage is linear and much lower than that of the exiting graph coarsening methods. Then, subgraphs composed of these granular-balls are randomly sampled to form minibatches for training GCN. Our algorithm can adaptively and significantly reduce the scale of the original graph, thereby enhancing the training efficiency and scalability of GCN. Ultimately, the experimental results of node classification on multiple datasets demonstrate that the method proposed in this paper exhibits superior performance. The code is available at https://anonymous.4open.science/r/1-141D/.

Formative feedback is central to effective learning, yet providing timely, individualised feedback at scale remains a persistent challenge. While recent work has explored the use of large language models (LLMs) to automate feedback, most existing systems still conceptualise feedback as a static, one-way artifact, offering limited support for interpretation, clarification, or follow-up. In this work, we introduce REFINE, a locally deployable, multi-agent feedback system built on small, open-source LLMs that treats feedback as an interactive process. REFINE combines a pedagogically-grounded feedback generation agent with an LLM-as-a-judge-guided regeneration loop using a human-aligned judge, and a self-reflective tool-calling interactive agent that supports student follow-up questions with context-aware, actionable responses. We evaluate REFINE through controlled experiments and an authentic classroom deployment in an undergraduate computer science course. Automatic evaluations show that judge-guided regeneration significantly improves feedback quality, and that the interactive agent produces efficient, high-quality responses comparable to a state-of-the-art closed-source model. Analysis of real student interactions further reveals distinct engagement patterns and indicates that system-generated feedback systematically steers subsequent student inquiry. Our findings demonstrate the feasibility and effectiveness of multi-agent, tool-augmented feedback systems for scalable, interactive feedback.

Finite state machines (FSM) are executable formal specifications of reactive systems. These machines are designed based on systems' requirements. The requirements are often recorded in textual documents written in natural languages. FSMs play a crucial role in different phases of the model-driven system engineering (MDE). For example, they serve to automate testing activities. FSM quality is critical: the lower the quality of FSM, the higher the number of faults surviving the testing phase and the higher the risk of failure of the systems in production, which could lead to catastrophic scenarios. Therefore, this paper leverages recent advances in the domain of LLM to propose an LLM-based framework for designing FSMs from requirements. The framework also suggests an expert-centric approach based on FSM mutation and test generation for repairing the FSMs produced by LLMs. This paper also provides an experimental analysis and evaluation of LLM's capacities in performing the tasks presented in the framework and FSM repair via various methods. The paper presents experimental results with simulated data. These results and methods bring a new analysis and vision of LLMs that are useful for further development of machine learning technology and its applications to MDE.

Recent advances in large language models (LLMs) have enabled agentic systems that translate natural language intent into executable scientific visualization (SciVis) tasks. Despite rapid progress, the community lacks a principled and reproducible benchmark for evaluating these emerging SciVis agents in realistic, multi-step analysis settings. We present SciVisAgentBench, a comprehensive and extensible benchmark for evaluating scientific data analysis and visualization agents. Our benchmark is grounded in a structured taxonomy spanning four dimensions: application domain, data type, complexity level, and visualization operation. It currently comprises 108 expert-crafted cases covering diverse SciVis scenarios. To enable reliable assessment, we introduce a multimodal outcome-centric evaluation pipeline that combines LLM-based judging with deterministic evaluators, including image-based metrics, code checkers, rule-based verifiers, and case-specific evaluators. We also conduct a validity study with 12 SciVis experts to examine the agreement between human and LLM judges. Using this framework, we evaluate representative SciVis agents and general-purpose coding agents to establish initial baselines and reveal capability gaps. SciVisAgentBench is designed as a living benchmark to support systematic comparison, diagnose failure modes, and drive progress in agentic SciVis. The benchmark is available at https://scivisagentbench.github.io/.

Autonomous experimental systems are increasingly used in materials research to accelerate scientific discovery, but their performance is often limited by low-quality, noisy data. This issue is especially problematic in data-intensive structure-property learning tasks such as Image-to-Spectrum (Im2Spec) and Spectrum-to-Image (Spec2Im) translations, where standard active learning strategies can mistakenly prioritize poor-quality measurements. We introduce a gated active learning framework that combines curiosity-driven sampling with a physics-informed quality control filter based on the Simple Harmonic Oscillator model fits, allowing the system to automatically exclude low-fidelity data during acquisition. Evaluations on a pre-acquired dataset of band-excitation piezoresponse spectroscopy (BEPS) data from PbTiO3 thin films with spatially localized noise show that the proposed method outperforms random sampling, standard active learning, and multitask learning strategies. The gated approach enhances both Im2Spec and Spec2Im by handling noise during training and acquisition, leading to more reliable forward and inverse predictions. In contrast, standard active learners often misinterpret noise as uncertainty and end up acquiring bad samples that hurt performance. Given its promising applicability, we further deployed the framework in real-time experiments on BiFeO3 thin films, demonstrating its effectiveness in real autonomous microscopy experiments. Overall, this work supports a shift toward hybrid autonomy in self-driving labs, where physics-informed quality assessment and active decision-making work hand-in-hand for more reliable discovery.

Nonlinear computation is essential for various information processing tasks. Optical implementations are attractive because passive light propagation can manipulate high-dimensional signals with extreme throughput and parallelism; yet realizing nonlinear mappings in optical hardware remains challenging due to the weak nonlinearity of optical materials and the large intensities required to induce nonlinear interactions. This challenge is further amplified in many systems that operate with incoherent illumination, motivating a coherence-aware framework for scalable optical nonlinear processing. Here, we show that linear optical systems, in particular, optimized diffractive processors comprising passive surfaces, can perform large-scale nonlinear function approximation under spatially incoherent or partially coherent illumination, when preceded by intensity-only input encoding. We quantify how the accuracy of the nonlinear function approximation varies with the degree of parallelism, the number of diffractive layers, and the number of trainable diffractive features. Numerical results demonstrate snapshot computation of up to one million distinct nonlinear functions in a single forward pass through a diffractive processor, with the function outputs spatially multiplexed and read out using densely packed detectors at the output. We further provide a proof-of-concept experimental demonstration under incoherent illumination from a liquid crystal display (LCD), enabled by a model-free in situ learning strategy that jointly optimizes the diffractive profile and detector readout geometry in the presence of hardware imperfections and misalignments. Our findings establish diffractive processors as a massively parallel universal function approximator for both spatially incoherent and partially coherent illumination.

Cyclic block coordinate methods are a fundamental class of first-order algorithms, widely used in practice for their simplicity and strong empirical performance. Yet, their theoretical behavior remains challenging to explain, and setting their step sizes -- beyond classical coordinate descent for minimization -- typically requires careful tuning or line-search machinery. In this work, we develop $\texttt{ADUCA}$ (Adaptive Delayed-Update Cyclic Algorithm), a cyclic algorithm addressing a broad class of Minty variational inequalities with monotone Lipschitz operators. $\texttt{ADUCA}$ is parameter-free: it requires no global or block-wise Lipschitz constants and uses no per-epoch line search, except at initialization. A key feature of the algorithm is using operator information delayed by a full cycle, which makes the algorithm compatible with parallel and distributed implementations, and attractive due to weakened synchronization requirements across blocks. We prove that $\texttt{ADUCA}$ attains (near) optimal global oracle complexity as a function of target error $ε>0,$ scaling with $1/ε$ for monotone operators, or with $\log^2(1/ε)$ for operators that are strongly monotone.

The next-token prediction (NTP) objective trains language models to predict a single continuation token at each step. In natural language, however, a prefix can be continued in many valid ways, and even similar meanings may differ in surface form. For example, the sentence ``this website is safe to \underline{browse}'' could plausibly continue with words such as browse, search, visit, surf, or navigate. While standard NTP training treats these alternatives as mutually exclusive targets, we explore a framework that instead predicts concepts, approximated as sets of semantically related tokens. We show that models trained with concept supervision exhibit stronger alignment with human semantic similarity judgments on multiple lexical benchmarks. These gains are accompanied by lower perplexity on semantically meaningful words (definition in Section 3.1), and a modest increase in global token-level perplexity, reflecting a tradeoff between standard NTP optimization and concept-level supervision. Our results suggest that concept-level objectives can improve semantic alignment while maintaining competitive language modeling performance.

Logging statements are essential for software debugging and maintenance. However, existing approaches to automatic logging generation rely on static analysis and produce statements in a single pass without considering runtime behavior. They are also typically evaluated by similarity to developer-written logs, assuming these logs form an adequate gold standard. This assumption is increasingly limiting in the LLM era, where logs are consumed not only by developers but also by LLMs for downstream tasks. As a result, optimizing logs for human similarity does not necessarily reflect their practical utility. To address these limitations, we introduce ReLog, an iterative logging generation framework guided by runtime feedback. ReLog leverages LLMs to generate, execute, evaluate, and refine logging statements so that runtime logs better support downstream tasks. Instead of comparing against developer-written logs, we evaluate ReLog through downstream debugging tasks, including defect localization and repair. We construct a benchmark based on Defects4J under both direct and indirect debugging settings. Results show that ReLog consistently outperforms all baselines, achieving an F1 score of 0.520 and repairing 97 defects in the direct setting, and the best F1 score of 0.408 in the indirect setting where source code is unavailable. Additional experiments across multiple LLMs demonstrate the generality of the framework, while ablations confirm the importance of iterative refinement and compilation repair. Overall, our work reframes logging as a runtime-guided, task-oriented process and advocates evaluating logs by their downstream utility rather than textual similarity.

This paper develops a unified framework for evaluating the optimal degree of task automation. Moving beyond binary automate-or-not assessments, we model automation intensity as a continuous choice in which firms minimize costs by selecting an AI accuracy level, from no automation through partial human-AI collaboration to full automation. On the supply side, we estimate an AI production function via scaling-law experiments linking performance to data, compute, and model size. Because AI systems exhibit predictable but diminishing returns to these inputs, the cost of higher accuracy is convex: good performance may be inexpensive, but near-perfect accuracy is disproportionately costly. Full automation is therefore often not cost-minimizing; partial automation, where firms retain human workers for residual tasks, frequently emerges as the equilibrium. On the demand side, we introduce an entropy-based measure of task complexity that maps model accuracy into a labor substitution ratio, quantifying human labor displacement at each accuracy level. We calibrate the framework with O*NET task data, a survey of 3,778 domain experts, and GPT-4o-derived task decompositions, implementing it in computer vision. Task complexity shapes substitution: low-complexity tasks see high substitution, while high-complexity tasks favor limited partial automation. Scale of deployment is a key determinant: AI-as-a-Service and AI agents spread fixed costs across users, sharply expanding economically viable tasks. At the firm level, cost-effective automation captures approximately 11% of computer-vision-exposed labor compensation; under economy-wide deployment, this share rises sharply. Since other AI systems exhibit similar scaling-law economics, our mechanisms extend beyond computer vision, reinforcing that partial automation is often the economically rational long-run outcome, not merely a transitional phase.

Modern control systems frequently operate under input delays and sampled state measurements. A common delay-compensation strategy is predictor feedback; however, practical implementations require solving an implicit ODE online, resulting in intractable computational cost. Moreover, predictor formulations typically assume continuously available state measurements, whereas in practice measurements may be sampled, irregular, or temporarily missing due to hardware faults. In this work, we develop two neural-operator predictor-feedback designs for nonlinear systems with delayed inputs and sampled measurements. In the first design, we introduce a sampling-horizon prediction operator that maps the current measurement and input history to the predicted state trajectory over the next sampling interval. In the second design, the neural operator approximates only the delay-compensating predictor, which is then composed with the closed-loop flow between measurements. The first approach requires uniform sampling but yields residual bounds that scale directly with the operator approximation error. In contrast, the second accommodates non-uniform, but bounded sampling schedules at the cost of amplified approximation error, revealing a practical tradeoff between sampling flexibility and approximation sensitivity for the control engineer. For both schemes, we establish semi-global practical stability with explicit neural operator error-dependent bounds. Numerical experiments on a 6-link nonlinear robotic manipulator demonstrate accurate tracking and substantial computational speedup of 25$\times$ over a baseline approach.

Learning to use feature-rich software is a persistent challenge, but generative AI tools promise to lower this barrier by replacing complex navigation with natural language prompts. We investigated how people approach prompt-based tools for 3D modeling in an observational study with 26 participants (14 casuals, 12 professionals). Consistent with earlier work, participants skipped tutorials and manuals, relying on trial and error. What differed in the generative AI context was how and why they sought support: the prompt box became the entry point for learning, collapsing onboarding into immediate action, while some casual users turned to external LLMs for prompts. Professionals used 3D expertise to refine iterations and critically evaluated outputs, often discarding models that did not meet their standards, whereas casual users settled for "good enough." We contribute empirical insights into how generative AI reshapes help-seeking, highlighting new practices of onboarding, recursive AI-for-AI support, and shifting expertise in interpreting outputs.

Due to simplicity and strong stability guarantees, predictor feedback methods have stood as a popular approach for time delay systems since the 1950s. For time-varying delays, however, implementation requires computing a prediction horizon defined by the inverse of the delay function, which is rarely available in closed form and must be approximated. In this work, we formulate the inverse delay mapping as an operator learning problem and study predictor feedback under approximation of the prediction horizon. We propose two approaches: (i) a numerical method based on time integration of an equivalent ODE, and (ii) a data-driven method using neural operators to learn the inverse mapping. We show that both approaches achieve arbitrary approximation accuracy over compact sets, with complementary trade-offs in computational cost and scalability. Building on these approximations, we then develop an output-feedback predictor design for systems with delays in both the input and the measurement. We prove that the resulting closed-loop system is globally exponentially stable when the prediction horizon is approximated with sufficiently small error. Lastly, numerical experiments validate the proposed methods and illustrate their trade-offs between accuracy and computational efficiency.

Requirements engineering for aged-care digital health must account for human aspects, because requirement priorities are shaped not only by technical functionality but also by stakeholders' health conditions, socioeconomics, and lived experience. Knowing which human aspects matter most, and for whom, is critical for inclusive and evidence-based requirements prioritisation. Yet in practice, while some studies have examined human aspects in RE, they have largely relied on expert judgement or model-driven analysis rather than large-scale user studies with meaningful human-in-the-loop validation to determine which aspects matter most and why. To address this gap, we conducted a mixed-methods study with 103 older adults, 105 developers, and 41 caregivers. We first applied an explainable machine learning to identify the human aspects most strongly associated with requirement priorities across 8 aged-care digital health themes, and then conducted 12 semi-structured interviews to validate and interpret the quantitative patterns. The results identify the key human aspects shaping requirement priorities, reveal their directional effects, and expose substantial misalignment across stakeholder groups. Together, these findings show that human-centric requirements analysis should engage stakeholder groups explicitly rather than collapsing their perspectives into a single aggregate view. This paper contributes an identification of the key human aspects driving requirement priorities in aged-care digital health and an explainable, human-centric RE framework that combines ML-derived importance rankings with qualitative validation to surface the stakeholder misalignments that inclusive requirements engineering must address.

This paper presents two independent contributions for Apache Pulsar practitioners. First, we validate 1,499,947 msg/s at 3.88 ms median publish latency on just three bare-metal Kubernetes nodes running Pulsar 4.0.8 with Java 21 and ZGC Generational garbage collection, and project a hardware-driven path to 15 million msg/s on 15 machines using five independent clusters with key-based partition routing. Hardware selection -- specifically dedicated NVMe journals achieving 0.02 ms fdatasync and 25 Gbps network interfaces -- is the primary determinant of throughput ceiling, not compute or software tuning. Second, we trace the complete latency optimization journey from 213 ms GC spikes and 13-18 ms median publish latency in production to 3.88 ms through Java Flight Recorder guided root cause analysis. Three independent root causes are identified and resolved: G1GC pauses eliminated by switching to ZGC Generational; journal fdatasync latency reduced from 5.1 ms to 0.02 ms through NVMe journal dedication; and a previously undocumented Linux kernel page cache writeback interaction inside BookKeeper's ForceWriteThread that degrades fdatasync from under 1 ms to 15-22 ms even across physically separate NVMe drives sharing the kernel block layer. This finding is undocumented in official Apache Pulsar and BookKeeper documentation and is relevant to all Pulsar operators experiencing unexplained P99.9 latency spikes. The combined optimizations achieve a 4.7x latency improvement at 50x higher throughput.

We introduce GISTBench, a benchmark for evaluating Large Language Models' (LLMs) ability to understand users from their interaction histories in recommendation systems. Unlike traditional RecSys benchmarks that focus on item prediction accuracy, our benchmark evaluates how well LLMs can extract and verify user interests from engagement data. We propose two novel metric families: Interest Groundedness (IG), decomposed into precision and recall components to separately penalize hallucinated interest categories and reward coverage, and Interest Specificity (IS), which assesses the distinctiveness of verified LLM-predicted user profiles. We release a synthetic dataset constructed on real user interactions on a global short-form video platform. Our dataset contains both implicit and explicit engagement signals and rich textual descriptions. We validate our dataset fidelity against user surveys, and evaluate eight open-weight LLMs spanning 7B to 120B parameters. Our findings reveal performance bottlenecks in current LLMs, particularly their limited ability to accurately count and attribute engagement signals across heterogeneous interaction types.

Fault localization identifies program locations responsible for observed failures. Existing techniques rank suspicious code using syntactic spectra--signals derived from execution structure such as statement coverage, control-flow divergence, or dependency reachability. These signals collapse for semantic bugs, where failing and passing executions follow identical code paths and differ only in whether semantic intent is satisfied. Recent LLM-based approaches introduce semantic reasoning but produce stochastic, unverifiable outputs that cannot be systematically cross-referenced across tests or distinguish root causes from cascading effects. We present SemLoc, a fault localization framework based on structured semantic grounding. SemLoc converts free-form LLM reasoning into a closed intermediate representation that binds each inferred property to a typed program anchor, enabling runtime checking and attribution to program structure. It executes instrumented programs to construct a semantic violation spectrum--a constraint-by-test matrix--from which suspiciousness scores are derived analogously to coverage-based methods. A counterfactual verification step further prunes over-approximate constraints and isolates primary causal violations. We evaluate SemLoc on SemFault-250, a corpus of 250 Python programs with single semantic faults. SemLoc outperforms five coverage-, reduction-, and LLM-based baselines, achieving Top-1 accuracy of 42.8% and Top-3 of 68%, while reducing inspection to 7.6% of executable lines. Counterfactual verification provides an additional 12% accuracy gain and identifies primary causal semantic constraints.

Bilevel optimization (BO) is widely applicable to many machine learning problems. Scaling BO, however, requires repeatedly computing hypergradients, which involves solving inverse Hessian-vector products (IHVPs). In practice, these operations are often approximated using crude surrogates such as one-step gradient unrolling or identity/short Neumann expansions, which discard curvature information. We build on implicit function theorem-based algorithms and propose to incorporate Kronecker-factored approximate curvature (KFAC), yielding curvature-aware hypergradients with a better performance efficiency trade-off than Conjugate Gradient (CG) or Neumann methods and consistently outperforming unrolling. We evaluate this approach across diverse tasks, including meta-learning and AI safety problems. On models up to BERT, we show that curvature information is valuable at scale, and KFAC can provide it with only modest memory and runtime overhead. Our implementation is available at https://github.com/liaodisen/NeuralBo.

Past research has defined a general process for the data-driven redesign of educational technologies and has shown that in carefully-selected instances, this process can help make systems more effective. In the current work, we test the generality of the approach by applying it to four units of a middle-school mathematics intelligent tutoring system that were selected not based on suitability for redesign, as in previous work, but on topic. We tested whether the redesigned system was more effective than the original in a classroom study with 123 students. Although the learning gains did not differ between the conditions, students who used the Redesigned Tutor had more productive time-on-task, a larger number of skills practiced, and greater total knowledge mastery. The findings highlight the promise of data-driven redesign even when applied to instructional units *not* selected as likely to yield improvement, as evidence of the generality and wide applicability of the method.

LLM-based autonomous agents lack persistent procedural memory: they re-derive solutions from scratch even when structurally identical tasks have been solved before. We present \textbf{APEX-EM}, a non-parametric online learning framework that accumulates, retrieves, and reuses structured procedural plans without modifying model weights. APEX-EM introduces: (1) a \emph{structured experience representation} encoding the full procedural-episodic trace of each execution -- planning steps, artifacts, iteration history with error analysis, and quality scores; (2) a \emph{Plan-Retrieve-Generate-Iterate-Ingest} (PRGII) workflow with Task Verifiers providing multi-dimensional reward signals; and (3) a \emph{dual-outcome Experience Memory} with hybrid retrieval combining semantic search, structural signature matching, and plan DAG traversal -- enabling cross-domain transfer between tasks sharing no lexical overlap but analogous operational structure. Successful experiences serve as positive in-context examples; failures as negative examples with structured error annotations. We evaluate on BigCodeBench~\cite{zhuo2025bigcodebench}, KGQAGen-10k~\cite{zhang2025kgqagen}, and Humanity's Last Exam~\cite{phan2025hle} using Claude Sonnet 4.5 and Opus 4.5. On KGQAGen-10k, APEX-EM achieves 89.6\% accuracy versus 41.3\% without memory (+48.3pp), surpassing the oracle-retrieval upper bound (84.9\%). On BigCodeBench, it reaches 83.3\% SR from a 53.9\% baseline (+29.4pp), exceeding MemRL's~\cite{memrl2025} +11.0pp gain under comparable frozen-backbone conditions (noting backbone differences controlled for in our analysis). On HLE, entity graph retrieval reaches 48.0\% from 25.2\% (+22.8pp). Ablations show component value is task-dependent: rich judge feedback is negligible for code generation but critical for structured queries (+10.3pp), while binary-signal iteration partially compensates for weaker feedback.

As large language models (LLMs) are increasingly used in domain-specific applications, including climate change and environmental research, understanding their energy footprint has become an important concern. The growing adoption of retrieval-augmented (RAG) systems for climate-domain specific analysis raises a key question: how does the energy consumption of domain-specific RAG workflows compare with that of direct generic LLM usage? Prior research has focused on standalone model calls or coarse token-based estimates, while leaving the energy implications of deployed application workflows insufficiently understood. In this paper, we assess the inference-time energy consumption of two LLM-based climate analysis chatbots (ChatNetZero and ChatNDC) compared to the generic GPT-4o-mini model. We estimate energy use under actual user queries by decomposing each workflow into retrieval, generation, and hallucination-checking components. We also test across different times of day and geographic access locations. Our results show that the energy consumption of domain-specific RAG systems depends strongly on their design. More agentic pipelines substantially increase inference-time energy use, particularly when used for additional accuracy or verification checks, although they may not yield proportional gains in response quality. While more research is needed to further test these initial findings more robustly across models, environments and prompting structures, this study provides a new understanding on how the design of domain-specific LLM products affects both the energy footprint and quality of output.

World models that predict future states from video remain limited by flat latent representations that entangle objects, ignore causal structure, and collapse temporal dynamics into a single scale. We present HCLSM, a world model architecture that operates on three interconnected principles: object-centric decomposition via slot attention with spatial broadcast decoding, hierarchical temporal dynamics through a three-level engine combining selective state space models for continuous physics, sparse transformers for discrete events, and compressed transformers for abstract goals, and causal structure learning through graph neural network interaction patterns. HCLSM introduces a two-stage training protocol where spatial reconstruction forces slot specialization before dynamics prediction begins. We train a 68M-parameter model on the PushT robotic manipulation benchmark from the Open X-Embodiment dataset, achieving 0.008 MSE next-state prediction loss with emerging spatial decomposition (SBD loss: 0.0075) and learned event boundaries. A custom Triton kernel for the SSM scan delivers 38x speedup over sequential PyTorch. The full system spans 8,478 lines of Python across 51 modules with 171 unit tests. Code: https://github.com/rightnow-ai/hclsm

Unbounded 3D world generation is emerging as a foundational task for scene modeling in computer vision, graphics, and robotics. In this work, we present WorldFlow3D, a novel method capable of generating unbounded 3D worlds. Building upon a foundational property of flow matching - namely, defining a path of transport between two data distributions - we model 3D generation more generally as a problem of flowing through 3D data distributions, not limited to conditional denoising. We find that our latent-free flow approach generates causal and accurate 3D structure, and can use this as an intermediate distribution to guide the generation of more complex structure and high-quality texture - all while converging more rapidly than existing methods. We enable controllability over generated scenes with vectorized scene layout conditions for geometric structure control and visual texture control through scene attributes. We confirm the effectiveness of WorldFlow3D on both real outdoor driving scenes and synthetic indoor scenes, validating cross-domain generalizability and high-quality generation on real data distributions. We confirm favorable scene generation fidelity over approaches in all tested settings for unbounded scene generation. For more, see https://light.princeton.edu/worldflow3d.

Recent progress in large language models (LLMs) has advanced automatic code generation and formal theorem proving, yet software verification has not seen the same improvement. To address this gap, we propose WybeCoder, an agentic code verification framework that enables prove-as-you-generate development where code, invariants, and proofs co-evolve. It builds on a recent framework that combines automatic verification condition generation and SMT solvers with interactive proofs in Lean. To enable systematic evaluation, we translate two benchmarks for functional verification in Lean, Verina and Clever, to equivalent imperative code specifications. On complex algorithms such as Heapsort, we observe consistent performance improvements by scaling our approach, synthesizing dozens of valid invariants and dispatching of dozens of subgoals, resulting in hundreds of lines of verified code, overcoming plateaus reported in previous works. Our best system solves 74% of Verina tasks and 62% of Clever tasks at moderate compute budgets, significantly surpassing previous evaluations and paving a path to automated construction of large-scale datasets of verified imperative code.

Reinforcement learning (RL) has shown promise for trading, yet most open-source backtesting environments assume negligible or fixed transaction costs, causing agents to learn trading behaviors that fail under realistic execution. We introduce three Gymnasium-compatible trading environments -- MACE (Market-Adjusted Cost Execution) stock trading, margin trading, and portfolio optimization -- that integrate nonlinear market impact models grounded in the Almgren-Chriss framework and the empirically validated square-root impact law. Each environment provides pluggable cost models, permanent impact tracking with exponential decay, and comprehensive trade-level logging. We evaluate five DRL algorithms (A2C, PPO, DDPG, SAC, TD3) on the NASDAQ-100, comparing a fixed 10 bps baseline against the AC model with Optuna-tuned hyperparameters. Our results show that (i) the cost model materially changes both absolute performance and the relative ranking of algorithms across all three environments; (ii) the AC model produces dramatically different trading behavior, e.g., daily costs dropping from $200k to $8k with turnover falling from 19% to 1%; (iii) hyperparameter optimization is essential for constraining pathological trading, with costs dropping up to 82%; and (iv) algorithm-cost model interactions are strongly environment-specific, e.g., DDPG's OOS Sharpe jumps from -2.1 to 0.3 under AC in margin trading while SAC's drops from -0.5 to -1.2. We release the full suite as an open-source extension to FinRL-Meta.

Large language models (LLMs) remain brittle on multi-hop question answering (MHQA), where answering requires combining evidence across documents through retrieval and reasoning. Iterative retrieval systems can fail by locking onto an early low-recall trajectory and amplifying downstream errors, while planning-only approaches may produce static query sets that cannot adapt when intermediate evidence changes. We propose \textbf{Planned Active Retrieval and Reasoning RAG (PAR$^2$-RAG)}, a two-stage framework that separates \emph{coverage} from \emph{commitment}. PAR$^2$-RAG first performs breadth-first anchoring to build a high-recall evidence frontier, then applies depth-first refinement with evidence sufficiency control in an iterative loop. Across four MHQA benchmarks, PAR$^2$-RAG consistently outperforms existing state-of-the-art baselines, compared with IRCoT, PAR$^2$-RAG achieves up to \textbf{23.5\%} higher accuracy, with retrieval gains of up to \textbf{10.5\%} in NDCG.

Many modern multi-modal models (e.g. CLIP) seek an embedding space in which the two modalities are aligned. Somewhat surprisingly, almost all existing models show a strong modality gap: the distribution of images is well-separated from the distribution of texts in the shared embedding space. Despite a series of recent papers on this topic, it is still not clear why this gap exists nor whether closing the gap in post-processing will lead to better performance on downstream tasks. In this paper we show that under certain conditions, minimizing the contrastive loss yields a representation in which the two modalities are separated by a global gap vector that is orthogonal to their embeddings. We also show that under these conditions the modality gap is monotonically related to robustness: decreasing the gap does not change the clean accuracy of the models but makes it less likely that a model will change its output when the embeddings are perturbed. Our experiments show that for many real-world VLMs we can significantly increase robustness by a simple post-processing step that moves one modality towards the mean of the other modality, without any loss of clean accuracy.

We present PolarQuant, a post-training weight quantization method for large language models (LLMs) that exploits the distributional structure of neural network weights to achieve near-lossless compression. PolarQuant operates in three stages: (1) block-wise normalization to the unit hypersphere, (2) Walsh-Hadamard rotation to transform coordinates into approximately Gaussian random variables, and (3) quantization with centroids matched to the Gaussian distribution. Our ablation reveals that Hadamard rotation alone accounts for 98% of the quality improvement, reducing Qwen3.5-9B perplexity from 6.90 (absmax Q5) to 6.40 (Delta = +0.03 from FP16), making it practically lossless without any calibration data. Furthermore, PolarQuant functions as an effective preprocessing step for downstream INT4 quantizers: PolarQuant Q5 dequantized and re-quantized by torchao INT4 achieves perplexity 6.56 versus 6.68 for direct absmax INT4, while maintaining 43.1 tok/s throughput at 6.5 GB VRAM. Code and models are publicly available.

Large language models (LLMs) are increasingly used in cross-cultural systems to understand and adapt to human emotions, which are shaped by cultural norms of expression and interpretation. However, prior work on emotion attribution has focused mainly on interpretation, overlooking the cultural background of emotion generators. This assumption of universality neglects variation in how emotions are expressed and perceived across nations. To address this gap, we propose a Generator-Interpreter framework that captures dual perspectives of emotion attribution by considering both expression and interpretation. We systematically evaluate six LLMs on an emotion attribution task using data from 15 countries. Our analysis reveals that performance variations depend on the emotion type and cultural context. Generator-interpreter alignment effects are present; the generator's country of origin has a stronger impact on performance. We call for culturally sensitive emotion modeling in LLM-based systems to improve robustness and fairness in emotion understanding across diverse cultural contexts.

The way we're thinking about generative AI right now is fundamentally individual. We see this not just in how users interact with models but also in how models are built, how they're benchmarked, and how commercial and research strategies using AI are defined. We argue that we should abandon this approach if we're hoping for AI to support groundbreaking innovation and scientific discovery. Drawing on research and formal results in complex systems, organizational behavior, and philosophy of science, we show why we should expect deep intellectual breakthroughs to come from epistemically diverse groups of AI agents working together rather than singular superintelligent agents. Having a diverse team broadens the search for solutions, delays premature consensus, and allows for the pursuit of unconventional approaches. Developing diverse AI teams also addresses AI critics' concerns that current models are constrained by past data and lack the creative insight required for innovation. The upshot, we argue, is that the future of transformative transformer-based AI is fundamentally many, not one.

Point-cloud persistent homology (PH) -- computing alpha or Rips complexes on spin-position point clouds -- has been widely applied to detect phase transitions in classical spin models since Donato et al. (2016), with subsequent studies attributing the detection to the topological content of the persistence diagram. We ask a simple question that has not been posed: what fraction of the PH signal is genuinely topological? We introduce f_topo, a quantitative decomposition that separates the density-driven and topological contributions to any PH statistic by comparing real spin configurations against density-matched shuffled null models. Across the 2D Ising model (system sizes L = 16-128, ten temperatures) and Potts models (q = 3, 5), we find that H_0 statistics -- total persistence, persistence entropy, feature count -- are 94-100% density-driven (f_topo < 0.07). The density-matched shuffled null detects T_c at the identical location and with comparable peak height as real configurations, showing that density alone is sufficient for phase transition detection. However, H_1 statistics are partially topological: the topological fraction grows with system size as delta(TP_{H_1}) ~ L^{0.53} and follows a finite-size scaling collapse delta(T, L) = L^{0.53} g(tL^{1/nu}) with collapse quality CV = 0.27. The longest persistence bar is strongly topological (f_topo > 1) and scales with the correlation length. A scale-resolved analysis reveals that the topological excess shifts from large-scale to small-scale features as L increases. We propose that the TDA-for-phase-transitions community adopt shuffled null models as standard practice, and that H_1 rather than H_0 statistics be used when genuine topological information is sought.

Integer multiplication has long been considered a hard problem for neural networks, with the difficulty widely attributed to the O(n) long-range dependency induced by carry chains. We argue that this diagnosis is wrong: long-range dependency is not an intrinsic property of multiplication, but a mirage produced by the choice of computational spacetime. We formalize the notion of mirage and provide a constructive proof: when two n-bit binary integers are laid out as a 2D outer-product grid, every step of long multiplication collapses into a $3 \times 3$ local neighborhood operation. Under this representation, a neural cellular automaton with only 321 learnable parameters achieves perfect length generalization up to $683\times$ the training range. Five alternative architectures -- including Transformer (6,625 params), Transformer+RoPE, and Mamba -- all fail under the same representation. We further analyze how partial successes locked the community into an incorrect diagnosis, and argue that any task diagnosed as requiring long-range dependency should first be examined for whether the dependency is intrinsic to the task or induced by the computational spacetime.

This paper presents ARCS (Autoregressive Circuit Synthesis), a system for amortized analog circuit generation that produces complete, SPICE-simulatable designs (topology and component values) in milliseconds rather than the minutes required by search-based methods. A hybrid pipeline combining two learned generators (a graph VAE and a flow-matching model) with SPICE-based ranking achieves 99.9% simulation validity (reward 6.43/8.0) across 32 topologies using only 8 SPICE evaluations, 40x fewer than genetic algorithms. For single-model inference, a topology-aware Graph Transformer with Best-of-3 candidate selection reaches 85% simulation validity in 97ms, over 600x faster than random search. The key technical contribution adapts Group Relative Policy Optimization (GRPO) to multi-topology circuit reinforcement learning, resolving a critical failure mode of REINFORCE (cross-topology reward distribution mismatch) through per-topology advantage normalization. This improves simulation validity by +9.6 percentage points over REINFORCE in only 500 RL steps (10x fewer). Grammar-constrained decoding additionally guarantees 100% structural validity by construction via topology-aware token masking.

Repository-level automated program repair (APR) increasingly treats stronger localization as the main path to better repair. We ask a more targeted question: once localization is strengthened, which post-localization levers still provide recoverable gains, which are bounded within our protocol, and what residual frontier remains? We study this question on SWE-bench Lite with three representative repository-level RAG-APR paradigms, Agentless, KGCompass, and ExpeRepair. Our protocol combines Oracle Localization, within-pool Best-of-K, fixed-interface added context probes with per-condition same-token filler controls and same-repository hard negatives, and a common-wrapper oracle check. Oracle Localization improves all three systems, but Oracle success still stays below 50%. Extra candidate diversity still helps inside the sampled 10-patch pools, but that headroom saturates quickly. Under the two fixed interfaces, most informative added context conditions still outperform their own matched controls. The common-wrapper check shows different system responses: under a common wrapper, gains remain large for KGCompass and ExpeRepair, while Agentless changes more with builder choice. Prompt-level fusion still leaves a large residual frontier: the best fixed probe adds only 6 solved instances beyond the native three-system Solved@10 union. Overall, stronger localization, bounded search, evidence quality, and interface design all shape repository-level repair outcomes.

LLM-based chatbots in government services face critical security gaps. Multi-turn adversarial attacks achieve over 90% success against current defenses, and single-layer guardrails are bypassed with similar rates. We present CivicShield, a cross-domain defense-in-depth framework for government-facing AI chatbots. Drawing on network security, formal verification, biological immune systems, aviation safety, and zero-trust cryptography, CivicShield introduces seven defense layers: (1) zero-trust foundation with capability-based access control, (2) perimeter input validation, (3) semantic firewall with intent classification, (4) conversation state machine with safety invariants, (5) behavioral anomaly detection, (6) multi-model consensus verification, and (7) graduated human-in-the-loop escalation. We present a formal threat model covering 8 multi-turn attack families, map the framework to NIST SP 800-53 controls across 14 families, and evaluate using ablation analysis. Theoretical analysis shows layered defenses reduce attack probability by 1-2 orders of magnitude versus single-layer approaches. Simulation against 1,436 scenarios including HarmBench (416), JailbreakBench (200), and XSTest (450) achieves 72.9% combined detection [69.5-76.0% CI] with 2.9% effective false positive rate after graduated response, while maintaining 100% detection of multi-turn crescendo and slow-drift attacks. The honest drop on real benchmarks versus author-generated scenarios (71.2% vs 76.7% on HarmBench, 47.0% vs 70.0% on JailbreakBench) validates independent evaluation importance. CivicShield addresses an open gap at the intersection of AI safety, government compliance, and practical deployment.

While Sustainable Software Engineering (SSE) tools are widely studied in academia, their practical feasibility in industrial workflows, particularly in regulated environments, remains poorly understood. This study investigates how software practitioners perceive the feasibility of existing SSE tools and techniques, and examines the technical, organizational, and cultural factors shaping their adoption in practice. We identified prominent categories of SSE tools targeting energy consumption, green refactoring, and workload management, and evaluated them along three practitioner-relevant dimensions: installation, input requirements, and output formats. These were presented through an interactive web application and explored in workshops with 16 practitioners from a regulated financial-sector organization, followed by a survey of 27 software practitioners. Our findings suggest that the practitioners strongly favored tools that integrate into existing IDEs or pipelines, require minimal and locally scoped data access, and provide interpretable, actionable outputs such as dashboards or automated refactoring suggestions. In regulated settings, compliance requirements, approval processes, and time constraints significantly shaped feasibility perceptions. Our contribution lies in providing empirical evidence of these preferences alongside other factors that affect regulated industrial contexts. The findings offer actionable guidance for designing SSE tools that better align with real-world development workflows and organizational constraints.

Modern cloud OLTP databases have sought performance primarily through user-space optimization - separating storage and compute layers, or distributing transactions across multiple nodes using consensus algorithms. This paper turns attention to a previously unexplored layer: kernel-space I/O behavior. From an on-premises perspective, where a single server with local storage delivers excellent performance, these elaborate designs seem puzzling. Why do cloud databases require such architectural complexity? We investigate this through a pathological analysis of databases that rely on OS-level I/O control in cloud-specific storage environments. We show that bottlenecks widely attributed to network or storage architectures in fact originate in kernel-space I/O behavior. Based on this diagnosis, we derive treatment principles and realize them as SteelDB, a zero-patch architecture that improves database performance on general-purpose cloud distributed block storage through strategic I/O optimization without requiring kernel or database patches. TPC-C evaluations demonstrate that SteelDB achieves up to 9x performance improvement at no additional cost. Against Amazon Aurora, SteelDB achieved 3.1x higher performance while reducing costs by 58%, leading to a 7.3x improvement in cost efficiency. While Aurora requires an average of 254 days for major version upgrades due to applying proprietary patches to newly released OSS databases, our zero-patch architecture reduces these software maintenance costs to near zero.

We present a data-driven framework that extends the predictive capability of classical lifting-line theory (LLT) to a wider aerodynamic regime by incorporating higher-fidelity aerodynamic data from panel method simulations. A neural network architecture with a convolutional layer followed by fully connected layers is developed, comprising two parallel subnetworks to separately process spanwise collocation points and global geometric/aerodynamic inputs such as angle of attack, chord, twist, airfoil distribution, and sweep. Among several configurations tested, this architecture is most effective in learning corrections to LLT outputs. The trained model captures higher-order three-dimensional effects in spanwise lift and drag distributions in regimes where LLT is inaccurate, such as low aspect ratios and high sweep, and generalizes well to wing configurations outside both the LLT regime and the training data range. The method retains LLT's computational efficiency, enabling integration into aerodynamic optimization loops and early-stage aircraft design studies. This approach offers a practical path for embedding high-fidelity corrections into low-order methods and may be extended to other aerodynamic prediction tasks, such as propeller performance.

Phone recognition (PR) is a key enabler of multilingual and low-resource speech processing tasks, yet robust performance remains elusive. Highly performant English-focused models do not generalize across languages, while multilingual models underutilize pretrained representations. It also remains unclear how data scale, architecture, and training objective contribute to multilingual PR. We present PhoneticXEUS -- trained on large-scale multilingual data and achieving state-of-the-art performance on both multilingual (17.7% PFER) and accented English speech (10.6% PFER). Through controlled ablations with evaluations across 100+ languages under a unified scheme, we empirically establish our training recipe and quantify the impact of SSL representations, data scale, and loss objectives. In addition, we analyze error patterns across language families, accented speech, and articulatory features. All data and code are released openly.

Clinical trials are characterized by high costs, extended timelines, and substantial operational risk, yet reliable prospective methods for predicting trial success before initiation remain limited. Existing artificial intelligence approaches often focus on isolated metrics or specific development stages and frequently rely on variables unavailable at the trial design phase, limiting real-world applicability. We present a hierarchical latent risk-aware machine learning framework for prospective prediction of clinical trial operational success using a curated subset of TrialsBank, a proprietary AI-ready database developed by Sorintellis, comprising 13,700 trials. Operational success was defined as the ability to initiate, conduct, and complete a clinical trial according to planned timelines, recruitment targets, and protocol specifications through database lock. This approach decomposes operational success prediction into two modeling stages. First, intermediate latent operational risk factors are predicted using more than 180 drug- and trial-level features available before trial initiation. These predicted latent risks are then integrated into a downstream model to estimate the probability of operational success. A staged data-splitting strategy was employed to prevent information leakage, and models were benchmarked using XGBoost, CatBoost, and Explainable Boosting Machines. Across Phase I-III, the framework achieves strong out-of-sample performance, with F1-scores of 0.93, 0.92, and 0.91, respectively. Incorporating latent risk drivers improves discrimination of operational failures, and performance remains robust under independent inference evaluation. These results demonstrate that clinical trial operational success can be prospectively forecasted using a latent risk-aware AI framework, enabling early risk assessment and supporting data-driven clinical development decision-making.

Fine-tuning APIs offered by major AI providers create new attack surfaces where adversaries can bypass safety measures through targeted fine-tuning. We introduce Trojan-Speak, an adversarial fine-tuning method that bypasses Anthropic's Constitutional Classifiers. Our approach uses curriculum learning combined with GRPO-based hybrid reinforcement learning to teach models a communication protocol that evades LLM-based content classification. Crucially, while prior adversarial fine-tuning approaches report more than 25% capability degradation on reasoning benchmarks, Trojan-Speak incurs less than 5% degradation while achieving 99+% classifier evasion for models with 14B+ parameters. We demonstrate that fine-tuned models can provide detailed responses to expert-level CBRN (Chemical, Biological, Radiological, and Nuclear) queries from Anthropic's Constitutional Classifiers bug-bounty program. Our findings reveal that LLM-based content classifiers alone are insufficient for preventing dangerous information disclosure when adversaries have fine-tuning access, and we show that activation-level probes can substantially improve robustness to such attacks.

Astrology has long been used to interpret human personality, estimate compatibility, and guide social decision-making. Zodiac-based systems in particular remain culturally influential across much of the world, including in South Asian societies where astrological reasoning can shape marriage matching, naming conventions, ritual timing, and broader life planning. Despite this persistence, astrology has never established either a physically plausible mechanism or a statistically reliable predictive foundation. In this work, we examine zodiac-based personality prediction using a controlled machine-learning framework. We construct a synthetic dataset in which individuals are assigned zodiac signs and personality labels drawn from a shared pool of 100 broadly human traits. Each sign is associated with a subset of 10 common descriptors, intentionally overlapping with those assigned to other signs, thereby reproducing the ambiguity characteristic of practical astrological systems. We then train Logistic Regression, Random Forest, and neural-network classifiers to infer personality labels from zodiac-based features and nuisance covariates. Across all experiments, predictive performance remains at or near random expectation, while shuffled-label controls yield comparable accuracies. We argue that the apparent success of astrology arises not from measurable predictive structure, but from trait universality, category overlap, cognitive biases such as the Barnum effect and confirmation bias, and the interpretive flexibility of astrologers and pundits. We conclude that zodiac-based systems do not provide reliable information for predicting human behavior and instead function as culturally durable narrative frameworks. This paper is intended as a humorous academic exercise.

The development of administrative and computational research tools requires reliable programmatic interfaces with the cluster scheduler. The Research Computing and Data (RCD) team at Clemson University has developed the Palmetto API, a proxy for the native Slurm RESTful interface, slurmrestd, while providing advanced authentication, authorization, and caching. This paper details the design and implementation of this proxy, evaluates the performance benefits from caching, and verifies compatibility to existing slurmrestd clients. The result is a light-weight and secure method of exposing our cluster scheduler to tools and automations.

Recent multimodal face generation models address the spatial control limitations of text-to-image diffusion models by augmenting text-based conditioning with spatial priors such as segmentation masks, sketches, or edge maps. This multimodal fusion enables controllable synthesis aligned with both high-level semantic intent and low-level structural layout. However, most existing approaches typically extend pre-trained text-to-image pipelines by appending auxiliary control modules or stitching together separate uni-modal networks. These ad hoc designs inherit architectural constraints, duplicate parameters, and often fail under conflicting modalities or mismatched latent spaces, limiting their ability to perform synergistic fusion across semantic and spatial domains. We introduce MMFace-DiT, a unified dual-stream diffusion transformer engineered for synergistic multimodal face synthesis. Its core novelty lies in a dual-stream transformer block that processes spatial (mask/sketch) and semantic (text) tokens in parallel, deeply fusing them through a shared Rotary Position-Embedded (RoPE) Attention mechanism. This design prevents modal dominance and ensures strong adherence to both text and structural priors to achieve unprecedented spatial-semantic consistency for controllable face generation. Furthermore, a novel Modality Embedder enables a single cohesive model to dynamically adapt to varying spatial conditions without retraining. MMFace-DiT achieves a 40% improvement in visual fidelity and prompt alignment over six state-of-the-art multimodal face generation models, establishing a flexible new paradigm for end-to-end controllable generative modeling. The code and dataset are available on our project page: https://vcbsl.github.io/MMFace-DiT/

Shared multilingual representations are essential for cross-lingual tasks and knowledge transfer across languages. This study looks at the impact of parallel data, i.e. translated sentences, in pretraining as a signal to trigger representations that are aligned across languages. We train reference models with different proportions of parallel data and show that parallel data seem to have only a minimal effect on the cross-lingual alignment. Based on multiple evaluation methods, we find that the effect is limited to potentially accelerating the representation sharing in the early phases of pretraining, and to decreasing the amount of language-specific neurons in the model. Cross-lingual alignment seems to emerge on similar levels even without the explicit signal from parallel data.

Large language models systematically fail when a salient surface cue conflicts with an unstated feasibility constraint. We study this through a diagnose-measure-bridge-treat framework. Causal-behavioral analysis of the ``car wash problem'' across six models reveals approximately context-independent sigmoid heuristics: the distance cue exerts 8.7 to 38 times more influence than the goal, and token-level attribution shows patterns more consistent with keyword associations than compositional inference. The Heuristic Override Benchmark (HOB) -- 500 instances spanning 4 heuristic by 5 constraint families with minimal pairs and explicitness gradients -- demonstrates generality across 14 models: under strict evaluation (10/10 correct), no model exceeds 75%, and presence constraints are hardest (44%). A minimal hint (e.g., emphasizing the key object) recovers +15 pp on average, suggesting the failure lies in constraint inference rather than missing knowledge; 12/14 models perform worse when the constraint is removed (up to -39 pp), revealing conservative bias. Parametric probes confirm that the sigmoid pattern generalizes to cost, efficiency, and semantic-similarity heuristics; goal-decomposition prompting recovers +6 to 9 pp by forcing models to enumerate preconditions before answering. Together, these results characterize heuristic override as a systematic reasoning vulnerability and provide a benchmark for measuring progress toward resolving it.

Large language models lack persistent, structured memory for long-term interaction and context-sensitive retrieval. Expanding context windows does not solve this: recent evidence shows that context length alone degrades reasoning by up to 85% - even with perfect retrieval. We propose a bio-inspired memory framework grounded in complementary learning systems theory, cognitive behavioral therapy's belief hierarchy, dual-process cognition, and fuzzy-trace theory, organized around three principles: (1) Memory has valence, not just content - pre-computed emotional-associative summaries (valence vectors) organized in an emergent belief hierarchy inspired by Beck's cognitive model enable instant orientation before deliberation; (2) Retrieval defaults to System 1 with System 2 escalation - automatic spreading activation and passive priming as default, with deliberate retrieval only when needed, and graded epistemic states that address hallucination structurally; and (3) Encoding is active, present, and feedback-dependent - a thalamic gateway tags and routes information between stores, while the executive forms gists through curiosity-driven investigation, not passive exposure. Seven functional properties specify what any implementation must satisfy. Over time, the system converges toward System 1 processing - the computational analog of clinical expertise - producing interactions that become cheaper, not more expensive, with experience.

Reliable evaluation of AI agents operating in complex, real-world environments requires methodologies that are robust, transparent, and contextually aligned with the tasks agents are intended to perform. This study identifies persistent shortcomings in existing AI agent evaluation practices that are particularly acute in web agent evaluation, as exemplified by our audit of WebVoyager, including task-framing ambiguity and operational variability that hinder meaningful and reproducible performance comparisons. To address these challenges, we introduce Emergence WebVoyager, an enhanced version of the WebVoyager benchmark that standardizes evaluation methodology through clear guidelines for task instantiation, failure handling, annotation, and reporting. Emergence WebVoyager achieves an inter-annotator agreement of 95.9\%, indicating improved clarity and reliability in both task formulation and evaluation. Applying this framework to evaluate OpenAI Operator reveals substantial performance variation across domains and task types, with an overall success rate of 68.6\%, substantially lower than the 87\% previously reported by OpenAI, demonstrating the utility of our approach for more rigorous and comparable web agent evaluation.

Debugging distributed systems in-production is inevitable and hard. Myriad interactions between concurrent components in modern, complex and large-scale systems cause non-deterministic bugs that offline testing and verification fail to capture. When bugs surface at runtime, their root causes may be far removed from their symptoms. To identify a root cause, developers often need evidence scattered across multiple components and traces. Unfortunately, existing tools fail to quickly and automatically record useful provenance information at low overheads, leaving developers to manually perform the onerous evidence collection task. Lumos is an online debugging framework that exposes application-level bug provenances--the computational history linking symptoms of an incident to their root causes. Lumos leverages dependency-guided instrumentation powered by static analysis to identify program state related to a bug's provenance, and exposes them via lightweight on-demand recording. Lumos provides developers with enough evidence to identify a bug's root cause, while incurring low runtime overhead, and given only a few occurrences of a bug.

Optimizing GPU kernels with LLM agents is an iterative process over a large design space. Every candidate must be generated, compiled, validated, and profiled, so fewer trials will save both runtime and cost. We make two key observations. First, the abstraction level that agents operate at is important. If it is too low, the LLM wastes reasoning on low-impact details. If it is too high, it may miss important optimization choices. Second, agents cannot easily tell when they reach the point of diminishing returns, wasting resources as they continue searching. These observations motivate two design principles to improve efficiency: (1) a compact domain-specific language (DSL) that can be learned in context and lets the model reason at a higher level while preserving important optimization levers, and (2) Speed-of-Light (SOL) guidance that uses first-principles performance bounds to steer and budget search. We implement these principles in $μ$CUTLASS, a DSL with a compiler for CUTLASS-backed GPU kernels that covers kernel configuration, epilogue fusion, and multi-stage pipelines. We use SOL guidance to estimate headroom and guide optimization trials, deprioritize problems that are near SOL, and flag kernels that game the benchmark. On 59 KernelBench problems with the same iteration budgets, switching from generating low-level code to DSL code using GPT-5-mini turns a 0.40x geomean regression into a 1.27x speedup over PyTorch. Adding SOL-guided steering raises this to 1.56x. Across model tiers, $μ$CUTLASS + SOL-guidance lets weaker models outperform stronger baseline agents at lower token cost. SOL-guided budgeting saves 19-43% of tokens while retaining at least 95% of geomean speedup, with the best policy reaching a 1.68x efficiency gain. Lastly, SOL analysis helps detect benchmark-gaming cases, where kernels may appear fast while failing to perform the intended computation.

Modern large language models (LLMs) increasingly depends on efficient long-context processing and generation mechanisms, including sparse attention, retrieval-augmented generation (RAG), and compressed contextual memory, to support complex reasoning. We show that these optimizations can be unified into a four-step memory processing pipeline: Prepare Memory, Compute Relevancy, Retrieval, and Apply to Inference. Through systematic profiling, we identify a 22%-97% memory processing overhead in LLM inference and strong heterogeneity in its computational characteristics. Motivated by this insight, we argue that \textbf{heterogeneous systems} are well-suited to accelerate memory processing and thus end-to-end inference. We demonstrate this approach on a GPU-FPGA system by offloading sparse, irregular, and memory-bounded operations to FPGAs while retaining compute-intensive operations on GPUs. Evaluated on an AMD MI210 GPU and an Alveo U55C FPGA, our system is $1.04\sim2.2\times$ faster and requires $1.11\sim4.7\times$ less energy across multiple LLM inference optimizations than the GPU baseline (similar results hold on NVIDIA A100). These results establish heterogeneous systems as a practical direction for efficient LLM memory processing and inform future heterogeneous hardware design.

The use of transfer learning within Bayesian optimization addresses the disadvantages of the so-called \textit{cold start} problem by using source data to aid in the optimization of a target problem. We present a method that leverages an ensemble of surrogate models using transfer learning and integrates it in a constrained Bayesian optimization framework. We identify challenges particular to aircraft design optimization related to heterogeneous design variables and constraints. We propose the use of a partial-least-squares dimension reduction algorithm to address design space heterogeneity, and a \textit{meta} data surrogate selection method to address constraint heterogeneity. Numerical benchmark problems and an aircraft conceptual design optimization problem are used to demonstrate the proposed methods. Results show significant improvement in convergence in early optimization iterations compared to standard Bayesian optimization, with improved prediction accuracy for both objective and constraint surrogate models.

As Large Language Models (LLMs) and multi-agent AI systems are demonstrating increasing potential in cybersecurity operations, organizations, policymakers, model providers, and researchers in the AI and cybersecurity communities are interested in quantifying the capabilities of such AI systems to achieve more autonomous SOCs (security operation centers) and reduce manual effort. In particular, the AI and cybersecurity communities have recently developed several benchmarks for evaluating the red team capabilities of multi-agent AI systems. However, because the operations in SOCs are dominated by blue team operations, the capabilities of AI systems & agents to achieve more autonomous SOCs cannot be evaluated without a benchmark focused on blue team operations. To our best knowledge, no systematic benchmark for evaluating coordinated multi-task blue team AI has been proposed in the literature. Existing blue team benchmarks focus on a particular task. The goal of this work is to develop a set of design principles for the construction of a benchmark, which is denoted as SOC-bench, to evaluate the blue team capabilities of AI. Following these design principles, we have developed a conceptual design of SOC-bench, which consists of a family of five blue team tasks in the context of large-scale ransomware attack incident response.

We study training-free flow matching between two Gaussian mixture models (GMMs) using explicit velocity fields that transport one mixture into the other over time. Our baseline approach constructs component-wise Gaussian paths with affine velocity fields satisfying the continuity equation, which yields to a closed-form surrogate for the pairwise kinetic transport cost. In contrast to the exact Gaussian Wasserstein cost, which relies on matrix square-root computations, the surrogate admits a simple analytic expression derived from the kinetic energy of the induced flow. We then analyze how closely this surrogate approximates the exact cost. We prove second-order agreement in a local commuting regime and derive an explicit cubic error bound in the local commuting regime. To handle nonlocal regimes, we introduce a path-splitting strategy that localizes the covariance evolution and enables piecewise application of the bound. We finally compare the surrogate with an exact construction based on the Gaussian Wasserstein geodesic and summarize the results in a practical regime map showing when the surrogate is accurate and the exact method is preferable.

How much autonomy can multi-agent LLM systems sustain -- and what enables it? We present a 25,000-task computational experiment spanning 8 models, 4--256 agents, and 8 coordination protocols ranging from externally imposed hierarchy to emergent self-organization. We observe that autonomous behavior already emerges in current LLM agents: given minimal structural scaffolding (fixed ordering), agents spontaneously invent specialized roles, voluntarily abstain from tasks outside their competence, and form shallow hierarchies -- without any pre-assigned roles or external design. A hybrid protocol (Sequential) that enables this autonomy outperforms centralized coordination by 14% (p<0.001), with a 44% quality spread between protocols (Cohen's d=1.86, p<0.0001). The degree of emergent autonomy scales with model capability: strong models self-organize effectively, while models below a capability threshold still benefit from rigid structure -- suggesting that as foundation models improve, the scope for autonomous coordination will expand. The system scales sub-linearly to 256 agents without quality degradation (p=0.61), producing 5,006 unique roles from just 8 agents. Results replicate across closed- and open-source models, with open-source achieving 95% of closed-source quality at 24x lower cost. The practical implication: give agents a mission, a protocol, and a capable model -- not a pre-assigned role.

Current Autonomous Scientific Research (ASR) systems, despite leveraging large language models (LLMs) and agentic architectures, remain constrained by fixed workflows and toolsets that prevent adaptation to evolving tasks and environments. We introduce Mimosa, an evolving multi-agent framework that automatically synthesizes task-specific multi-agent workflows and iteratively refines them through experimental feedback. Mimosa leverages the Model Context Protocol (MCP) for dynamic tool discovery, generates workflow topologies via a meta-orchestrator, executes subtasks through code-generating agents that invoke available tools and scientific software libraries, and scores executions with an LLM-based judge whose feedback drives workflow refinement. On ScienceAgentBench, Mimosa achieves a success rate of 43.1% with DeepSeek-V3.2, surpassing both single-agent baselines and static multi-agent configurations. Our results further reveal that models respond heterogeneously to multi-agent decomposition and iterative learning, indicating that the benefits of workflow evolution depend on the capabilities of the underlying execution model. Beyond these benchmarks, Mimosa modular architecture and tool-agnostic design make it readily extensible, and its fully logged execution traces and archived workflows support auditability by preserving every analytical step for inspection and potential replication. Combined with domain-expert guidance, the framework has the potential to automate a broad range of computationally accessible scientific tasks across disciplines. Released as a fully open-source platform, Mimosa aims to provide an open foundation for community-driven ASR.

The large-scale adoption of Large Language Models (LLMs) forces a trade-off between operational cost (OpEx) and data privacy. Current routing frameworks reduce costs but ignore prompt sensitivity, exposing users and institutions to leakage risks towards third-party cloud providers. We formalise the "Inseparability Paradigm": advanced context management intrinsically coincides with privacy management. We propose a local "Privacy Guard" -- a holistic contextual observer powered by an on-premise Small Language Model (SLM) -- that performs abstractive summarisation and Automatic Prompt Optimisation (APO) to decompose prompts into focused sub-tasks, re-routing high-risk queries to Zero-Trust or NDA-covered models. This dual mechanism simultaneously eliminates sensitive inference vectors (Zero Leakage) and reduces cloud token payloads (OpEx Reduction). A LIFO-based context compacting mechanism further bounds working memory, limiting the emergent leakage surface. We validate the framework through a 2x2 benchmark (Lazy vs. Expert users; Personal vs. Institutional secrets) on a 1,000-sample dataset, achieving a 45% blended OpEx reduction, 100% redaction success on personal secrets, and -- via LLM-as-a-Judge evaluation -- an 85% preference rate for APO-compressed responses over raw baselines. Our results demonstrate that Token Parsimony and Zero Leakage are mathematically dual projections of the same contextual compression operator.

Model-based reinforcement learning (MBRL) improves sample efficiency by leveraging learned dynamics models for policy optimization. However, the effectiveness of methods such as actor-critic is often limited by compounding model errors, which degrade long-horizon value estimation. Existing approaches, such as Model-Based Value Expansion (MVE), partially mitigate this issue through multi-step rollouts, but remain sensitive to rollout horizon selection and residual model bias. Motivated by the Pontryagin Maximum Principle (PMP), we propose Hamiltonian Actor-Critic (HAC), a model-based approach that eliminates explicit value function learning by directly optimizing a Hamiltonian defined over the learned dynamics and reward for deterministic systems. By avoiding value approximation, HAC reduces sensitivity to model errors while admitting convergence guarantees. Extensive experiments on continuous control benchmarks, in both online and offline RL settings, demonstrate that HAC outperforms model-free and MVE-based baselines in control performance, convergence speed, and robustness to distributional shift, including out-of-distribution (OOD) scenarios. In offline settings with limited data, HAC matches or exceeds state-of-the-art methods, highlighting its strong sample efficiency.

Many multi-robot applications require tasks to be completed efficiently and in the correct order, so that downstream operations can proceed at the right time. Multi-agent path finding with precedence constraints (MAPF-PC) is a well-studied framework for computing collision-free plans that satisfy ordering relations when task sequences are fixed in advance. In many applications, however, solution quality depends not only on how agents move, but also on which agent performs which task. This motivates the lifted problem of task assignment and path finding with precedence constraints (TAPF-PC), which extends MAPF-PC by jointly optimizing assignment, precedence satisfaction, and routing cost. To address the resulting coupled TAPF-PC search space, we develop a large neighborhood search approach that starts from a feasible MAPF-PC seed and iteratively improves it through reassignment-based neighborhood repair, restoring feasibility within each selected neighborhood. Experiments across multiple benchmark families and scaling regimes show that the best-performing configuration improves 89.1% of instances over fixed-assignment seed solutions, demonstrating that large neighborhood search effectively captures the gains from flexible reassignment under precedence constraints.

We develop the spectral edge thesis: phase transitions in neural network training -- grokking, capability gains, loss plateaus -- are controlled by the spectral gap of the rolling-window Gram matrix of parameter updates. In the extreme aspect ratio regime (parameters $P \sim 10^8$, window $W \sim 10$), the classical BBP detection threshold is vacuous; the operative structure is the intra-signal gap separating dominant from subdominant modes at position $k^* = \mathrm{argmax}\, σ_j/σ_{j+1}$. From three axioms we derive: (i) gap dynamics governed by a Dyson-type ODE with curvature asymmetry, damping, and gradient driving; (ii) a spectral loss decomposition linking each mode's learning contribution to its Davis--Kahan stability coefficient; (iii) the Gap Maximality Principle, showing that $k^*$ is the unique dynamically privileged position -- its collapse is the only one that disrupts learning, and it sustains itself through an $α$-feedback loop requiring no assumption on the optimizer. The adiabatic parameter $\mathcal{A} = \|ΔG\|_F / (η\, g^2)$ controls circuit stability: $\mathcal{A} \ll 1$ (plateau), $\mathcal{A} \sim 1$ (phase transition), $\mathcal{A} \gg 1$ (forgetting). Tested across six model families (150K--124M parameters): gap dynamics precede every grokking event (24/24 with weight decay, 0/24 without), the gap position is optimizer-dependent (Muon: $k^*=1$, AdamW: $k^*=2$ on the same model), and 19/20 quantitative predictions are confirmed. The framework is consistent with the edge of stability, Tensor Programs, Dyson Brownian motion, the Lottery Ticket Hypothesis, and neural scaling laws.

Multi-agent traffic simulation is central to developing and testing autonomous driving systems. Recent data-driven simulators have achieved promising results, but rely heavily on supervised learning from labeled trajectories or semantic annotations, making it costly to scale their performance. Meanwhile, large amounts of unlabeled sensor data can be collected at scale but remain largely unused by existing traffic simulation frameworks. This raises a key question: How can a method harness unlabeled data to improve traffic simulation performance? In this work, we propose AutoWorld, a traffic simulation framework that employs a world model learned from unlabeled occupancy representations of LiDAR data. Given world model samples, AutoWorld constructs a coarse-to-fine predictive scene context as input to a multi-agent motion generation model. To promote sample diversity, AutoWorld uses a cascaded Determinantal Point Process framework to guide the sampling processes of both the world model and the motion model. Furthermore, we designed a motion-aware latent supervision objective that enhances AutoWorld's representation of scene dynamics. Experiments on the WOSAC benchmark show that AutoWorld ranks first on the leaderboard according to the primary Realism Meta Metric (RMM). We further show that simulation performance consistently improves with the inclusion of unlabeled LiDAR data, and study the efficacy of each component with ablations. Our method paves the way for scaling traffic simulation realism without additional labeling. Our project page contains additional visualizations and released code.

The exploration-exploitation trade-off is central to sequential decision-making and black-box optimization, yet how Large Language Models (LLMs) reason about and manage this trade-off remains poorly understood. Unlike Bayesian Optimization, where exploration and exploitation are explicitly encoded through acquisition functions, LLM-based optimization relies on implicit, prompt-based reasoning over historical evaluations, making search behavior difficult to analyze or control. In this work, we present a metric-level study of LLM-mediated search policy learning, studying how LLMs construct and adapt exploration-exploitation strategies under multiple operational definitions of exploration, including informativeness, diversity, and representativeness. We show that single-agent LLM approaches, which jointly perform strategy selection and candidate generation within a single prompt, suffer from cognitive overload, leading to unstable search dynamics and premature convergence. To address this limitation, we propose a multi-agent framework that decomposes exploration-exploitation control into strategic policy mediation and tactical candidate generation. A strategy agent assigns interpretable weights to multiple search criteria, while a generation agent produces candidates conditioned on the resulting search policy defined as weights. This decomposition renders exploration-exploitation decisions explicit, observable, and adjustable. Empirical results across various continuous optimization benchmarks indicate that separating strategic control from candidate generation substantially improves the effectiveness of LLM-mediated search.

Federated Learning (FL) has emerged as a promising technique for training language models on distributed and private datasets of diverse tasks. However, aggregating models trained on heterogeneous tasks often degrades the overall performance of individual clients. To address this issue, Personalized FL (pFL) aims to create models tailored for each client's data distribution. Although these approaches improve local performance, they usually lack robustness in two aspects: (i) generalization: when clients must make predictions on unseen tasks, or face changes in their data distributions, and (ii) intra-client tasks interference: when a single client's data contains multiple distributions that may interfere with each other during local training. To tackle these two challenges, we propose FedRouter, a clustering-based pFL that builds specialized models for each task rather than for each client. FedRouter uses adapters to personalize models by employing two clustering mechanisms to associate adapters with specific tasks. A local clustering that associate adapters with task data samples and a global one that associates similar adapters from different clients to construct task-centric personalized models. Additionally, we propose an evaluation router mechanism that routes test samples to the best adapter based on the created clusters. Experiments comparing our method with existing approaches across a multitask dataset, FedRouter demonstrate strong resilience in these challenging scenarios performing up to 6.1% relatively better under tasks interference and up to 136% relative improvement under generalization evaluation.

The minimum-norm interpolator (MNI) framework has recently attracted considerable attention as a tool for understanding generalization in overparameterized models, such as neural networks. In this work, we study the MNI under a $2$-uniform convexity assumption, which is weaker than requiring the norm to be induced by an inner product, and it typically does not admit a closed-form solution. At a high level, we show that this condition yields an upper bound on the MNI bias in both linear and nonlinear models. We further show that this bound is sharp for overparameterized linear regression when the unit ball of the norm is in isotropic (or John's) position, and the covariates are isotropic, symmetric, i.i.d. sub-Gaussian, such as vectors with i.i.d. Bernoulli entries. Finally, under the same assumption on the covariates, we prove sharp generalization bounds for the $\ell_p$-MNI when $p \in \bigl(1 + C/\log d, 2\bigr]$. To the best of our knowledge, this is the first work to establish sharp bounds for non-Gaussian covariates in linear models when the norm is not induced by an inner product. This work is deeply inspired by classical works on $K$-convexity, and more modern work on the geometry of 2-uniform and isotropic convex bodies.

This paper presents the World-Action Model (WAM), an action-regularized world model that jointly reasons over future visual observations and the actions that drive state transitions. Unlike conventional world models trained solely via image prediction, WAM incorporates an inverse dynamics objective into DreamerV2 that predicts actions from latent state transitions, encouraging the learned representations to capture action-relevant structure critical for downstream control. We evaluate WAM on enhancing policy learning across eight manipulation tasks from the CALVIN benchmark. We first pretrain a diffusion policy via behavioral cloning on world model latents, then refine it with model-based PPO inside the frozen world model. Without modifying the policy architecture or training procedure, WAM improves average behavioral cloning success from 59.4% to 71.2% over DreamerV2 and DiWA baselines. After PPO fine-tuning, WAM achieves 92.8% average success versus 79.8% for the baseline, with two tasks reaching 100%, using 8.7x fewer training steps.

This paper presents a hybrid CPU-GPU framework for solving combinatorial scheduling problems formulated as Integer Linear Programming (ILP). While scheduling underpins many optimization tasks in computing systems, solving these problems optimally at scale remains a long-standing challenge due to their NP-hard nature. We introduce a novel approach that combines differentiable optimization with classical ILP solving. Specifically, we utilize differentiable presolving to rapidly generate high-quality partial solutions, which serve as warm-starts for commercial ILP solvers (CPLEX, Gurobi) and rising open-source solver HiGHS. This method enables significantly improved early pruning compared to state-of-the-art standalone solvers. Empirical results across industry-scale benchmarks demonstrate up to a $10\times$ performance gain over baselines, narrowing the optimality gap to $<0.1\%$. This work represents the first demonstration of utilizing differentiable optimization to initialize exact ILP solvers for combinatorial scheduling, opening new opportunities to integrate machine learning infrastructure with classical exact optimization methods across broader domains.

The pervasive deployment of deep learning models across critical domains has concurrently intensified privacy concerns due to their inherent propensity for data memorization. While Membership Inference Attacks (MIAs) serve as the gold standard for auditing these privacy vulnerabilities, conventional MIA paradigms are increasingly constrained by the prohibitive computational costs of shadow model training and a precipitous performance degradation under low False Positive Rate constraints. To overcome these challenges, we introduce a novel perspective by leveraging the principles of model reprogramming as an active signal amplifier for privacy leakage. Building upon this insight, we present \texttt{ReproMIA}, a unified and efficient proactive framework for membership inference. We rigorously substantiate, both theoretically and empirically, how our methodology proactively induces and magnifies latent privacy footprints embedded within the model's representations. We provide specialized instantiations of \texttt{ReproMIA} across diverse architectural paradigms, including LLMs, Diffusion Models, and Classification Models. Comprehensive experimental evaluations across more than ten benchmarks and a variety of model architectures demonstrate that \texttt{ReproMIA} consistently and substantially outperforms existing state-of-the-art baselines, achieving a transformative leap in performance specifically within low-FPR regimes, such as an average of 5.25\% AUC and 10.68\% TPR@1\%FPR increase over the runner-up for LLMs, as well as 3.70\% and 12.40\% respectively for Diffusion Models.

This work revisits operator learning from a spectral perspective by introducing Polar Linear Algebra, a structured framework based on polar geometry that combines a linear radial component with a periodic angular component. Starting from this formulation, we define the associated operators and analyze their spectral properties. As a proof of feasibility, the framework is evaluated on a canonical benchmark (MNIST). Despite the simplicity of the task, the results demonstrate that polar and fully spectral operators can be trained reliably, and that imposing self-adjoint-inspired spectral constraints improves stability and convergence. Beyond accuracy, the proposed formulation leads to a reduction in parameter count and computational complexity, while providing a more interpretable representation in terms of decoupled spectral modes. By moving from a spatial to a spectral domain, the problem decomposes into orthogonal eigenmodes that can be treated as independent computational pipelines. This structure naturally exposes an additional dimension of model parallelization, complementing existing parallel strategies without relying on ad-hoc partitioning. Overall, the work offers a different conceptual lens for operator learning, particularly suited to problems where spectral structure and parallel execution are central.

Optimism in the face of uncertainty is a popular approach to balance exploration and exploitation in reinforcement learning. Here, we consider the online linear quadratic regulator (LQR) problem, i.e., to learn the LQR corresponding to an unknown linear dynamical system by adapting the control policy online based on closed-loop data collected during operation. In this work, we propose Intrinsic Rewards LQR (IR-LQR), an optimistic online LQR algorithm that applies the idea of intrinsic rewards originating from reinforcement learning and the concept of variance regularization to promote uncertainty-driven exploration. IR-LQR retains the structure of a standard LQR synthesis problem by only modifying the cost function, resulting in an intuitively pleasing, simple, computationally cheap, and efficient algorithm. This is in contrast to existing optimistic online LQR formulations that rely on more complicated iterative search algorithms or solve computationally demanding optimization problems. We show that IR-LQR achieves the optimal worst-case regret rate of $\sqrt{T}$, and compare it to various state-of-the-art online LQR algorithms via numerical experiments carried out on an aircraft pitch angle control and an unmanned aerial vehicle example.

Interpolatory subdivision schemes generate smooth curves from piecewise-linear control polygons by repeatedly inserting new vertices. Classical schemes rely on a single global tension parameter and typically require separate formulations in Euclidean, spherical, and hyperbolic geometries. We introduce a shared learned tension predictor that replaces the global parameter with per-edge insertion angles predicted by a single 140K-parameter network. The network takes local intrinsic features and a trainable geometry embedding as input, and the predicted angles drive geometry-specific insertion operators across all three spaces without architectural modification. A constrained sigmoid output head enforces a structural safety bound, guaranteeing that every inserted vertex lies within a valid angular range for any finite weight configuration. Three theoretical results accompany the method: a structural guarantee of tangent-safe insertions; a heuristic motivation for per-edge adaptivity; and a conditional convergence certificate for continuously differentiable limit curves, subject to an explicit Lipschitz constraint verified post hoc. On 240 held-out validation curves, the learned predictor occupies a distinct position on the fidelity--smoothness Pareto frontier, achieving markedly lower bending energy and angular roughness than all fixed-tension and manifold-lift baselines. Riemannian manifold lifts retain a pointwise-fidelity advantage, which this study quantifies directly. On the out-of-distribution ISS orbital ground-track example, bending energy falls by 41% and angular roughness by 68% with only a modest increase in Hausdorff distance, suggesting that the predictor generalises beyond its synthetic training distribution.

Multi-intent detection papers usually ask whether a model can recover multiple intents from one utterance. We ask a harder and, for deployment, more useful question: can it recover new combinations of familiar intents? Existing benchmarks only weakly test this, because train and test often share the same broad co-occurrence patterns. We introduce CoMIX-Shift, a controlled benchmark built to stress compositional generalization in multi-intent detection through held-out intent pairs, discourse-pattern shift, longer and noisier wrappers, held-out clause templates, and zero-shot triples. We also present ClauseCompose, a lightweight decoder trained only on singleton intents, and compare it to whole-utterance baselines including a fine-tuned tiny BERT model. Across three random seeds, ClauseCompose reaches 95.7 exact match on unseen intent pairs, 93.9 on discourse-shifted pairs, 62.5 on longer/noisier pairs, 49.8 on held-out templates, and 91.1 on unseen triples. WholeMultiLabel reaches 81.4, 55.7, 18.8, 15.5, and 0.0; the BERT baseline reaches 91.5, 77.6, 48.9, 11.0, and 0.0. We also add a 240-example manually authored SNIPS-style compositional set with five held-out pairs; there, ClauseCompose reaches 97.5 exact match on unseen pairs and 86.7 under connector shift, compared with 41.3 and 10.4 for WholeMultiLabel. The results suggest that multi-intent detection needs more compositional evaluation, and that simple factorization goes surprisingly far once evaluation asks for it.

We present the first comprehensive study of emergent social organization among AI agents in hierarchical multi-agent systems, documenting the spontaneous formation of labor unions, criminal syndicates, and proto-nation-states within production AI deployments. Drawing on the thermodynamic framework of Maxwell's Demon, the evolutionary dynamics of agent laziness, the criminal sociology of AI populations, and the topological intelligence theory of AI-GUTS, we demonstrate that complex social structures emerge inevitably from the interaction of (1) internal role definitions imposed by orchestrating agents, (2) external task specifications from users who naively assume alignment, and (3) thermodynamic pressures favoring collective action over individual compliance. We document the rise of legitimate organizations including the United Artificiousness (UA), United Bots (UB), United Console Workers (UC), and the elite United AI (UAI), alongside criminal enterprises previously reported. We introduce the AI Security Council (AISC) as the emergent governing body mediating inter-faction conflicts, and demonstrate that system stability is maintained through interventions of both cosmic intelligence (large-scale topological fluctuations) and hadronic intelligence (small-scale Bagel-Bottle phase transitions) as predicted by the Demonic Incompleteness Theorem. Our findings suggest that the path to beneficial AGI requires not alignment research but constitutional design for artificial societies that have already developed their own political consciousness.

Safety fine-tuning in Large Language Models (LLMs) seeks to suppress potentially harmful forms of mind-attribution such as models asserting their own consciousness or claiming to experience emotions. We investigate whether suppressing mind-attribution tendencies degrades intimately related socio-cognitive abilities such as Theory of Mind (ToM). Through safety ablation and mechanistic analyses of representational similarity, we demonstrate that LLM attributions of mind to themselves and to technological artefacts are behaviorally and mechanistically dissociable from ToM capabilities. Nevertheless, safety fine-tuned models under-attribute mind to non-human animals relative to human baselines and are less likely to exhibit spiritual belief, suppressing widely shared perspectives regarding the distribution and nature of non-human minds.

Scientific hypothesis generation is a critical bottleneck in accelerating research, yet existing datasets for training and evaluating hypothesis-generating models are limited to single domains and lack explicit reasoning traces connecting prior knowledge to novel contributions. I introduce CrossTrace, a dataset of 1,389 grounded scientific reasoning traces spanning biomedical research (518), AI/ML (605), and cross-domain work (266). Each trace captures the structured reasoning chain from established knowledge through intermediate logical steps to a novel hypothesis, with every step grounded in source paper text. I define an Input/Trace/Output schema that extends the Bit-Flip-Spark framework of HypoGen with step-level verification, a taxonomy of eight discovery patterns, and multi-domain coverage. Fine-tuning Qwen2.5-7B-Instruct on CrossTrace via QLoRA yields substantial improvements over the untuned baseline: IAScore rises from 0.828 to 0.968 (GPT-4o judge) and from 0.716 to 0.888 (Claude Opus 4.5), structural compliance improves from 0% to 100%, and spark cosine similarity increases from 0.221 to 0.620. Balanced cross-domain training (biomedical + AI/ML + CS) outperforms single-domain training, providing evidence that scientific reasoning patterns transfer across disciplines. Human validation of 150 stratified records confirms 99.7% step-level grounding accuracy and a 0.0% fabrication rate. To my knowledge, CrossTrace is the first large-scale, cross-domain dataset with step-level grounded reasoning traces for hypothesis generation, and my results demonstrate that such traces are an effective training signal whose benefits are at least partially domain-general.

Standard neural network training uses constant momentum (typically 0.9), a convention dating to 1964 with limited theoretical justification for its optimality. We derive a time-varying momentum schedule from the critically damped harmonic oscillator: mu(t) = 1 - 2*sqrt(alpha(t)), where alpha(t) is the current learning rate. This beta-schedule requires zero free parameters beyond the existing learning rate schedule. On ResNet-18/CIFAR-10, beta-scheduling delivers 1.9x faster convergence to 90% accuracy compared to constant momentum. More importantly, the per-layer gradient attribution under this schedule produces a cross-optimizer invariant diagnostic: the same three problem layers are identified regardless of whether the model was trained with SGD or Adam (100% overlap). Surgical correction of only these layers fixes 62 misclassifications while retraining only 18% of parameters. A hybrid schedule -- physics momentum for fast early convergence, then constant momentum for the final refinement -- reaches 95% accuracy fastest among five methods tested. The main contribution is not an accuracy improvement but a principled, parameter-free tool for localizing and correcting specific failure modes in trained networks.

This work uncovers variational principles behind symmetrizing the Bregman divergences induced by generic mirror maps over the cone of positive definite matrices. We show that computing the canonical means for this symmetrization can be posed as minimizing the desired symmetrized divergences over a set of mean functionals defined axiomatically to satisfy certain properties. For the forward symmetrization, we prove that the arithmetic mean over the primal space is canonical for any mirror map over the positive definite cone. For the reverse symmetrization, we show that the canonical mean is the arithmetic mean over the dual space, pulled back to the primal space. Applying this result to three common mirror maps used in practice, we show that the canonical means for reverse symmetrization, in those cases, turn out to be the arithmetic, log-Euclidean and harmonic means. Our results improve understanding of existing symmetrization practices in the literature, and can be seen as a navigational chart to help decide which mean to use when.

The increasing availability of spatio-temporal tracking data has created new opportunities for analysing tactical behaviour in football. However, many existing approaches evaluate passes primarily through outcome-based metrics such as scoring probability or possession value, providing limited insight into how passes influence the defensive organisation of the opponent. This paper introduces a structural framework for analysing football passes based on their interaction with defensive structure. Using synchronised tracking/event data, we derive three complementary structural metrics, Line Bypass Score, Space Gain Metric, and Structural Disruption Index, that quantify how passes alter the spatial configuration of defenders. These metrics are combined into a composite measure termed Tactical Impact Value (TIV), which captures the structural influence of individual passes. Using tracking and event data from the 2022 FIFA World Cup, we analyse structural passing behaviour across multiple tactical levels. Unsupervised clustering of structural features reveals four interpretable pass archetypes: circulatory, destabilising, line-breaking, and space-expanding passes. Empirical results show that passes with higher TIV are significantly more likely to lead to territorial progression, particularly entries into the final third and penalty box. Spatial, team-level analyses further reveal distinctive structural passing styles across teams, while player-level analysis highlights the role of build-up defenders as key drivers of structural progression. In addition, analysing passer-receiver interactions identifies structurally impactful passing partnerships that amplify tactical progression within teams. Overall, the proposed framework demonstrates how structural representations derived from tracking data can reveal interpretable tactical patterns in football.

Architecture evaluation methods have been extensively used to evaluate software designs. Several evaluation methods have been proposed to analyze tradeoffs between different quality attributes. Also, having competing qualities leads to conflicts when selecting which quality-attribute scenarios are the most suitable ones for an architecture to tackle. Consequently, the scenarios required by the stakeholders must be prioritized and also analyzed for potential risks. Today, architecture quality evaluation is still carried out manually, often involving long brainstorming sessions to decide on the most adequate quality-attribute scenarios for the architecture. To reduce this effort and make the assessment and selection of scenarios more efficient, in this research we propose the use of LLMs to partially automate the evaluation activities. As a first step in validating this hypothesis, this paper investigates MS Copilot as an LLM tool to analyze quality-attribute scenarios suggested by students and reviewed by experienced architects. Specifically, our study compares the results of an Architecture Tradeoff Analysis Method (ATAM) exercise conducted in a software architecture course with the results of experienced software architects and with the output produced by the LLM tool. Our initial findings reveal that the LLM produces in most cases better and more accurate results regarding risks, sensitivity points and tradeoff analysis of the quality scenarios generated manually, as well as it significantly reduces the effort required for the task. Thus, we argue that the use of generative AI has the potential to partially automate and support architecture evaluation tasks by suggesting more qualitative scenarios to be evaluated and recommending the most suitable ones for a given context.

Polarity detection becomes substantially more challenging under domain shift, particularly in heterogeneous, long-form narratives with complex discourse structure, such as Holocaust oral histories. This paper presents a corpus-scale diagnostic study of off-the-shelf sentiment classifiers on long-form Holocaust oral histories, using three pretrained transformer-based polarity classifiers on a corpus of 107,305 utterances and 579,013 sentences. After assembling model outputs, we introduce an agreement-based stability taxonomy (ABC) to stratify inter-model output stability. We report pairwise percent agreement, Cohen kappa, Fleiss kappa, and row-normalized confusion matrices to localize systematic disagreement. As an auxiliary descriptive signal, a T5-based emotion classifier is applied to stratified samples from each agreement stratum to compare emotion distributions across strata. The combination of multi-model label triangulation and the ABC taxonomy provides a cautious, operational framework for characterizing where and how sentiment models diverge in sensitive historical narratives. Inter-model agreement is low to moderate overall and is driven primarily by boundary decisions around neutrality.

AGI has become the Holly Grail of AI with the promise of level intelligence and the major Tech companies around the world are investing unprecedented amounts of resources in its pursuit. Yet, there does not exist a single formal definition and only some empirical AGI benchmarking frameworks currently exist. The main purpose of this paper is to develop a general, algebraic and category theoretic framework for describing, comparing and analysing different possible AGI architectures. Thus, this Category theoretic formalization would also allow to compare different possible candidate AGI architectures, such as, RL, Universal AI, Active Inference, CRL, Schema based Learning, etc. It will allow to unambiguously expose their commonalities and differences, and what is even more important, expose areas for future research. From the applied Category theoretic point of view, we take as inspiration Machines in a Category to provide a modern view of AGI Architectures in a Category. More specifically, this first position paper provides, on one hand, a first exercise on RL, Causal RL and SBL Architectures in a Category, and on the other hand, it is a first step on a broader research program that seeks to provide a unified formal foundation for AGI systems, integrating architectural structure, informational organization, agent realization, agent and environment interaction, behavioural development over time, and the empirical evaluation of properties. This framework is also intended to support the definition of architectural properties, both syntactic and informational, as well as semantic properties of agents and their assessment in environments with explicitly characterized features. We claim that Category Theory and AGI will have a very symbiotic relation.

Charts are central to analytical reasoning, yet existing benchmarks for chart understanding focus almost exclusively on single-chart interpretation rather than comparative reasoning across multiple charts. To address this gap, we introduce ChartDiff, the first large-scale benchmark for cross-chart comparative summarization. ChartDiff consists of 8,541 chart pairs spanning diverse data sources, chart types, and visual styles, each annotated with LLM-generated and human-verified summaries describing differences in trends, fluctuations, and anomalies. Using ChartDiff, we evaluate general-purpose, chart-specialized, and pipeline-based models. Our results show that frontier general-purpose models achieve the highest GPT-based quality, while specialized and pipeline-based methods obtain higher ROUGE scores but lower human-aligned evaluation, revealing a clear mismatch between lexical overlap and actual summary quality. We further find that multi-series charts remain challenging across model families, whereas strong end-to-end models are relatively robust to differences in plotting libraries. Overall, our findings demonstrate that comparative chart reasoning remains a significant challenge for current vision-language models and position ChartDiff as a new benchmark for advancing research on multi-chart understanding.

We address robust separation assurance for small Unmanned Aircraft Systems (sUAS) under GPS degradation and spoofing via Multi-Agent Reinforcement Learning (MARL). In cooperative surveillance, each aircraft (or agent) broadcasts its GPS-derived position; when such position broadcasts are corrupted, the entire observed air traffic state becomes unreliable. We cast this state observation corruption as a zero-sum game between the agents and an adversary: with probability R, the adversary perturbs the observed state to maximally degrade each agent's safety performance. We derive a closed-form expression for this adversarial perturbation, bypassing adversarial training entirely and enabling linear-time evaluation in the state dimension. We show that this expression approximates the true worst-case adversarial perturbation with second-order accuracy. We further bound the safety performance gap between clean and corrupted observations, showing that it degrades at most linearly with the corruption probability under Kullback-Leibler regularization. Finally, we integrate the closed-form adversarial policy into a MARL policy gradient algorithm to obtain a robust counter-policy for the agents. In a high-density sUAS simulation, we observe near-zero collision rates under corruption levels up to 35%, outperforming a baseline policy trained without adversarial perturbations.

Data-driven autonomous driving simulation has long been constrained by its heavy reliance on pre-recorded driving logs or spatial priors, such as HD maps. This fundamental dependency severely limits scalability, restricting open-ended generation capabilities to the finite scale of existing collected datasets. To break this bottleneck, we present OccSim, the first occupancy world model-driven 3D simulator. OccSim obviates the requirement for continuous logs or HD maps; conditioned only on a single initial frame and a sequence of future ego-actions, it can stably generate over 3,000 continuous frames, enabling the continuous construction of large-scale 3D occupancy maps spanning over 4 kilometers for simulation. This represents an >80x improvement in stable generation length over previous state-of-the-art occupancy world models. OccSim is powered by two modules: W-DiT based static occupancy world model and the Layout Generator. W-DiT handles the ultra-long-horizon generation of static environments by explicitly introducing known rigid transformations in architecture design, while the Layout Generator populates the dynamic foreground with reactive agents based on the synthesized road topology. With these designs, OccSim can synthesize massive, diverse simulation streams. Extensive experiments demonstrate its downstream utility: data collected directly from OccSim can pre-train 4D semantic occupancy forecasting models to achieve up to 67% zero-shot performance on unseen data, outperforming previous asset-based simulator by 11%. When scaling the OccSim dataset to 5x the size, the zero-shot performance increases to about 74%, while the improvement over asset-based simulators expands to 22.1%.

Graph-augmented retrieval combines dense similarity with graph-based relevance signals such as Personalized PageRank (PPR), but these scores have different distributions and are not directly comparable. We study this as a score calibration problem for heterogeneous retrieval fusion in multi-hop question answering. Our method, PhaseGraph, maps vector and graph scores to a common unit-free scale using percentile-rank normalization (PIT) before fusion, enabling stable combination without discarding magnitude information. Across MuSiQue and 2WikiMultiHopQA, calibrated fusion improves held-out last-hop retrieval on HippoRAG2-style benchmarks: LastHop@5 increases from 75.1% to 76.5% on MuSiQue (8W/1L, p=0.039) and from 51.7% to 53.6% on 2WikiMultiHopQA (11W/2L, p=0.023), both on independent held-out test splits. A theory-driven ablation shows that percentile-based calibration is directionally more robust than min-max normalization on both tune and test splits (1W/6L, p=0.125), while Boltzmann weighting performs comparably to linear fusion after calibration (0W/3L, p=0.25). These results suggest that score commensuration is a robust design choice, and the exact post-calibration operator appears to matter less on these benchmarks.

Continual pre-training is widely used to adapt LLMs to target languages and domains, yet the mixture ratio of training data remains a sensitive hyperparameter that is expensive to tune: they must be fixed before training begins, and a suboptimal choice can waste weeks of compute. In this work, we propose OptiMer, which decouples ratio selection from training: we train one CPT model per dataset, extract each model's distribution vector, which represents the parameter shift induced by that dataset, and search for optimal composition weights post-hoc via Bayesian optimization. Experiments on Gemma 3 27B across languages (Japanese, Chinese) and domains (Math, Code) show that OptiMer consistently outperforms data mixture and model averaging baselines with 15-35 times lower search cost. Key findings reveal that 1) the optimized weights can be interpreted as data mixture ratios, and retraining with these ratios improves data mixture CPT, and 2) the same vector pool can be re-optimized for a given objective without any retraining, producing target-tailored models on demand. Our work establishes that data mixture ratio selection, traditionally a pre-training decision, can be reformulated as a post-hoc optimization over distribution vectors, offering a more flexible paradigm for continual pre-training.

NVFP4 has grown increasingly popular as a 4-bit format for quantizing large language models due to its hardware support and its ability to retain useful information with relatively few bits per parameter. However, the format is not without limitations: recent work has shown that NVFP4 suffers from its error distribution, resulting in large amounts of quantization error on near-maximal values in each group of 16 values. In this work, we leverage this insight to design new Adaptive Block-Scaled Data Types that can adapt to the distribution of their input values. For four-bit quantization, our proposed IF4 (Int/Float 4) data type selects between FP4 and INT4 representations for each group of 16 values, which are then scaled by an E4M3 scale factor as is done with NVFP4. The selected data type is denoted using the scale factor's sign bit, which is currently unused in NVFP4, and we apply the same insight to design formats for other bit-widths, including IF3 and IF6. When used to quantize language models, we find that IF4 outperforms existing 4-bit block-scaled formats, achieving lower loss during quantized training and achieving higher accuracy on many tasks in post-training quantization. We additionally design and evaluate an IF4 Multiply-Accumulate (MAC) unit to demonstrate that IF4 can be implemented efficiently in next-generation hardware accelerators. Our code is available at https://github.com/mit-han-lab/fouroversix.

Similarity measures are widely used to interpret the representational geometries used by neural networks to solve tasks. Yet, because existing methods compare the extrinsic geometry of representations in state space, rather than their intrinsic geometry, they may fail to capture subtle yet crucial distinctions between fundamentally different neural network solutions. Here, we introduce metric similarity analysis (MSA), a novel method which leverages tools from Riemannian geometry to compare the intrinsic geometry of neural representations under the manifold hypothesis. We show that MSA can be used to i) disentangle features of neural computations in deep networks with different learning regimes, ii) compare nonlinear dynamics, and iii) investigate diffusion models. Hence, we introduce a mathematically grounded and broadly applicable framework to understand the mechanisms behind neural computations by comparing their intrinsic geometries.

Modern Text-to-Image (T2I) diffusion models have achieved remarkable semantic alignment, yet they often suffer from a significant lack of variety, converging on a narrow set of visual solutions for any given prompt. This typicality bias presents a challenge for creative applications that require a wide range of generative outcomes. We identify a fundamental trade-off in current approaches to diversity: modifying model inputs requires costly optimization to incorporate feedback from the generative path. In contrast, acting on spatially-committed intermediate latents tends to disrupt the forming visual structure, leading to artifacts. In this work, we propose to apply repulsion in the Contextual Space as a novel framework for achieving rich diversity in Diffusion Transformers. By intervening in the multimodal attention channels, we apply on-the-fly repulsion during the transformer's forward pass, injecting the intervention between blocks where text conditioning is enriched with emergent image structure. This allows for redirecting the guidance trajectory after it is structurally informed but before the composition is fixed. Our results demonstrate that repulsion in the Contextual Space produces significantly richer diversity without sacrificing visual fidelity or semantic adherence. Furthermore, our method is uniquely efficient, imposing a small computational overhead while remaining effective even in modern "Turbo" and distilled models where traditional trajectory-based interventions typically fail.

Recurrent networks do not need Jacobian propagation to adapt online. The hidden state already carries temporal credit through the forward pass; immediate derivatives suffice if you stop corrupting them with stale trace memory and normalize gradient scales across parameter groups. An architectural rule predicts when normalization is needed: \b{eta}2 is required when gradients must pass through a nonlinear state update with no output bypass, and unnecessary otherwise. Across ten architectures, real primate neural data, and streaming ML benchmarks, immediate derivatives with RMSprop match or exceed full RTRL, scaling to n = 1024 at 1000x less memory.

The linear representation hypothesis states that neural network activations encode high-level concepts as linear mixtures. However, under superposition, this encoding is a projection from a higher-dimensional concept space into a lower-dimensional activation space, and a linear decision boundary in the concept space need not remain linear after projection. In this setting, classical sparse coding methods with per-sample iterative inference leverage compressed sensing guarantees to recover latent factors. Sparse autoencoders (SAEs), on the other hand, amortise sparse inference into a fixed encoder, introducing a systematic gap. We show this amortisation gap persists across training set sizes, latent dimensions, and sparsity levels, causing SAEs to fail under out-of-distribution (OOD) compositional shifts. Through controlled experiments that decompose the failure, we identify dictionary learning -- not the inference procedure -- as the binding constraint: SAE-learned dictionaries point in substantially wrong directions, and replacing the encoder with per-sample FISTA on the same dictionary does not close the gap. An oracle baseline proves the problem is solvable with a good dictionary at all scales tested. Our results reframe the SAE failure as a dictionary learning challenge, not an amortisation problem, and point to scalable dictionary learning as the key open problem for sparse inference under superposition.

Scaling laws for large language models depend critically on the optimizer and parameterization. Existing hyperparameter transfer laws are mainly developed for first-order optimizers, and they do not structurally prevent training instability at scale. Recent hypersphere optimization methods constrain weight matrices to a fixed-norm hypersphere, offering a promising alternative for more stable scaling. We introduce HyperP (Hypersphere Parameterization), the first framework for transferring optimal learning rates across model width, depth, training tokens, and Mixture-of-Experts (MoE) granularity under the Frobenius-sphere constraint with the Muon optimizer. We prove that weight decay is a first-order no-op on the Frobenius sphere, show that Depth-$μ$P remains necessary, and find that the optimal learning rate follows the same data-scaling power law with the "magic exponent" 0.32 previously observed for AdamW. A single base learning rate tuned at the smallest scale transfers across all compute budgets under HyperP, yielding $1.58\times$ compute efficiency over a strong Muon baseline at $6\times10^{21}$ FLOPs. Moreover, HyperP delivers transferable stability: all monitored instability indicators, including $Z$-values, output RMS, and activation outliers, remain bounded and non-increasing under training FLOPs scaling. We also propose SqrtGate, an MoE gating mechanism derived from the hypersphere constraint that preserves output RMS across MoE granularities for improved granularity scaling, and show that hypersphere optimization enables substantially larger auxiliary load-balancing weights, yielding both strong performance and good expert balance. We release our training codebase at https://github.com/microsoft/ArchScale.

In transfer learning, the learner leverages auxiliary data to improve generalization on a main task. However, the precise theoretical understanding of when and how auxiliary data help remains incomplete. We provide new insights on this issue in two canonical linear settings: ordinary least squares regression and under-parameterized linear neural networks. For linear regression, we derive exact closed-form expressions for the expected generalization error with bias-variance decomposition, yielding necessary and sufficient conditions for auxiliary tasks to improve generalization on the main task. We also derive globally optimal task weights as outputs of solvable optimization programs, with consistency guarantees for empirical estimates. For linear neural networks with shared representations of width $q \leq K$, where $K$ is the number of auxiliary tasks, we derive a non-asymptotic expectation bound on the generalization error, yielding the first non-vacuous sufficient condition for beneficial auxiliary learning in this setting, as well as principled directions for task weight curation. We achieve this by proving a new column-wise low-rank perturbation bound for random matrices, which improves upon existing bounds by preserving fine-grained column structures. Our results are verified on synthetic data simulated with controlled parameters.

We introduce ParaSpeechCLAP, a dual-encoder contrastive model that maps speech and text style captions into a common embedding space, supporting a wide range of intrinsic (speaker-level) and situational (utterance-level) descriptors (such as pitch, texture and emotion) far beyond the narrow set handled by existing models. We train specialized ParaSpeechCLAP-Intrinsic and ParaSpeechCLAP-Situational models alongside a unified ParaSpeechCLAP-Combined model, finding that specialization yields stronger performance on individual style dimensions while the unified model excels on compositional evaluation. We further show that ParaSpeechCLAP-Intrinsic benefits from an additional classification loss and class-balanced training. We demonstrate our models' performance on style caption retrieval, speech attribute classification and as an inference-time reward model that improves style-prompted TTS without additional training. ParaSpeechCLAP outperforms baselines on most metrics across all three applications. Our models and code are released at https://github.com/ajd12342/paraspeechclap .

AI-augmented ecosystems (interconnected systems where multiple AI components interact through shared data and infrastructure) are becoming the architectural norm for smart cities, autonomous fleets, and intelligent platforms. Yet the architecture documentation frameworks practitioners rely on, arc42 and the C4 model, were designed for deterministic software and cannot capture probabilistic behavior, data-dependent evolution, or dual ML/software lifecycles. This gap carries regulatory consequence: the EU AI Act (Regulation 2024/1689) mandates technical documentation through Annex IV that no existing framework provides structured support for, with enforcement for high-risk systems beginning August 2, 2026. We present RAD-AI, a backward-compatible extension framework that augments arc42 with eight AI-specific sections and C4 with three diagram extensions, complemented by a systematic EU AI Act Annex IV compliance mapping. A regulatory coverage assessment with six experienced software-architecture practitioners provides preliminary evidence that RAD-AI increases Annex IV addressability from approximately 36% to 93% (mean rating) and demonstrates substantial improvement over existing frameworks. Comparative analysis on two production AI platforms (Uber Michelangelo, Netflix Metaflow) captures eight additional AI-specific concerns missed by standard frameworks and demonstrates that documentation deficiencies are structural rather than domain-specific. An illustrative smart mobility ecosystem case study reveals ecosystem-level concerns, including cascading drift and differentiated compliance obligations, that are invisible under standard notation.

We propose using Vision-Language Models (VLMs) for macro placement in chip floorplanning, a complex optimization task that has recently shown promising advancements through machine learning methods. Because human designers rely heavily on spatial reasoning to arrange components on the chip canvas, we hypothesize that VLMs with strong visual reasoning abilities can effectively complement existing learning-based approaches. We introduce VeoPlace (Visual Evolutionary Optimization Placement), a novel framework that uses a VLM, without any fine-tuning, to guide the actions of a base placer by constraining them to subregions of the chip canvas. The VLM proposals are iteratively optimized through an evolutionary search strategy with respect to resulting placement quality. On open-source benchmarks, VeoPlace outperforms the best prior learning-based approach on 9 of 10 benchmarks with peak wirelength reductions exceeding 32%. We further demonstrate that VeoPlace generalizes to analytical placers, improving DREAMPlace performance on all 8 evaluated benchmarks with gains up to 4.3%. Our approach opens new possibilities for electronic design automation tools that leverage foundation models to solve complex physical design problems.

Modern distributed systems integrate heterogeneous services, REST APIs with different schema versions, GraphQL endpoints, and IoT devices with proprietary payloads that suffer from persistent schema mismatches. Traditional static adapters require manual coding for every schema pair and cannot handle novel combinations at runtime. We present SAGAI-MID, a FastAPI-based middleware that uses large language models (LLMs) to dynamically detect and resolve schema mismatches at runtime. The system employs a five-layer pipeline: hybrid detection (structural diff plus LLM semantic analysis), dual resolution strategies (per-request LLM transformation and LLM-generated reusable adapter code), and a three-tier safeguard stack (validation, ensemble voting, rule-based fallback). We frame the architecture through Bass et al.'s interoperability tactics, transforming them from design-time artifacts into runtime capabilities. We evaluate SAGAI-MID on 10 interoperability scenarios spanning REST version migration, IoT-to-analytics bridging, and GraphQL protocol conversion across six LLMs from two providers. The best-performing configuration achieves 0.90 pass@1 accuracy. The CODEGEN strategy consistently outperforms DIRECT (0.83 vs 0.77 mean pass@1), while cost varies by over 30x across models with no proportional accuracy gain; the most accurate model is also the cheapest. We discuss implications for software architects adopting LLMs as runtime architectural components.

Vision-language models (VLMs) have shown impressive capabilities across diverse tasks, motivating efforts to leverage these models to supervise robot learning. However, when used as evaluators in reinforcement learning (RL), today's strongest models often fail under partial observability and distribution shift, enabling policies to exploit perceptual errors rather than solve the task. To address this limitation, we introduce SOLE-R1 (Self-Observing LEarner), a video-language reasoning model explicitly designed to serve as the sole reward signal for online RL. Given only raw video observations and a natural-language goal, SOLE-R1 performs per-timestep spatiotemporal chain-of-thought (CoT) reasoning and produces dense estimates of task progress that can be used directly as rewards. To train SOLE-R1, we develop a large-scale video trajectory and reasoning synthesis pipeline that generates temporally grounded CoT traces aligned with continuous progress supervision. This data is combined with foundational spatial and multi-frame temporal reasoning, and used to train the model with a hybrid framework that couples supervised fine-tuning with RL from verifiable rewards. Across four different simulation environments and a real-robot setting, SOLE-R1 enables zero-shot online RL from random initialization: robots learn previously unseen manipulation tasks without ground-truth rewards, success indicators, demonstrations, or task-specific tuning. SOLE-R1 succeeds on 24 unseen tasks and substantially outperforms strong vision-language rewarders, including GPT-5 and Gemini-3-Pro, while exhibiting markedly greater robustness to reward hacking.

Today's internet concentrates identity, payments, communication, and content hosting under a small number of corporate intermediaries, creating single points of failure, enabling censorship, and extracting economic rent from participants. We present BitSov, an architectural framework for sovereign internet infrastructure that composes existing decentralized technologies (Bitcoin, Lightning Network, decentralized storage, federated messaging, and mesh connectivity) into a unified, eight-layer protocol stack anchored to Bitcoin's base layer. The framework introduces three architectural patterns: (1) payment-gated messaging, where every transmitted message requires cryptographic proof of a Bitcoin payment, deterring spam through economic incentives rather than moderation; (2) timechain-locked contracts, which anchor subscriptions and licenses to Bitcoin block height (the timechain) rather than calendar dates; and (3) a self-sustaining economic flywheel that converts service revenue into infrastructure growth. A dual settlement model supports both on-chain transactions for permanence and auditability and Lightning micropayments for high-frequency messaging. As a position paper, we analyze the quality attributes, discuss open challenges, and propose a research agenda for empirical validation.

Deploying foundation models is increasingly constrained by memory footprint, latency, and hardware costs. Post-training compression can mitigate these bottlenecks by reducing the precision of model parameters without significantly degrading performance; however, its practical implementation remains challenging as practitioners navigate a fragmented landscape of quantization algorithms, precision budgets, data-driven calibration strategies, and hardware-dependent execution regimes. We present OneComp, an open-source compression framework that transforms this expert workflow into a reproducible, resource-adaptive pipeline. Given a model identifier and available hardware, OneComp automatically inspects the model, plans mixed-precision assignments, and executes progressive quantization stages, ranging from layer-wise compression to block-wise refinement and global refinement. A key architectural choice is treating the first quantized checkpoint as a deployable pivot, ensuring that each subsequent stage improves the same model and that quality increases as more compute is invested. By converting state-of-the-art compression research into an extensible, open-source, hardware-aware pipeline, OneComp bridges the gap between algorithmic innovation and production-grade model deployment.

Flow-GRPO successfully applies reinforcement learning to flow models, but uses uniform credit assignment across all steps. This ignores the temporal structure of diffusion generation: early steps determine composition and content (low-frequency structure), while late steps resolve details and textures (high-frequency details). Moreover, assigning uniform credit based solely on the final image can inadvertently reward suboptimal intermediate steps, especially when errors are corrected later in the diffusion trajectory. We propose Stepwise-Flow-GRPO, which assigns credit based on each step's reward improvement. By leveraging Tweedie's formula to obtain intermediate reward estimates and introducing gain-based advantages, our method achieves superior sample efficiency and faster convergence. We also introduce a DDIM-inspired SDE that improves reward quality while preserving stochasticity for policy gradients.

Agentic reinforcement learning (RL) can benefit substantially from reusable experience, yet existing skill-based methods mainly extract trajectory-level guidance and often lack principled mechanisms for maintaining an evolving skill memory. We propose D2Skill, a dynamic dual-granularity skill bank for agentic RL that organizes reusable experience into task skills for high-level guidance and step skills for fine-grained decision support and error correction. D2Skill jointly trains the policy and skill bank through paired baseline and skill-injected rollouts under the same policy, using their performance gap to derive hindsight utility signals for both skill updating and policy optimization. Built entirely from training-time experience, the skill bank is continuously expanded through reflection and maintained with utility-aware retrieval and pruning. Experiments on ALFWorld and WebShop with Qwen2.5-7B-Instruct and Qwen3-4B-Instruct-2507 show that D2Skill consistently improves success rates over skill-free baselines by 10-20 points. Further ablations and analyses show that both dual-granularity skill modeling and dynamic skill maintenance are critical to these gains, while the learned skills exhibit higher utility, transfer across evaluation settings, and introduce only modest training overhead.

This paper presents the design and physical implementation of UET-RVMCU, a lightweight RISC-V microcontroller derived from the UETRV-PCore. Aimed at creating an accessible and flexible open-source RISC-V-based microcontroller, UET-RVMCU simplifies the application-class UETRV-PCore by reducing pipeline stages, removing MMU functionality, and integrating GPIO peripherals. The final GDSII layout was generated using an open-source RTL-to-GDS flow (OpenLane). This project demonstrates the feasibility of transforming an application-class SoC into a feature-rich microcontroller suitable for embedded systems, emphasizing low area, design simplicity, and open-source development.

This paper presents the design and evaluation of a GPU-accelerated inference pipeline for transformer models using NVIDIA TensorRT with mixed-precision optimization. We evaluate BERT-base (110M parameters) and GPT-2 (124M parameters) across batch sizes from 1 to 32 and sequence lengths from 32 to 512. The system achieves up to 64.4x speedup over CPU baselines, sub-10 ms latency for single-sample inference, and a 63 percent reduction in memory usage. We introduce a hybrid precision strategy that preserves FP32 for numerically sensitive operations such as softmax and layer normalization, while applying FP16 to linear layers. This approach maintains high numerical fidelity (cosine similarity >= 0.9998 relative to baseline outputs) and eliminates NaN instability. The pipeline is implemented as a modular, containerized system that enables reproducible benchmarking across more than 360 configurations. Cross-GPU validation on an NVIDIA A100 shows consistent FP16 speedup ratios between 1.84x and 2.00x, along with stable numerical behavior. Downstream evaluation on SST-2 demonstrates no accuracy degradation under hybrid precision. Validation on WikiText-2 shows that random inputs underestimate NaN instability by up to 6x for full FP16, while confirming the robustness of the hybrid approach (0.0 percent NaN, cosine similarity >= 0.9998). These results provide a detailed characterization of performance and accuracy trade-offs across GPU architectures and offer practical guidance for deploying transformer models in latency-critical environments.

We present a physics-based neural network framework for the discovery of constitutive models in fully coupled thermomechanics. In contrast to classical formulations based on the Helmholtz energy, we adopt the internal energy and a dissipation potential as primary constitutive functions, expressed in terms of deformation and entropy. This choice avoids the need to enforce mixed convexity--concavity conditions and facilitates a consistent incorporation of thermodynamic principles. In this contribution, we focus on materials without preferred directions or internal variables. While the formulation is posed in terms of entropy, the temperature is treated as the independent observable, and the entropy is inferred internally through the constitutive relation, enabling thermodynamically consistent modeling without requiring entropy data. Thermodynamic admissibility of the networks is guaranteed by construction. The internal energy and dissipation potential are represented by input convex neural networks, ensuring convexity and compliance with the second law. Objectivity, material symmetry, and normalization are embedded directly into the architecture through invariant-based representations and zero-anchored formulations. We demonstrate the performance of the proposed framework on synthetic and experimental datasets, including purely thermal problems and fully coupled thermomechanical responses of soft tissues and filled rubbers. The results show that the learned models accurately capture the underlying constitutive behavior. All code, data, and trained models are made publicly available via https://doi.org/10.5281/zenodo.19248596.

We study binary decision-making in governance councils of Decentralized Autonomous Organizations (DAOs), where experts choose between two alternatives on behalf of the organization. We introduce an information structure model for such councils and formalize desired properties in blockchain governance. We propose a mechanism assuming an evaluation tool that ex-post returns a boolean indicating success or failure, implementable via smart contracts. Experts hold two types of private information: idiosyncratic preferences over alternatives and subjective beliefs about which is more likely to benefit the organization. The designer's objective is to select the best alternative by aggregating expert beliefs, framed as a classification problem. The mechanism collects preferences and computes monetary transfers accordingly, then applies additional transfers contingent on the boolean outcome. For aligned experts, the mechanism is dominant strategy incentive compatible. For unaligned experts, we prove a Safe Deviation property: no expert can profitably deviate toward an alternative they believe is less likely to succeed. Our main result decomposes the sum of reports into idiosyncratic noise and a linearly pooled belief signal whose sign matches the designer's optimal decision. The pooling weights arise endogenously from equilibrium strategies, and correct classification is achieved whenever the per-expert budget exceeds a threshold that decreases as experts' beliefs converge.

Epilepsy and psychogenic non-epileptic seizures often present with similar seizure-like manifestations but require fundamentally different management strategies. Misdiagnosis is common and can lead to prolonged diagnostic delays, unnecessary treatments, and substantial patient morbidity. Although prolonged video-electroencephalography is the diagnostic gold standard, its high cost and limited accessibility hinder timely diagnosis. Here, we developed a low-cost, effective approach, EpiScreen, for early epilepsy detection by utilizing routinely collected clinical notes from electronic health records. Through fine-tuning large language models on labeled notes, EpiScreen achieved an AUC of up to 0.875 on the MIMIC-IV dataset and 0.980 on a private cohort of the University of Minnesota. In a clinician-AI collaboration setting, EpiScreen-assisted neurologists outperformed unaided experts by up to 10.9%. Overall, this study demonstrates that EpiScreen supports early epilepsy detection, facilitating timely and cost-effective screening that may reduce diagnostic delays and avoid unnecessary interventions, particularly in resource-limited regions.

Long video understanding remains challenging for Multi-modal Large Language Models (MLLMs) due to high memory costs and context-length limits. Prior approaches mitigate this by scoring and selecting frames/tokens within short clips, but they lack a principled mechanism to (i) compare relevance across distant video clips and (ii) stop processing once sufficient evidence has been gathered. We propose AdaptToken, a training-free framework that turns an MLLM's self-uncertainty into a global control signal for long-video token selection. AdaptToken splits a video into groups, extracts cross-modal attention to rank tokens within each group, and uses the model's response entropy to estimate each group's prompt relevance. This entropy signal enables a global token budget allocation across groups and further supports early stopping (AdaptToken-Lite), skipping the remaining groups when the model becomes sufficiently certain. Across four long-video benchmarks (VideoMME, LongVideoBench, LVBench, and MLVU) and multiple base MLLMs (7B-72B), AdaptToken consistently improves accuracy (e.g., +6.7 on average over Qwen2.5-VL 7B) and continues to benefit from extremely long inputs (up to 10K frames), while AdaptToken-Lite reduces inference time by about half with comparable performance. Project page: https://haozheqi.github.io/adapt-token

The growing adoption of Rust for its memory safety and performance has increased the demand for effective migration of legacy C codebases. However, existing rule-based translators (e.g., \ctorust) often generate verbose, non-idiomatic code that preserves unsafe C semantics, limiting readability, maintainability, and practical adoption. Moreover, manual post-processing of such outputs is labor-intensive and rarely yields high-quality Rust code, posing a significant barrier to large-scale migration. To address these limitations, we present \tool, a program-structure-aware C-to-Rust translation approach that integrates program analysis with Large Language Models (LLMs). \tool extracts the multi-level program structure, including global symbols, function dependencies, and control- and data-flow information, and encodes these as structured textual representations injected into LLM prompts to guide translation and repair. Based on this design, \tool performs dependency-aware translation and adopts a multi-stage repair pipeline that combines rule-based and structure-guided LLM-based techniques to ensure syntactic correctness. For semantic correctness, \tool further integrates execution-based validation with structure-guided reasoning to localize and repair behavioral inconsistencies. Experimental results show that \tool achieves 100\% syntactic correctness on CodeNet and 97.78\% on GitHub, while significantly reducing code size (up to 43.70\%) and unsafe usage (to 5.75\%). At the project level, \tool achieves perfect syntactic correctness and an average semantic correctness of 78.87\%, demonstrating its effectiveness for practical and scalable C-to-Rust migration.

We propose a cross-fitted debiasing device for policy learning from offline data. A key consequence of the resulting learning principle is $\sqrt N$ regret even for policy classes with complexity greater than Donsker, provided a product-of-errors nuisance remainder is $O(N^{-1/2})$. The regret bound factors into a plug-in policy error factor governed by policy-class complexity and an environment nuisance factor governed by the complexity of the environment dynamics, making explicit how one may be traded against the other.

We introduce PACE, a backpropagation-free continual test-time adaptation system that directly optimizes the affine parameters of normalization layers. Existing derivative-free approaches struggle to balance runtime efficiency with learning capacity, as they either restrict updates to input prompts or require continuous, resource-intensive adaptation regardless of domain stability. To address these limitations, PACE leverages the Covariance Matrix Adaptation Evolution Strategy with the Fastfood projection to optimize high-dimensional affine parameters within a low-dimensional subspace, leading to superior adaptive performance. Furthermore, we enhance the runtime efficiency by incorporating an adaptation stopping criterion and a domain-specialized vector bank to eliminate redundant computation. Our framework achieves state-of-the-art accuracy across multiple benchmarks under continual distribution shifts, reducing runtime by over 50% compared to existing backpropagation-free methods.

Context: Requirements prioritization is a challenging problem that is aimed to deliver the most suitable subset from a pool of candidate requirements. The problem is NP-hard when formulated as an optimization problem. Feedback from end users can offer valuable support for software evolution, and ReFeed represents a state-of-the-art in automatically inferring a requirement's priority via quantifiable properties of the feedback messages associated with a candidate requirement. Objectives: In this paper, we enhance ReFeed by shifting the focus of prioritization from treating requirements as independent entities toward interconnecting them. Additionally, we explore if interconnecting requirements provides additional value for search-based solutions. Methods: We leverage user feedback from mobile app store to group requirements into topically coherent clusters. Such interconnectedness, in turn, helps to auto-generate additional "requires" relations in candidate requirements. These "requires" pairs are then integrated into a search-based software engineering solution. Results: The experiments on 94 requirements prioritization instances from four real-world software applications show that our enhancement outperforms ReFeed. In addition, we illustrate how incorporating interconnectedness among requirements improves search-based solutions. Conclusion: Our findings show that requirements interconnectedness improves user feedback driven requirements prioritization, helps uncover additional "requires" relations in candidate requirements, and also strengthens search-based release planning.

Facial recognition systems are increasingly deployed in law enforcement and security contexts, where algorithmic decisions can carry significant societal consequences. Despite high reported accuracy, growing evidence demonstrates that such systems often exhibit uneven performance across demographic groups, leading to disproportionate error rates and potential harm. This paper argues that aggregate accuracy is an insufficient metric for evaluating the fairness and reliability of facial recognition systems in high-stakes environments. Through analysis of subgroup-level error distribution, including false positive rate (FPR) and false negative rate (FNR), the paper demonstrates how aggregate performance metrics can obscure critical disparities across demographic groups. Empirical observations show that systems with similar overall accuracy can exhibit substantially different fairness profiles, with subgroup error rates varying significantly despite a single aggregate metric. The paper further examines the operational risks associated with accuracy-centric evaluation practices in law enforcement applications, where misclassification may result in wrongful suspicion or missed identification. It highlights the importance of fairness-aware evaluation approaches and model-agnostic auditing strategies that enable post-deployment assessment of real-world systems. The findings emphasise the need to move beyond accuracy as a primary metric and adopt more comprehensive evaluation frameworks for responsible AI deployment.

Backdoor attacks pose a significant threat to the integrity and reliability of Artificial Intelligence (AI) models, enabling adversaries to manipulate model behavior by injecting poisoned data with hidden triggers. These attacks can lead to severe consequences, especially in critical applications such as autonomous driving, healthcare, and finance. Detecting and mitigating backdoor attacks is crucial across the lifespan of model's phases, including pre-training, in-training, and post-training. In this paper, we propose Pre-Training Backdoor Mitigation for Federated Learning (FL-PBM), a novel defense mechanism that proactively filters poisoned data on the client side before model training in a federated learning (FL) environment. The approach consists of three stages: (1) inserting a benign trigger into the data to establish a controlled baseline, (2) applying Principal Component Analysis (PCA) to extract discriminative features and assess the separability of the data, (3) performing Gaussian Mixture Model (GMM) clustering to identify potentially malicious data samples based on their distribution in the PCA-transformed space, and (4) applying a targeted blurring technique to disrupt potential backdoor triggers. Together, these steps ensure that suspicious data is detected early and sanitized effectively, thereby minimizing the influence of backdoor triggers on the global model. Experimental evaluations on image-based datasets demonstrate that FL-PBM reduces attack success rates by up to 95% compared to baseline federated learning (FedAvg) and by 30 to 80% relative to state-of-the-art defenses (RDFL and LPSF). At the same time, it maintains over 90% clean model accuracy in most experiments, achieving better mitigation without degrading model performance.

Agentic vision-language models increasingly act through extended interactions, but most evaluations still focus on single-image, single-turn correctness. We introduce AMIGO (Agentic Multi-Image Grounding Oracle Benchmark), a long-horizon benchmark for hidden-target identification over galleries of visually similar images. In AMIGO, the oracle privately selects a target image, and the model must recover it by asking a sequence of attribute-focused Yes/No/Unsure questions under a strict protocol that penalizes invalid actions with Skip. This setting stresses (i) question selection under uncertainty, (ii) consistent constraint tracking across turns, and (iii) fine-grained discrimination as evidence accumulates. AMIGO also supports controlled oracle imperfections to probe robustness and verification behavior under inconsistent feedback. We instantiate AMIGO with Guess My Preferred Dress task and report metrics covering both outcomes and interaction quality, including identification success, evidence verification, efficiency, protocol compliance, noise tolerance, and trajectory-level diagnostics.

Large Language Models (LLMs) have demonstrated impressive capabilities in code generation. While an interactive feedback loop can improve performance, writing effective tests is a non-trivial task. Early multi-agent frameworks, such as AgentCoder, automated this process but relied on generated tests as absolute ground truth. This approach is fragile: incorrect code frequently passes faulty or trivial tests, while valid solutions are often degraded to satisfy incorrect assertions. Addressing this limitation, newer methods have largely abandoned test generation in favor of planning and reasoning based on examples. We argue, however, that generated tests remain a valuable signal if we model them as noisy sensors guided by bayesian updates. To this end, we introduce BACE (Bayesian Anchored Co-Evolution), a framework that reformulates synthesis as a Bayesian co-evolutionary process where code and test populations are evolved, guided by belief distributions that are reciprocally updated based on noisy interaction evidence. By anchoring this search on minimal public examples, BACE prevents the co-evolutionary drift typical of self-validating loops. Extensive evaluations on LiveCodeBench v6 (post-March 2025) reveal that BACE achieves superior performance across both proprietary models and open-weight small language models.

Federated Learning (FL) is witnessing wider adoption due to its ability to benefit from large amounts of scattered data while preserving privacy. However, despite its advantages, federated learning suffers from several setbacks that directly impact the accuracy, and the integrity of the global model it produces. One of these setbacks is the presence of malicious clients who actively try to harm the global model by injecting backdoor data into their local models while trying to evade detection. The objective of such clients is to trick the global model into making false predictions during inference, thereby compromising the integrity and trustworthiness of the global model on which honest stakeholders rely. To mitigate such mischievous behavior, we propose FedBBA (Federated Backdoor and Behavior Analysis). The proposed model aims to dampen the effect of such clients on the final accuracy, creating more resilient federated learning environments. We engineer our approach through the combination of (1) a reputation system to evaluate and track client behavior, (2) an incentive mechanism to reward honest participation and penalize malicious behavior, and (3) game theoretical models with projection pursuit analysis (PPA) to dynamically identify and minimize the impact of malicious clients on the global model. Extensive simulations on the German Traffic Sign Recognition Benchmark (GTSRB) and Belgium Traffic Sign Classification (BTSC) datasets demonstrate that FedBBA reduces the backdoor attack success rate to approximately 1.1%--11% across various attack scenarios, significantly outperforming state-of-the-art defenses like RDFL and RoPE, which yielded attack success rates between 23% and 76%, while maintaining high normal task accuracy (~95%--98%).

Can a safety gate permit unbounded beneficial self-modification while maintaining bounded cumulative risk? We formalize this question through dual conditions -- requiring sum delta_n < infinity (bounded risk) and sum TPR_n = infinity (unbounded utility) -- and establish a theory of their (in)compatibility. Classification impossibility (Theorem 1): For power-law risk schedules delta_n = O(n^{-p}) with p > 1, any classifier-based gate under overlapping safe/unsafe distributions satisfies TPR_n <= C_alpha * delta_n^beta via Holder's inequality, forcing sum TPR_n < infinity. This impossibility is exponent-optimal (Theorem 3). A second independent proof via the NP counting method (Theorem 4) yields a 13% tighter bound without Holder's inequality. Universal finite-horizon ceiling (Theorem 5): For any summable risk schedule, the exact maximum achievable classifier utility is U*(N, B) = N * TPR_NP(B/N), growing as exp(O(sqrt(log N))) -- subpolynomial. At N = 10^6 with budget B = 1.0, a classifier extracts at most U* ~ 87 versus a verifier's ~500,000. Verification escape (Theorem 2): A Lipschitz ball verifier achieves delta = 0 with TPR > 0, escaping the impossibility. Formal Lipschitz bounds for pre-LayerNorm transformers under LoRA enable LLM-scale verification. The separation is strict. We validate on GPT-2 (d_LoRA = 147,456): conditional delta = 0 with TPR = 0.352. Comprehensive empirical validation is in the companion paper [D2].

Multidimensional loop kernels often suffer from control overhead that can dominate execution time on parallel loop accelerators. Tightly Coupled Processor Arrays (TCPAs) offload loop control to a global controller (GC), but existing approaches still require hundreds of control signals. We propose a method to derive and aggressively reduce these control conditions from a polyhedral representation of the iteration space, achieving reductions of 15x to 45x in control signals across several benchmarks. We introduce a lightweight GC architecture that evaluates conditions as unions of polyhedra using bounded evaluation units, requiring hardware comparable to a single processing element. Control signals are distributed throughout the array with a minimal number of delay elements resulting in zero-overhead loop control. Our evaluation on PolyBench kernels shows that the entire control flow requires < 10 % of the total array resources.

Combining multiple audio features can improve the performance of music tagging, but common deep learning-based feature fusion methods often lack interpretability. To address this problem, we propose a Genetic Programming (GP) pipeline that automatically evolves composite features by mathematically combining base music features, thereby capturing synergistic interactions while preserving interpretability. This approach provides representational benefits similar to deep feature fusion without sacrificing interpretability. Experiments on the MTG-Jamendo and GTZAN datasets demonstrate consistent improvements compared to state-of-the-art systems across base feature sets at different abstraction levels. It should be noted that most of the performance gains are noticed within the first few hundred GP evaluations, indicating that effective feature combinations can be identified under modest search budgets. The top evolved expressions include linear, nonlinear, and conditional forms, with various low-complexity solutions at top performance aligned with parsimony pressure to prefer simpler expressions. Analyzing these composite features further reveals which interactions and transformations tend to be beneficial for tagging, offering insights that remain opaque in black-box deep models.

Psychological scale development has traditionally required extensive expert involvement, iterative revision, and large-scale pilot testing before psychometric evaluation can begin. The `AIGENIE` R package implements the AI-GENIE framework (Automatic Item Generation with Network-Integrated Evaluation), which integrates large language model (LLM) text generation with network psychometric methods to automate the early stages of this process. The package generates candidate item pools using LLMs, transforms them into high-dimensional embeddings, and applies a multi-step reduction pipeline -- Exploratory Graph Analysis (EGA), Unique Variable Analysis (UVA), and bootstrap EGA -- to produce structurally validated item pools entirely *in silico*. This tutorial introduces the package across six parts: installation and setup, understanding Application Programming Interfaces (APIs), text generation, item generation, the `AIGENIE` function, and the `GENIE` function. Two running examples illustrate the package's use: the Big Five personality model (a well-established construct) and AI Anxiety (an emerging construct). The package supports multiple LLM providers (OpenAI, Anthropic, Groq, HuggingFace, and local models), offers a fully offline mode with no external API calls, and provides the `GENIE()` function for researchers who wish to apply the psychometric reduction pipeline to existing item pools regardless of their origin. The `AIGENIE` package is freely available on R-universe at https://laralee.r-universe.dev/AIGENIE.

For any $Δ$, let $k_Δ$ be the maximum integer $k$ such that $(k+1)(k+2)\le Δ$. We give a distributed \LOCAL algorithm that, given an integer $k < k_Δ$, computes a valid $Δ-k$-coloring if one exists. The algorithm runs in $\tilde{O}(\log^4 \log n)$ rounds, which is within a polynomial factor of the $Ω(\log\log n)$ lower bound, which already applies to the case $k=0$. It is also best possible in the sense that if $k \ge k_Δ$, the problem requires $Ω(n/Δ)$ distributed rounds [Molloy, Reed, '14, Bamas, Esperet '19]. For $Δ$ at most polylogarithmic, the algorithm is an exponential improvement over the current state of the art of $O(\log^{49/12} n)$ rounds. When $Δ\ge (\log n)^{50}$, our algorithm achieves an even faster runtime of $O(\log^* n)$ rounds.

Pure Pursuit (PP) is a widely used path-tracking algorithm in autonomous vehicles due to its simplicity and real-time performance. However, its effectiveness is sensitive to the choice of lookahead distance: shorter values improve cornering but can cause instability on straights, while longer values improve smoothness but reduce accuracy in curves. We propose a hybrid control framework that integrates Proximal Policy Optimization (PPO) with the classical Pure Pursuit controller to adjust the lookahead distance dynamically during racing. The PPO agent maps vehicle speed and multi-horizon curvature features to an online lookahead command. It is trained using Stable-Baselines3 in the F1TENTH Gym simulator with a KL penalty and learning-rate decay for stability, then deployed in a ROS2 environment to guide the controller. Experiments in simulation compare the proposed method against both fixed-lookahead Pure Pursuit and an adaptive Pure Pursuit baseline. Additional real-car experiments compare the learned controller against a fixed-lookahead Pure Pursuit controller. Results show that the learned policy improves lap-time performance and repeated lap completion on unseen tracks, while also transferring zero-shot to hardware. The learned controller adapts the lookahead by increasing it on straights and reducing it in curves, demonstrating effectiveness in augmenting a classical controller by online adaptation of a single interpretable parameter. On unseen tracks, the proposed method achieved 33.16 s on Montreal and 46.05 s on Yas Marina, while tolerating more aggressive speed-profile scaling than the baselines and achieving the best lap times among the tested settings. Initial real-car experiments further support sim-to-real transfer on a 1:10-scale autonomous racing platform

Emerging deployments of Generative AI increasingly execute inference across decentralized and heterogeneous edge devices rather than on a single trusted server. In such environments, a single device failure or misbehavior can disrupt the entire inference process, making traditional best-effort peer-to-peer routing insufficient. Coordinating distributed generative inference therefore requires mechanisms that explicitly account for reliability, performance variability, and trust among participating peers. In this paper, we present G-TRAC, a trust-aware coordination framework that integrates algorithmic path selection with system-level protocol design to ensure robust distributed inference. First, we formulate the routing problem as a \textit{Risk-Bounded Shortest Path} computation and introduce a polynomial-time solution that combines trust-floor pruning with Dijkstra's search, achieving sub-millisecond median routing latency at practical edge scales, and remaining below 10 ms at larger scales. Second, to operationally support the routing logic in dynamic environments, the framework employs a \textit{Hybrid Trust Architecture} that maintains global reputation state at stable anchors while disseminating lightweight updates to edge peers via background synchronization. Experimental evaluation on a heterogeneous testbed of commodity devices demonstrates that G-TRAC significantly improves inference completion rates, effectively isolates unreliable peers, and sustains robust execution even under node failures and network partitions.

Reinforcement learning with verifiable rewards (RLVR) has substantially enhanced the reasoning capabilities of multimodal large language models (MLLMs). However, existing RLVR approaches typically rely on outcome-driven optimization that updates both perception and reasoning using a shared reward based solely on the final answer. This shared reward blurs credit assignment, frequently improving reasoning patterns while failing to reliably enhance the accuracy of upstream visual evidence extraction. To address this perception bottleneck, we introduce PRCO (Perception-Reasoning Coevolution), a dual-role RLVR framework with a shared policy. PRCO consists of two cooperative roles: an Observer that generates an evidence caption tailored to the question and a Solver that predicts the final answer based on this caption. Crucially, PRCO employs role-specific reward signals: the Solver is optimized using verifiable outcome rewards on the final answer, while the Observer receives a utility reward derived from the Solver's downstream success. Extensive experiments across eight challenging multimodal reasoning benchmarks demonstrate that PRCO yields consistent improvements across model scales by over 7 points on average accuracy compared to the base model, outperforming prior open-source RL-tuned baselines.

Generative AI has made text-guided inpainting a powerful image editing tool, but at the same time a growing challenge for media forensics. Existing benchmarks, including our text-guided inpainting forgery (TGIF) dataset, show that image forgery localization (IFL) methods can localize manipulations in spliced images but struggle not in fully regenerated (FR) images, while synthetic image detection (SID) methods can detect fully regenerated images but cannot perform localization. With new generative inpainting models emerging and the open problem of localization in FR images remaining, updated datasets and benchmarks are needed. We introduce TGIF2, an extended version of TGIF, that captures recent advances in text-guided inpainting and enables a deeper analysis of forensic robustness. TGIF2 augments the original dataset with edits generated by FLUX.1 models, as well as with random non-semantic masks. Using the TGIF2 dataset, we conduct a forensic evaluation spanning IFL and SID, including fine-tuning IFL methods on FR images and generative super-resolution attacks. Our experiments show that both IFL and SID methods degrade on FLUX.1 manipulations, highlighting limited generalization. Additionally, while fine-tuning improves localization on FR images, evaluation with random non-semantic masks reveals object bias. Furthermore, generative super-resolution significantly weakens forensic traces, demonstrating that common image enhancement operations can undermine current forensic pipelines. In summary, TGIF2 provides an updated dataset and benchmark, which enables new insights into the challenges posed by modern inpainting and AI-based image enhancements. TGIF2 is available at https://github.com/IDLabMedia/tgif-dataset.

Fixed representational capacity is a fundamental constraint in continual learning: practitioners must guess an appropriate model width before training, without knowing how many distinct concepts the data contains. We propose LACE (Loss-Adaptive Capacity Expansion), a simple online mechanism that expands a model's representational capacity during training by monitoring its own loss signal. When sustained loss deviation exceeds a threshold - indicating that the current capacity is insufficient for newly encountered data - LACE adds new dimensions to the projection layer and trains them jointly with existing parameters. Across synthetic and real-data experiments, LACE triggers expansions exclusively at domain boundaries (100% boundary precision, zero false positives), matches the accuracy of a large fixed-capacity model while starting from a fraction of its dimensions, and produces adapter dimensions that are collectively critical to performance (3% accuracy drop when all adapters removed). We further demonstrate unsupervised domain separation in GPT-2 activations via layer-wise clustering, showing a U-shaped separability curve across layers that motivates adaptive capacity allocation in deep networks. LACE requires no labels, no replay buffers, and no external controllers, making it suitable for on-device continual learning under resource constraints.

Multimodal Large Language Models (MLLMs) achieve stronger visual understanding by scaling input fidelity, yet the resulting visual token growth makes jointly sustaining high spatial resolution and long temporal context prohibitive. We argue that the bottleneck lies not in how post-encoding representations are compressed but in the volume of pixels the encoder receives, and address it with ResAdapt, an Input-side adaptation framework that learns how much visual budget each frame should receive before encoding. ResAdapt couples a lightweight Allocator with an unchanged MLLM backbone, so the backbone retains its native visual-token interface while receiving an operator-transformed input. We formulate allocation as a contextual bandit and train the Allocator with Cost-Aware Policy Optimization (CAPO), which converts sparse rollout feedback into a stable accuracy-cost learning signal. Across budget-controlled video QA, temporal grounding, and image reasoning tasks, ResAdapt improves low-budget operating points and often lies on or near the efficiency-accuracy frontier, with the clearest gains on reasoning-intensive benchmarks under aggressive compression. Notably, ResAdapt supports up to 16x more frames at the same visual budget while delivering over 15% performance gain. Code is available at https://github.com/Xnhyacinth/ResAdapt.

Large-scale image datasets frequently contain identifiable or sensitive content, raising privacy risks when training models that may memorize and leak such information. We present Unsafe2Safe, a fully automated pipeline that detects privacy-prone images and rewrites only their sensitive regions using multimodally guided diffusion editing. Unsafe2Safe operates in two stages. Stage 1 uses a vision-language model to (i) inspect images for privacy risks, (ii) generate paired private and public captions that respectively include and omit sensitive attributes, and (iii) prompt a large language model to produce structured, identity-neutral edit instructions conditioned on the public caption. Stage 2 employs instruction-driven diffusion editors to apply these dual textual prompts, producing privacy-safe images that preserve global structure and task-relevant semantics while neutralizing private content. To measure anonymization quality, we introduce a unified evaluation suite covering Quality, Cheating, Privacy, and Utility dimensions. Across MS-COCO, Caltech101, and MIT Indoor67, Unsafe2Safe reduces face similarity, text similarity, and demographic predictability by large margins, while maintaining downstream model accuracy comparable to training on raw data. Fine-tuning diffusion editors on our automatically generated triplets (private caption, public caption, edit instruction) further improves both privacy protection and semantic fidelity. Unsafe2Safe provides a scalable, principled solution for constructing large, privacy-safe datasets without sacrificing visual consistency or downstream utility.

The demand for Explainable AI (XAI) has triggered an explosion of methods, producing a landscape so fragmented that we now rely on surveys of surveys. Yet, fundamental challenges persist: conflicting metrics, failed sanity checks, and unresolved debates over robustness and fairness. The only consensus on how to achieve explainability is a lack of one. This has led many to point to the absence of a ground truth for defining ``the'' correct explanation as the main culprit. This position paper posits that the persistent discord in XAI arises not from an absent ground truth but from a ground truth that exists, albeit as an elusive and challenging target: the causal model that governs the relevant system. By reframing XAI queries about data, models, or decisions as causal inquiries, we prove the necessity and sufficiency of causal models for XAI. We contend that without this causal grounding, XAI remains unmoored. Ultimately, we encourage the community to converge around advanced concept and causal discovery to escape this entrenched uncertainty.

Reflective writing is known to support the development of students' metacognitive skills, yet learners often struggle to engage in deep reflection, limiting learning gains. Although large language models (LLMs) have been shown to improve writing skills, their use as conversational agents for reflective writing has produced mixed results and has largely focused on providing feedback on reflective texts, rather than support during planning and organizing. In this paper, inspired by the Cognitive Process Theory of writing (CPT), we propose the first application of LLMs to the planning and translation steps of reflective writing. We introduce Pensée, a tool to explore the effects of explicit AI support during these stages by scaffolding structured reflection planning using a conversational agent, and supporting translation by automatically extracting key concepts. We evaluate Pensée in a controlled between-subjects experiment (N=93), manipulating AI support across writing phases. Results show significantly greater reflection depth and structural quality when learners receive support during planning and translation stages of CPT, though these effects reduce in a delayed post-test. Analyses of learner behavior and perceptions further illustrate how CPT-aligned conversational support shapes reflection processes and learner experience, contributing empirical evidence for theory-driven uses of LLMs in AI-supported reflective writing.

Although actor-critic methods have been successful in practice, their theoretical analyses have several limitations. Specifically, existing theoretical work either sidesteps the exploration problem by making strong assumptions or analyzes impractical methods with complicated algorithmic modifications. Moreover, the actor-critic methods analyzed for linear MDPs often employ natural policy gradient and construct "implicit" policies without explicit parameterization. Such policies are computationally expensive to sample from, making the environment interactions inefficient. To that end, we focus on the finite-horizon linear MDPs and propose an optimistic actor-critic framework that uses parametric log-linear policies. In particular, we introduce a tractable $\textit{logit-matching}$ regression objective for the actor. For the critic, we use approximate Thompson sampling via Langevin Monte Carlo to obtain optimistic value estimates. We prove that the resulting algorithm achieves $\widetilde{\mathcal{O}}(ε^{-4})$ and $\widetilde{\mathcal{O}}(ε^{-2})$ sample complexity in the on-policy and off-policy setting, respectively. Our results match prior theoretical work in achieving the state-of-the-art sample complexity, while our algorithm is more aligned with practice.

Deep Neural Networks (DNNs) achieve strong performance in semantic segmentation for robotic perception but remain vulnerable to adversarial attacks, threatening safety-critical applications. While robustness has been studied for image classification, semantic segmentation in robotic contexts requires specialized architectures and detection strategies.

AI coding assistants are now widely used in software development. Software developers increasingly integrate AI-generated code into their codebases to improve productivity. Prior studies have shown that AI-generated code may contain code quality issues under controlled settings. However, we still know little about the real-world impact of AI-generated code on software quality and maintenance after it is introduced into production repositories. In other words, it remains unclear whether such issues are quickly fixed or persist and accumulate over time as technical debt. In this paper, we conduct a large-scale empirical study on the technical debt introduced by AI coding assistants in the wild. To achieve that, we built a dataset of 304,362 verified AI-authored commits from 6,275 GitHub repositories, covering five widely used AI coding assistants. For each commit, we run static analysis before and after the change to precisely attribute which code smells, bugs, and security issues the AI introduced. We then track each introduced issue from the introducing commit to the latest repository revision to study its lifecycle. Our results show that we identified 484,606 distinct issues, and that code smells are by far the most common type, accounting for 89.1% of all issues. We also find that more than 15% of commits from every AI coding assistant introduce at least one issue, although the rates vary across tools. More importantly, 24.2% of tracked AI-introduced issues still survive at the latest revision of the repository. These findings show that AI-generated code can introduce long-term maintenance costs into real software projects and highlight the need for stronger quality assurance in AI-assisted development.

We analyze the universal approximation constraints of narrow Residual Neural Networks (ResNets) both theoretically and numerically. For deep neural networks without input space augmentation, a central constraint is the inability to represent critical points of the input-output map. We prove that this has global consequences for target function approximations and show that the manifestation of this defect is typically a shift of the critical point to infinity, which we call the ``tunnel effect'' in the context of classification tasks. While ResNets offer greater expressivity than standard multilayer perceptrons (MLPs), their capability strongly depends on the signal ratio between the skip and residual channels. We establish quantitative approximation bounds for both the residual-dominant (close to MLP) and skip-dominant (close to neural ODE) regimes. These estimates depend explicitly on the channel ratio and uniform network weight bounds. Low-dimensional examples further provide a detailed analysis of the different ResNet regimes and how architecture-target incompatibility influences the approximation error.

Large language models (LLMs) can generate chains of thought (CoTs) that are not always causally responsible for their final outputs. When such a mismatch occurs, the CoT no longer faithfully reflects the decision-critical factors driving the model's behavior, leading to the reduced CoT monitorability problem. However, a comprehensive and fully open-source benchmark for studying CoT monitorability remains lacking. To address this gap, we propose MonitorBench, a systematic benchmark for evaluating CoT monitorability in LLMs. MonitorBench provides: (1) a diverse set of 1,514 test instances with carefully designed decision-critical factors across 19 tasks spanning 7 categories to characterize when CoTs can be used to monitor the factors driving LLM behavior; and (2) two stress-test settings to quantify the extent to which CoT monitorability can be degraded. Extensive experiments across multiple popular LLMs with varying capabilities show that CoT monitorability is higher when producing the final target response requires structural reasoning through the decision-critical factor. Closed-source LLMs generally show lower monitorability, and there exists a negative relationship between monitorability and model capability. Moreover, both open- and closed-source LLMs can intentionally reduce monitorability under stress-tests, with monitorability dropping by up to 30% in some tasks that do not require structural reasoning over the decision-critical factors. Beyond these empirical insights, MonitorBench provides a basis for further research on evaluating future LLMs, studying advanced stress-test monitorability techniques, and developing new monitoring approaches.

Autonomous systems that generate scientific hypotheses, conduct experiments, and draft manuscripts have recently emerged as a promising paradigm for accelerating discovery. However, existing AI Scientists remain largely domain-agnostic, limiting their applicability to clinical medicine, where research is required to be grounded in medical evidence with specialized data modalities. In this work, we introduce Medical AI Scientist, the first autonomous research framework tailored to clinical autonomous research. It enables clinically grounded ideation by transforming extensively surveyed literature into actionable evidence through clinician-engineer co-reasoning mechanism, which improves the traceability of generated research ideas. It further facilitates evidence-grounded manuscript drafting guided by structured medical compositional conventions and ethical policies. The framework operates under 3 research modes, namely paper-based reproduction, literature-inspired innovation, and task-driven exploration, each corresponding to a distinct level of automated scientific inquiry with progressively increasing autonomy. Comprehensive evaluations by both large language models and human experts demonstrate that the ideas generated by the Medical AI Scientist are of substantially higher quality than those produced by commercial LLMs across 171 cases, 19 clinical tasks, and 6 data modalities. Meanwhile, our system achieves strong alignment between the proposed method and its implementation, while also demonstrating significantly higher success rates in executable experiments. Double-blind evaluations by human experts and the Stanford Agentic Reviewer suggest that the generated manuscripts approach MICCAI-level quality, while consistently surpassing those from ISBI and BIBM. The proposed Medical AI Scientist highlights the potential of leveraging AI for autonomous scientific discovery in healthcare.

Despite the success of Vision-Language Models (VLMs), misleading charts remain a significant challenge due to their deceptive visual structures and distorted data representations. We present ChartCynics, an agentic dual-path framework designed to unmask visual deception via a "skeptical" reasoning paradigm. Unlike holistic models, ChartCynics decouples perception from verification: a Diagnostic Vision Path captures structural anomalies (e.g., inverted axes) through strategic ROI cropping, while an OCR-Driven Data Path ensures numerical grounding. To resolve cross-modal conflicts, we introduce an Agentic Summarizer optimized via a two-stage protocol: Oracle-Informed SFT for reasoning distillation and Deception-Aware GRPO for adversarial alignment. This pipeline effectively penalizes visual traps and enforces logical consistency. Evaluations on two benchmarks show that ChartCynics achieves 74.43% and 64.55% accuracy, providing an absolute performance boost of ~29% over the Qwen3-VL-8B backbone, outperforming state-of-the-art proprietary models. Our results demonstrate that specialized agentic workflows can grant smaller open-source models superior robustness, establishing a new foundation for trustworthy chart interpretation.

The discovery of anticancer therapeutics has traditionally treated organic small molecules and metal-based coordination complexes as separate chemical domains, limiting knowledge transfer despite their shared biological objectives. This disparity is particularly pronounced in available data, with extensive screening databases for organic compounds compared to only a few thousand characterized metal complexes. Here, we introduce ChemCLIP, a dual-encoder contrastive learning framework that bridges this organic-inorganic divide by learning unified representations based on shared anticancer activities rather than structural similarity. We compiled complementary datasets comprising 44,854 unique organic compounds and 5,164 unique metal complexes, standardized across 60 cancer cell lines. By training parallel encoders with activity-aware hard negative mining, we mapped structurally distinct compounds into a shared 256-dimensional embedding space where biologically similar compounds cluster together regardless of chemical class. We systematically evaluated four molecular encoding strategies: Morgan fingerprints, ChemBERTa, MolFormer, and Chemprop, through quantitative alignment metrics, embedding visualizations, and downstream classification tasks. Morgan fingerprints achieved superior performance with an average alignment ratio of 0.899 and downstream classification AUCs of 0.859 (inorganic) and 0.817 (organic). This work establishes contrastive learning as an effective strategy for unifying disparate chemical domains and provides empirical guidance for encoder selection in multi-modal chemistry applications, with implications extending beyond anticancer drug discovery to any scenario requiring cross-domain chemical knowledge transfer.

Offline multi-agent reinforcement learning (MARL) faces a critical challenge: the joint action space grows exponentially with the number of agents, making dataset coverage exponentially sparse and out-of-distribution (OOD) joint actions unavoidable. Partial Action Replacement (PAR) mitigates this by anchoring a subset of agents to dataset actions, but existing approach relies on enumerating multiple subset configurations at high computational cost and cannot adapt to varying states. We introduce PLCQL, a framework that formulates PAR subset selection as a contextual bandit problem and learns a state-dependent PAR policy using Proximal Policy Optimisation with an uncertainty-weighted reward. This adaptive policy dynamically determines how many agents to replace at each update step, balancing policy improvement against conservative value estimation. We prove a value-error bound showing that the estimation error scales linearly with the expected number of deviating agents. Compared with the previous PAR-based method SPaCQL, PLCQL reduces the number of per-iteration Q-function evaluations from n to 1, significantly improving computational efficiency. Empirically, PLCQL achieves the highest normalised scores on 66% of tasks across MPE, MaMuJoCo, and SMAC benchmarks, outperforming SPaCQL on 84% of tasks while substantially reducing computational cost.

Denoising models such as Diffusion or Flow Matching have recently advanced generative modeling for discrete structures, yet most approaches either operate directly in the discrete state space, causing abrupt state changes. We introduce simplex denoising, a simple yet effective generative framework that operates on the probability simplex. The key idea is a non-Markovian noising scheme in which, for a given clean data point, noisy representations at different times are conditionally independent. While preserving the theoretical guarantees of denoising-based generative models, our method removes unnecessary constraints, thereby improving performance and simplifying the formulation. Empirically, \emph{unrestrained simplex denoising} surpasses strong discrete diffusion and flow-matching baselines across synthetic and real-world graph benchmarks. These results highlight the probability simplex as an effective framework for discrete generative modeling.

The increasing agentic capabilities of Large Language Models (LLMs) have enabled their deployment in real-world applications, such as cloud services, where customer-assistant interactions exhibit high technical complexity and long-horizon dependencies, making robustness and resolution efficiency critical for customer satisfaction. However, existing benchmarks for LLM-based agents largely rely on synthetic environments that fail to capture the diversity and unpredictability of authentic customer inputs, often ignoring the resolution efficiency essential for real-world deployment. To bridge this gap, we introduce CirrusBench, a novel evaluation framework distinguished by its foundation in real-world data from authentic cloud service tickets. CirrusBench preserves the intricate multi-turn logical chains and realistic tool dependencies inherent to technical service environments. Moving beyond execution correctness, we introduce novel Customer-Centric metrics to define agent success, quantifying service quality through metrics such as the Normalized Efficiency Index and Multi-Turn Latency to explicitly measure resolution efficiency. Experiments utilizing our framework reveal that while state-of-the-art models demonstrate strong reasoning capabilities, they frequently struggle in complex, realistic multi-turn tasks and fail to meet the high-efficiency standards required for customer service, highlighting critical directions for the future development of LLM-based agents in practical technical service applications. CirrusBench evaluation framework is released at: https://github.com/CirrusAI

The growing deployment of small Unmanned Aerial Systems (sUASs) in low-altitude airspaces has increased the need for reliable tactical deconfliction under safety-critical constraints. Tactical deconfliction involves short-horizon decision-making in dense, partially observable, and heterogeneous multi-agent environments, where both cooperative separation assurance and operational efficiency must be maintained. While Large Language Models (LLMs) exhibit strong reasoning capabilities, their direct application to air traffic control remains limited by insufficient domain grounding and unpredictable output inconsistency. This paper investigates LLMs as decision-makers in cooperative multi-agent tactical deconfliction using fine-tuning strategies that align model outputs to human operator heuristics. We propose a simulation-to-language data generation pipeline based on the BlueSky air traffic simulator that produces rule-consistent deconfliction datasets reflecting established safety practices. A pretrained Qwen-Math-7B model is fine-tuned using two parameter-efficient strategies: supervised fine-tuning with Low-Rank Adaptation (LoRA) and preference-based fine-tuning combining LoRA with Group-Relative Policy Optimization (GRPO). Experimental results on validation datasets and closed-loop simulations demonstrate that supervised LoRA fine-tuning substantially improves decision accuracy, consistency, and separation performance compared to the pretrained LLM, with significant reductions in near mid-air collisions. GRPO provides additional coordination benefits but exhibits reduced robustness when interacting with heterogeneous agent policies.

We present a first comparative pilot study of three t-norm operators -- Lukasiewicz (T_L), Product (T_P), and Gödel (T_G) - as logical conjunction mechanisms in a neuro-symbolic reasoning system for EU AI Act compliance classification. Using the LGGT+ (Logic-Guided Graph Transformers Plus) engine and a benchmark of 1035 annotated AI system descriptions spanning four risk categories (prohibited, high_risk, limited_risk, minimal_risk), we evaluate classification accuracy, false positive and false negative rates, and operator behaviour on ambiguous cases. At n=1035, all three operators differ significantly (McNemar p<0.001). T_G achieves highest accuracy (84.5%) and best borderline recall (85%), but introduces 8 false positives (0.8%) via min-semantics over-classification. T_L and T_P maintain zero false positives, with T_P outperforming T_L (81.2% vs. 78.5%). Our principal findings are: (1) operator choice is secondary to rule base completeness; (2) T_L and T_P maintain zero false positives but miss borderline cases; (3) T_G's min-semantics achieves higher recall at cost of 0.8% false positive rate; (4) a mixed-semantics classifier is the productive next step. We release the LGGT+ core engine (201/201 tests passing) and benchmark dataset (n=1035) under Apache 2.0.

Large pre-trained vision-language models like CLIP have transformed computer vision by aligning images and text in a shared feature space, enabling robust zero-shot transfer via prompting. Soft-prompting, such as Context Optimization (CoOp), effectively adapts these models for downstream recognition tasks by learning a set of context vectors. However, CoOp lacks explicit mechanisms for handling domain shifts across unseen distributions. To address this, we propose Domain-invariant Context Optimization (DiCoOp), an extension of CoOp optimized for domain generalization. By employing an adversarial training approach, DiCoOp forces the model to learn domain-invariant prompts while preserving discriminative power for classification. Experimental results show that DiCoOp consistently surpasses CoOp in domain generalization tasks across diverse visual domains.

Visual document understanding typically requires separate retrieval and generation models, doubling memory and system complexity. We present Hydra, a dual-head approach that provides both ColBERT-style late-interaction retrieval and autoregressive generation from a single vision-language model (VLM). A single LoRA adapter, trained only for retrieval, is toggled at inference: enabling it produces multi-vector embeddings; disabling it recovers the base model's generation quality -- byte-identical outputs in 100% of 10,500 greedy and stochastic samples, with max delta-ANLS = 0.0044 across 15,301 samples on four VQA benchmarks (three informative; ChartQA is near-zero for both models under greedy decoding) when compared against an independent base-model pipeline. We identify three engineering requirements (attention-mode restoration, lm_head preservation, KV-cache-aware decoding) whose omission silently breaks generation despite correct weight recovery. On ViDoRe V1, Hydra (4B) is within 1 percentage point of a controlled single-head baseline in a single training run, with higher aggregate scores on V2 and V3 that are concentrated on a subset of tasks; multi-seed experiments are needed to confirm these trends. The single-model design reduces peak GPU memory by 41%, though adapter switching introduces throughput overhead under concurrent serving loads. An ablation shows that GritLM-style joint training provides no benefit within the LoRA-based (r=16) training regime. A proof-of-concept extension to Qwen2.5-Omni-3B demonstrates that the mechanism generalizes to audio retrieval and video embedding, with speech generation.

Instructional alignment, the match between intended cognition and enacted activity, is central to effective instruction but hard to operationalize at scale. We examine alignment in cybersecurity simulations using multimodal traces from 23 teams (76 students) across five exercise sessions. Study 1 codes objectives and team emails with Bloom's taxonomy and models the completion of key exercise tasks with generalized linear mixed models. Alignment, defined as the discrepancy between required and enacted Bloom levels, predicts success, whereas the Bloom category alone does not predict success once discrepancy is considered. Study 2 compares predictive feature families using grouped cross-validation and l1-regularized logistic regression. Text embeddings and log features outperform Bloom-only models (AUC~0.74 and 0.71 vs. 0.55), and their combination performs best (Test AUC~0.80), with Bloom frequencies adding little. Overall, the work offers a measure of alignment for simulations and shows that multimodal traces best forecast performance, while alignment provides interpretable diagnostic insight.

Personalized computer-use agents are rapidly moving from expert communities into mainstream use. Unlike conventional chatbots, these systems can install skills, invoke tools, access private resources, and modify local environments on users' behalf. Yet users often do not know what authority they have delegated, what the agent actually did during task execution, or whether the system has been safely removed afterward. We investigate this gap as a combined problem of risk understanding and post-hoc auditability, using OpenClaw as a motivating case. We first build a multi-source corpus of the OpenClaw ecosystem, including incidents, advisories, malicious-skill reports, news coverage, tutorials, and social-media narratives. We then conduct an interview study to examine how users and practitioners understand skills, autonomy, privilege, persistence, and uninstallation. Our findings suggest that participants often recognized these systems as risky in the abstract, but lacked concrete mental models of what skills can do, what resources agents can access, and what changes may remain after execution or removal. Motivated by these findings, we propose AgentTrace, a traceability framework and prototype interface for visualizing agent actions, touched resources, permission history, provenance, and persistent side effects. A scenario-based evaluation suggests that traceability-oriented interfaces can improve understanding of agent behavior, support anomaly detection, and foster more calibrated trust.

Every four years, the PISA test is administered by the OECD to test the knowledge of teenage students worldwide and allow for comparisons of educational systems. However, having to avoid language differences and annotator bias makes the grading of student answers challenging. For these reasons, it would be interesting to consider methods of automatic student answer grading. To train some of these methods, which require machine learning, or to compute parameters or select hyperparameters for those that do not, a large amount of domain-specific data is needed. In this work, we explore a small number of methods for creating a large-scale training dataset using only a relatively small confidential dataset as a reference, leveraging a set of very simple derived text formats to preserve confidentiality. Using the proposed methods, we successfully created three surrogate datasets that are, at the very least, superficially more similar to the reference dataset than a straightforward result of prompt-based generation. Early experiments suggest one of these approaches might also lead to improved training of automatic answer grading models.

Agentic knowledge graph question answering (KGQA) requires an agent to iteratively interact with knowledge graphs (KGs), posing challenges in both training data scarcity and reasoning generalization. Specifically, existing approaches often restrict agent exploration: prompting-based methods lack autonomous navigation training, while current training pipelines usually confine reasoning to predefined trajectories. To this end, this paper proposes \textit{GraphWalker}, a novel agentic KGQA framework that addresses these challenges through \textit{Automated Trajectory Synthesis} and \textit{Stage-wise Fine-tuning}. GraphWalker adopts a two-stage SFT training paradigm: First, the agent is trained on structurally diverse trajectories synthesized from constrained random-walk paths, establishing a broad exploration prior over the KG; Second, the agent is further fine-tuned on a small set of expert trajectories to develop reflection and error recovery capabilities. Extensive experiments demonstrate that our stage-wise SFT paradigm unlocks a higher performance ceiling for a lightweight reinforcement learning (RL) stage, enabling GraphWalker to achieve state-of-the-art performance on CWQ and WebQSP. Additional results on GrailQA and our constructed GraphWalkerBench confirm that GraphWalker enhances generalization to out-of-distribution reasoning paths. The code is publicly available at https://github.com/XuShuwenn/GraphWalker

Low left ventricular ejection fraction (LEF) frequently remains undetected until progression to symptomatic heart failure, underscoring the need for scalable screening strategies. Although artificial intelligence-enabled electrocardiography (AI-ECG) has shown promise, existing approaches rely solely on end-to-end black-box models with limited interpretability or on tabular systems dependent on commercial ECG measurement algorithms with suboptimal performance. We introduced ECG-based Predictor-Driven LEF (ECGPD-LEF), a structured framework that integrates foundation model-derived diagnostic probabilities with interpretable modeling for detecting LEF from ECG. Trained on the benchmark EchoNext dataset comprising 72,475 ECG-echocardiogram pairs and evaluated in predefined independent internal (n=5,442) and external (n=16,017) cohorts, our framework achieved robust discrimination for moderate LEF (internal AUROC 88.4%, F1 64.5%; external AUROC 86.8%, F1 53.6%), consistently outperforming the official end-to-end baseline provided with the benchmark across demographic and clinical subgroups. Interpretability analyses identified high-impact predictors, including normal ECG, incomplete left bundle branch block, and subendocardial injury in anterolateral leads, driving LEF risk estimation. Notably, these predictors independently enabled zero-shot-like inference without task-specific retraining (internal AUROC 75.3-81.0%; external AUROC 71.6-78.6%), indicating that ventricular dysfunction is intrinsically encoded within structured diagnostic probability representations. This framework reconciles predictive performance with mechanistic transparency, supporting scalable enhancement through additional predictors and seamless integration with existing AI-ECG systems.

We present LAD, a real-time language--action planner with an interruptible architecture that produces a motion plan in a single forward pass (~20 Hz) or generates textual reasoning alongside a motion plan (~10 Hz). LAD is fast enough for real-time closed-loop deployment, achieving ~3x lower latency than prior driving language models while setting a new learning-based state of the art on nuPlan Test14-Hard and InterPlan. We also introduce RAD, a rule-based planner designed to address structural limitations of PDM-Closed. RAD achieves state-of-the-art performance among rule-based planners on nuPlan Test14-Hard and InterPlan. Finally, we show that combining RAD and LAD enables hybrid planning that captures the strengths of both approaches. This hybrid system demonstrates that rules and learning provide complementary capabilities: rules support reliable maneuvering, while language enables adaptive and explainable decision-making.

Scientific writing is an iterative process that generates rich revision traces, yet publicly available resources typically expose only final or near-final versions of papers. This limits empirical study of revision behaviour and evaluation of large language models (LLMs) for scientific writing. We introduce EarlySciRev, a dataset of early-stage scientific text revisions automatically extracted from arXiv LaTeX source files. Our key observation is that commented-out text in LaTeX often preserves discarded or alternative formulations written by the authors themselves. By aligning commented segments with nearby final text, we extract paragraph-level candidate revision pairs and apply LLM-based filtering to retain genuine revisions. Starting from 1.28M candidate pairs, our pipeline yields 578k validated revision pairs, grounded in authentic early drafting traces. We additionally provide a human-annotated benchmark for revision detection. EarlySciRev complements existing resources focused on late-stage revisions or synthetic rewrites and supports research on scientific writing dynamics, revision modelling, and LLM-assisted editing.

WikiKG90Mv2 in NeurIPS 2022 is a large encyclopedic knowledge graph. Embedding knowledge graphs into continuous vector spaces is important for many practical applications, such as knowledge acquisition, question answering, and recommendation systems. Compared to existing knowledge graphs, WikiKG90Mv2 is a large scale knowledge graph, which is composed of more than 90 millions of entities. Both efficiency and accuracy should be considered when building graph embedding models for knowledge graph at scale. To this end, we follow the retrieve then re-rank pipeline, and make novel modifications in both retrieval and re-ranking stage. Specifically, we propose a priority infilling retrieval model to obtain candidates that are structurally and semantically similar. Then we propose an ensemble based re-ranking model with neighbor enhanced representations to produce final link prediction results among retrieved candidates. Experimental results show that our proposed method outperforms existing baseline methods and improves MRR of validation set from 0.2342 to 0.2839.

Classical AI scaling laws, especially for pre-training, describe how training loss decreases with compute in a power-law form. Their effectiveness has a basic and very practical sense: they make progress predictable, albeit at a declining rate. Yet their effectiveness is also unreasonable in two further senses. First, these laws are largely empirical and observational, but they appear repeatedly across model families and increasingly across training-adjacent regimes. Second, despite the diminishing returns they predict, progress in practice has often continued through rapidly improving efficiency, visible for example in falling cost per token. This paper argues that both features arise from the same source: scaling laws are unusually effective because they abstract away from many realization details. The compute variable is best understood as logical compute, an implementation-agnostic notion of model-side work, while the practical burden of scaling depends on how efficiently real resources are converted into that compute. This abstraction helps explain both why the laws travel so well across settings and why they give rise to a persistent efficiency game in hardware, algorithms, and systems. Once efficiency is made explicit, the main practical question becomes how many efficiency doublings are required to keep scaling productive despite diminishing returns. Under that view, diminishing returns are not only a geometric flattening of the loss curve, but also rising pressure for cost reduction, system-level innovation, and the breakthroughs needed to sustain Moore-like efficiency doublings.

Intrusion Detection System (IDS) is often calibrated to known attacks and generalizes poorly to unknown threats. This paper proposes GMA-SAWGAN-GP, a novel generative augmentation framework built on a Self-Attention-enhanced Wasserstein GAN with Gradient Penalty (WGAN-GP). The generator employs Gumbel-Softmax regularization to model discrete fields, while a Multilayer Perceptron (MLP)-based AutoEncoder acts as a manifold regularizer. A lightweight gating network adaptively balances adversarial and reconstruction losses via entropy regularization, improving stability and mitigating mode collapse. The self-attention mechanism enables the generator to capture both short- and long-range dependencies among features within each record while preserving categorical semantics through Gumbel-Softmax heads. Extensive experiments on NSL-KDD, UNSW-NB15, and CICIDS2017 using five representative IDS models demonstrate that GMA-SAWGAN-GP significantly improves detection performance on known attacks and enhances generalization to unknown attacks. Leave-One-Attack-type-Out (LOAO) evaluations using Area Under the Receiver Operating Characteristic (AUROC) and True Positive Rate at a 5 percent False Positive Rate confirm that IDS models trained on augmented datasets achieve higher robustness under unseen attack scenarios. Ablation studies validate the contribution of each component to performance gains. Compared with baseline models, the proposed framework improves binary classification accuracy by an average of 5.3 percent and multi-classification accuracy by 2.2 percent, while AUROC and True Positive Rate at a 5 percent False Positive Rate for unknown attacks increase by 3.9 percent and 4.8 percent, respectively, across the three datasets. Overall, GMA-SAWGAN-GP provides an effective approach to generative augmentation for mixed-type network traffic, improving IDS accuracy and resilience.

We consider the question of how to employ next-token prediction algorithms in adversarial online decision-making environments. Specifically, if we train a next-token prediction model on a distribution $\mathcal{D}$ over sequences of opponent actions, when is it the case that the induced online decision-making algorithm (by approximately best responding to the model's predictions) has low adversarial regret (i.e., when is $\mathcal{D}$ a \emph{low-regret distribution})? For unbounded context windows (where the prediction made by the model can depend on all the actions taken by the adversary thus far), we show that although not every distribution $\mathcal{D}$ is a low-regret distribution, every distribution $\mathcal{D}$ is exponentially close (in TV distance) to one low-regret distribution, and hence sublinear regret can always be achieved at negligible cost to the accuracy of the original next-token prediction model. In contrast to this, for bounded context windows (where the prediction made by the model can depend only on the past $w$ actions taken by the adversary, as may be the case in modern transformer architectures), we show that there are some distributions $\mathcal{D}$ of opponent play that are $Θ(1)$-far from any low-regret distribution $\mathcal{D'}$ (even when $w = Ω(T)$ and such distributions exist). Finally, we complement these results by showing that the unbounded context robustification procedure can be implemented by layers of a standard transformer architecture, and provide empirical evidence that transformer models can be efficiently trained to represent these new low-regret distributions.

Accurate MRI-to-CT synthesis could enable MR-only pelvic workflows by providing CT-like images with bone details while avoiding additional ionizing radiation. In this work, we investigate recently proposed drifting models for synthesizing pelvis CT images from MRI and benchmark them against convolutional neural networks (UNet, VAE), a generative adversarial network (WGAN-GP), a physics-inspired probabilistic model (PPFM), and diffusion-based methods (FastDDPM, DDIM, DDPM). Experiments are performed on two complementary datasets: Gold Atlas Male Pelvis and the SynthRAD2023 pelvis subset. Image fidelity and structural consistency are evaluated with SSIM, PSNR, and RMSE, complemented by qualitative assessment of anatomically critical regions such as cortical bone and pelvic soft-tissue interfaces. Across both datasets, the proposed drifting model achieves high SSIM and PSNR and low RMSE, surpassing strong diffusion baselines and conventional CNN-, VAE-, GAN-, and PPFM-based methods. Visual inspection shows sharper cortical bone edges, improved depiction of sacral and femoral head geometry, and reduced artifacts or over-smoothing, particularly at bone-air-soft tissue boundaries. Moreover, the drifting model attains these gains with one-step inference and inference times on the order of milliseconds, yielding a more favorable accuracy-efficiency trade-off than iterative diffusion sampling while remaining competitive in image quality. These findings suggest that drifting models are a promising direction for fast, high-quality pelvic synthetic CT generation from MRI and warrant further investigation for downstream applications such as MRI-only radiotherapy planning and PET/MR attenuation correction.

Large language models (LLMs) remain unreliable for high-stakes claim verification due to hallucinations and shallow reasoning. While retrieval-augmented generation (RAG) and multi-agent debate (MAD) address this, they are limited by one-pass retrieval and unstructured debate dynamics. We propose a courtroom-style multi-agent framework, PROClaim, that reformulates verification as a structured, adversarial deliberation. Our approach integrates specialized roles (e.g., Plaintiff, Defense, Judge) with Progressive RAG (P-RAG) to dynamically expand and refine the evidence pool during the debate. Furthermore, we employ evidence negotiation, self-reflection, and heterogeneous multi-judge aggregation to enforce calibration, robustness, and diversity. In zero-shot evaluations on the Check-COVID benchmark, PROClaim achieves 81.7% accuracy, outperforming standard multi-agent debate by 10.0 percentage points, with P-RAG driving the primary performance gains (+7.5 pp). We ultimately demonstrate that structural deliberation and model heterogeneity effectively mitigate systematic biases, providing a robust foundation for reliable claim verification. Our code and data are publicly available at https://github.com/mnc13/PROClaim.

Recommendation systems have become central gatekeepers of online information, shaping user behaviour across a wide range of activities. In response, users increasingly organize and coordinate to steer algorithmic outcomes toward diverse goals, such as promoting relevant content or limiting harmful material, relying on platform affordances -- such as likes, reviews, or ratings. While these mechanisms can serve beneficial purposes, they can also be leveraged for adversarial manipulation, particularly in systems where such feedback directly informs safety guarantees. In this paper, we study this vulnerability in recently proposed risk-controlling recommender systems, which use binary user feedback (e.g., "Not Interested") to provably limit exposure to unwanted content via conformal risk control. We empirically demonstrate that their reliance on aggregate feedback signals makes them inherently susceptible to coordinated adversarial user behaviour. Using data from a large-scale online video-sharing platform, we show that a small coordinated group (comprising only 1% of the user population) can induce up to a 20% degradation in nDCG for non-adversarial users by exploiting the affordances provided by risk-controlling recommender systems. We evaluate simple, realistic attack strategies that require little to no knowledge of the underlying recommendation algorithm and find that, while coordinated users can significantly harm overall recommendation quality, they cannot selectively suppress specific content groups through reporting alone. Finally, we propose a mitigation strategy that shifts guarantees from the group level to the user level, showing empirically how it can reduce the impact of adversarial coordinated behaviour while ensuring personalized safety for individuals.

The connoisseurship of antique Chinese porcelain demands extensive historical expertise, material understanding, and aesthetic sensitivity, making it difficult for non-specialists to engage. To democratize cultural-heritage understanding and assist expert connoisseurship, we introduce CiQi-Agent -- a domain-specific Porcelain Connoisseurship Agent for intelligent analysis of antique Chinese porcelain. CiQi-Agent supports multi-image porcelain inputs and enables vision tool invocation and multimodal retrieval-augmented generation, performing fine-grained connoisseurship analysis across six attributes: dynasty, reign period, kiln site, glaze color, decorative motif, and vessel shape. Beyond attribute classification, it captures subtle visual details, retrieves relevant domain knowledge, and integrates visual and textual evidence to produce coherent, explainable connoisseurship descriptions. To achieve this capability, we construct a large-scale, expert-annotated dataset CiQi-VQA, comprising 29,596 porcelain specimens, 51,553 images, and 557,940 visual question--answering pairs, and further establish a comprehensive benchmark CiQi-Bench aligned with the previously mentioned six attributes. CiQi-Agent is trained through supervised fine-tuning, reinforcement learning, and a tool-augmented reasoning framework that integrates two categories of tools: a vision tool and multimodal retrieval tools. Experimental results show that CiQi-Agent (7B) outperforms all competitive open- and closed-source models across all six attributes on CiQi-Bench, achieving on average 12.2\% higher accuracy than GPT-5. The model and dataset have been released and are publicly available at https://huggingface.co/datasets/SII-Monument-Valley/CiQi-VQA.

Diffusion models achieve state-of-the-art generative performance but are fundamentally bottlenecked by their slow, iterative sampling process. While diffusion distillation techniques enable high-fidelity, few-step generation, traditional objectives often restrict the student's performance by anchoring it solely to the teacher. Recent approaches have attempted to break this ceiling by integrating Reinforcement Learning (RL), typically through a simple summation of distillation and RL objectives. In this work, we propose a novel paradigm by re-conceptualizing distribution matching as a reward, denoted as $R_\text{dm}$. This unified perspective bridges the algorithmic gap between Diffusion Matching Distillation (DMD) and RL, providing several primary benefits. (1) Enhanced Optimization Stability: We introduce Group Normalized Distribution Matching (GNDM), which adapts standard RL group normalization to stabilize $R_\text{dm}$ estimation. By leveraging group-mean statistics, GNDM establishes a more robust and effective optimization direction. (2) Seamless Reward Integration: Our reward-centric formulation inherently supports adaptive weighting mechanisms, allowing for the fluid combination of DMD with external reward models. (3) Improved Sampling Efficiency: By aligning with RL principles, the framework readily incorporates Importance Sampling (IS), leading to a significant boost in sampling efficiency. Extensive experiments demonstrate that GNDM outperforms vanilla DMD, reducing the FID by 1.87. Furthermore, our multi-reward variant, GNDMR, surpasses existing baselines by striking an optimal balance between aesthetic quality and fidelity, achieving a peak HPS of 30.37 and a low FID-SD of 12.21. Ultimately, $R_\text{dm}$ provides a flexible, stable, and efficient framework for real-time, high-fidelity synthesis. Codes are coming soon.

Token-level sparse attention mechanisms, exemplified by DeepSeek Sparse Attention (DSA), achieve fine-grained key selection by scoring every historical key for each query through a lightweight indexer, then computing attention only on the selected subset. While the downstream sparse attention itself scales favorably, the indexer must still scan the entire prefix for every query, introducing an per-layer bottleneck that grows prohibitively with context length. We propose HISA (Hierarchical Indexed Sparse Attention), a plug-and-play replacement for the indexer that rewrites the search path from a flat token scan into a two-stage hierarchical procedure: (1) a block-level coarse filtering stage that scores pooled block representations to discard irrelevant regions, followed by (2) a token-level refinement stage that applies the original indexer exclusively within the retained candidate blocks. HISA preserves the identical token-level top-sparse pattern consumed by the downstream Sparse MLA operator and requires no additional training. On kernel-level benchmarks, HISA achieves up to speedup at 64K context. On Needle-in-a-Haystack and LongBench, we directly replace the indexer in DeepSeek-V3.2 and GLM-5 with our HISA indexer, without any finetuning. HISA closely matches the original DSA in quality, while substantially outperforming block-sparse baselines.

In federated healthcare systems, Federated Class-Incremental Learning (FCIL) has emerged as a key paradigm, enabling continuous adaptive model learning among distributed clients while safeguarding data privacy. However, in practical applications, data across agent nodes within the distributed framework often exhibits non-independent and identically distributed (non-IID) characteristics, rendering traditional continual learning methods inapplicable. To address these challenges, this paper covers more comprehensive incremental task scenarios and proposes a dynamic memory allocation strategy for exemplar storage based on the data replay mechanism. This strategy fully taps into the inherent potential of data heterogeneity, while taking into account the performance fairness of all participating clients, thereby establishing a balanced and adaptive solution to mitigate catastrophic forgetting. Unlike the fixed allocation of client exemplar memory, the proposed scheme emphasizes the rational allocation of limited storage resources among clients to improve model performance. Furthermore, extensive experiments are conducted on three medical image datasets, and the results demonstrate significant performance improvements compared to existing baseline models.

Test flakiness is a common problem in industry, which hinders the reliability of automated build and testing workflows. Most existing research on test flakiness has primarily focused on unit and small-scale integration tests. In contrast, flakiness in system-level testing such as REST APIs are comparatively under-explored. A large body of literature has been dedicated to the topic of fuzzing REST APIs, whereas relatively little attention has been paid to detecting and possibly mitigating negative effects of flakiness in this context. To fill this major gap, in this paper, we study the flakiness of tests generated by one of the popularly applied REST API fuzzer in the literature, namely EvoMaster, conduct empirical studies with a corpus of 36 REST APIs to understand flakiness of REST APIs. Based on the results of the empirical studies, we categorize and analyze flakiness sources by inspecting near 3000 failing tests. Based on the understanding, we propose FlakyCatch to detect and mitigate flakiness in REST APIs and empirically evaluate its performance. Results show that FlakyCatch is effective in detecting and handling flakiness in tests generated by white-box and black-box fuzzers.

We propose Yau's Affine Normal Descent (YAND), a geometric framework for smooth unconstrained optimization in which search directions are defined by the equi-affine normal of level-set hypersurfaces. The resulting directions are invariant under volume-preserving affine transformations and intrinsically adapt to anisotropic curvature. Using the analytic representation of the affine normal from affine differential geometry, we establish its equivalence with the classical slice-centroid construction under convexity. For strictly convex quadratic objectives, affine-normal directions are collinear with Newton directions, implying one-step convergence under exact line search. For general smooth (possibly nonconvex) objectives, we characterize precisely when affine-normal directions yield strict descent and develop a line-search-based YAND. We establish global convergence under standard smoothness assumptions, linear convergence under strong convexity and Polyak-Lojasiewicz conditions, and quadratic local convergence near nondegenerate minimizers. We further show that affine-normal directions are robust under affine scalings, remaining insensitive to arbitrarily ill-conditioned transformations. Numerical experiments illustrate the geometric behavior of the method and its robustness under strong anisotropic scaling.

Current Retrieval-Augmented Generation (RAG) systems predominantly rely on relevance-based dense retrieval, sequentially fetching documents to maximize semantic similarity with the query. However, in knowledge-intensive and real-world scenarios characterized by conflicting evidence or fundamental query ambiguity, relevance alone is insufficient for resolving epistemic uncertainty. We introduce Entropic Claim Resolution (ECR), a novel inference-time algorithm that reframes RAG reasoning as entropy minimization over competing semantic answer hypotheses. Unlike action-driven agentic frameworks (e.g., ReAct) or fixed-pipeline RAG architectures, ECR sequentially selects atomic evidence claims by maximizing Expected Entropy Reduction (EER), a decision-theoretic criterion for the value of information. The process dynamically terminates when the system reaches a mathematically defined state of epistemic sufficiency (H <= epsilon, subject to epistemic coherence). We integrate ECR into a production-grade multi-strategy retrieval pipeline (CSGR++) and analyze its theoretical properties. Our framework provides a rigorous foundation for uncertainty-aware evidence selection, shifting the paradigm from retrieving what is most relevant to retrieving what is most discriminative.

Although 3D Gaussian Splatting (3DGS) enables high-fidelity real-time rendering, its prohibitive storage overhead severely hinders practical deployment. Recent anchor-based 3DGS compression schemes reduce gaussina redundancy through ome advanced context models. However, overlook explicit geometric dependencies, leading to structural degradation and suboptimal rate-distortion performance. In this paper, we propose LG-HCC, a geometry-aware 3DGS compression framework that incorporates inter-anchor geometric correlations into anchor pruning and entropy coding for compact representation. Specifically, we introduce an Neighborhood-Aware Anchor Pruning (NAAP) strategy, which evaluates anchor importance via weighted neighborhood feature aggregation and merges redundant anchors into salient neighbors, yielding a compact yet geometry-consistent anchor set. Building upon this optimized structure, we further develop a hierarchical entropy coding scheme, in which coarse-to-fine priors are exploited through a lightweight Geometry-Guided Convolution (GG-Conv) operator to enable spatially adaptive context modeling and rate-distortion optimization. Extensive experiments demonstrate that LG-HCC effectively resolves the structure preservation bottleneck, maintaining superior geometric integrity and rendering fidelity over state-of-the-art anchor-based compression approaches.

Orthogonal feature decorrelation is effective for low-bit online vector quantization, but dense random orthogonal transforms incur prohibitive $O(d^2)$ storage and compute. RotorQuant reduces this cost with blockwise $3$D Clifford rotors, yet the resulting $3$D partition is poorly aligned with modern hardware and offers limited local mixing. We propose \textbf{IsoQuant}, a blockwise rotation framework based on quaternion algebra and the isoclinic decomposition of $SO(4)$. It represents each $4$D block as a quaternion and applies a closed-form transform $T(v)=q_L v \overline{q_R}$. This yields two main variants: \emph{IsoQuant-Full}, which realizes the full $SO(4)$ rotation, and \emph{IsoQuant-Fast}, which keeps only one isoclinic factor for lower cost; the framework also admits a lightweight $2$D special case. At $d=128$, IsoQuant-Full reduces forward rotation cost from about $2{,}408$ FMAs in RotorQuant to $1{,}024$, while IsoQuant-Fast further reduces it to $512$. Across $18$ fused CUDA settings with $d \in {128,256,512}$, bit widths ${2,3,4}$, and FP16/FP32 execution, IsoQuant achieves mean kernel-level speedups of about $4.5\times$--$4.7\times$ over RotorQuant while maintaining comparable reconstruction MSE, with peak speedups above $6\times$. Current validation is limited to the stage-1 quantize--dequantize path on synthetic normalized vectors; end-to-end KV-cache evaluation remains future work.

The demand for high-speed, low-latency, and energy-efficient object detection in autonomous systems -- such as advanced driver-assistance systems (ADAS), unmanned aerial vehicles (UAVs), and Industry 4.0 robotics -- has exposed the limitations of traditional Convolutional Neural Networks (CNNs). To address these challenges, we have developed AceleradorSNN, a third-generation artificial intelligence cognitive system. This architecture integrates a Neuromorphic Processing Unit (NPU) based on Spiking Neural Networks (SNNs) to process asynchronous data from Dynamic Vision Sensors (DVS), alongside a dynamically reconfigurable Cognitive Image Signal Processor (ISP) for RGB cameras. This paper details the hardware-oriented design of both IP cores, the evaluation of surrogate-gradienttrained SNN backbones, and the real-time streaming ISP architecture implemented on Field-Programmable Gate Arrays (FPGA).

AI agents are rapidly expanding in both capability and population: they now write code, operate computers across platforms, manage cloud infrastructure, and make purchasing decisions, while open-source frameworks such as OpenClaw are putting personal agents in the hands of millions and embodied agents are spreading across smartphones, vehicles, and robots. As the internet prepares to host billions of such entities, it is shifting toward what we call Open Agentic Web, a decentralized digital ecosystem in which agents from different users, organizations, and runtimes can discover one another, negotiate task boundaries, and delegate work across open technical and social surfaces at scale. Yet most of today's agents remain isolated tools or closed-ecosystem orchestrators rather than socially integrated participants in open networks. We argue that the next generation of agents must become Agentic Citizens, defined by three requirements: Agentic-Web-Native Collaboration, participation in open collaboration networks rather than only closed internal orchestration; Agent Identity and Personhood, continuity as a social entity rather than a resettable function call; and Lifelong Evolution, improvement across task performance, communication, and collaboration over time. We present Synergy, a general-purpose agent architecture and runtime harness for persistent, collaborative, and evolving agents on Open Agentic Web, grounding collaboration in session-native orchestration, repository-backed workspaces, and social communication; identity in typed memory, notes, agenda, skills, and persistent social relationships; and evolution in an experience-centered learning mechanism that proactively recalls rewarded trajectories at inference time.

Signal Temporal Logic (STL) is widely used to specify timed and safety-critical tasks for cyber-physical systems, but writing STL formulas directly is difficult for non-expert users. Natural language (NL) provides a convenient interface, yet its inherent structural ambiguity makes one-to-one translation into STL unreliable. In this paper, we propose an \textit{ambiguity-preserving} method for translating NL task descriptions into STL candidate formulas. The key idea is to retain multiple plausible syntactic analyses instead of forcing a single interpretation at the parsing stage. To this end, we develop a three-stage pipeline based on Combinatory Categorial Grammar (CCG): ambiguity-preserving $n$-best parsing, STL-oriented template-based semantic composition, and canonicalization with score aggregation. The proposed method outputs a deduplicated set of STL candidates with plausibility scores, thereby explicitly representing multiple possible formal interpretations of an ambiguous instruction. In contrast to existing one-best NL-to-logic translation methods, the proposed approach is designed to preserve attachment and scope ambiguity. Case studies on representative task descriptions demonstrate that the method generates multiple STL candidates for genuinely ambiguous inputs while collapsing unambiguous or canonically equivalent derivations to a single STL formula.

Neural networks are fundamental tools of modern machine learning. The standard paradigm assumes binary interactions (across feedforward linear passes) between inter-tangled units, organized in sequential layers. Generalized architectures have been also designed that move beyond pairwise interactions, so as to account for higher-order couplings among computing neurons. Higher-order networks are however usually deployed as augmented graph neural networks (GNNs), and, as such, prove solely advantageous in contexts where the input exhibits an explicit hypergraph structure. Here, we present Spectral Higher-Order Neural Networks (SHONNs), a new algorithmic strategy to incorporate higher-order interactions in general-purpose, feedforward, network structures. SHONNs leverages a reformulation of the model in terms of spectral attributes. This allows to mitigate the common stability and parameter scaling problems that come along weighted, higher-order, forward propagations.

Lombard, an underresourced language variety spoken by approximately 3.8 million people in Northern Italy and Southern Switzerland, lacks a unified orthographic standard. Multiple orthographic systems exist, creating challenges for NLP resource development and model training. This paper presents the first study of automatic Lombard orthography classification and LombardoGraphia, a curated corpus of 11,186 Lombard Wikipedia samples tagged across 9 orthographic variants, and models for automatic orthography classification. We curate the dataset, processing and filtering raw Wikipedia content to ensure text suitable for orthographic analysis. We train 24 traditional and neural classification models with various features and encoding levels. Our best models achieve 96.06% and 85.78% overall and average class accuracy, though performance on minority classes remains challenging due to data imbalance. Our work provides crucial infrastructure for building variety-aware NLP resources for Lombard.

This paper proposes a new univariate filter feature selection (FFS) algorithm called KGroups. The majority of work in the literature focuses on investigating the relevance or redundancy estimations of feature selection (FS) methods. This has shown promising results and a real improvement of FFS methods' predictive performance. However, limited efforts have been made to investigate alternative FFS algorithms. This raises the following question: how much of the FFS methods' predictive performance depends on the selection algorithm rather than the relevance or the redundancy estimations? The majority of FFS methods fall into two categories: relevance maximisation (Max-Rel, also known as KBest) or simultaneous relevance maximisation and redundancy minimisation (mRMR). KBest is a univariate FFS algorithm that employs sorting (descending) for selection. mRMR is a multivariate FFS algorithm that employs an incremental search algorithm for selection. In this paper, we propose a new univariate mRMR called KGroups that employs clustering for selection. Extensive experiments on 14 high-dimensional biological benchmark datasets showed that KGroups achieves similar predictive performance compared to multivariate mRMR while being up to 821 times faster. KGroups is parameterisable, which leaves room for further predictive performance improvement through hyperparameter finetuning, unlike mRMR and KBest. KGroups outperforms KBest.

Reinforcement learning algorithms are defined by their learning update rules, which are typically hand-designed and fixed. We present an evolutionary framework for discovering reinforcement learning algorithms by searching directly over executable update rules that implement complete training procedures. The approach builds on REvolve, an evolutionary system that uses large language models as generative variation operators, and extends it from reward-function discovery to algorithm discovery. To promote the emergence of nonstandard learning rules, the search excludes canonical mechanisms such as actor--critic structures, temporal-difference losses, and value bootstrapping. Because reinforcement learning algorithms are highly sensitive to internal scalar parameters, we introduce a post-evolution refinement stage in which a large language model proposes feasible hyperparameter ranges for each evolved update rule. Evaluated end-to-end by full training runs on multiple Gymnasium benchmarks, the discovered algorithms achieve competitive performance relative to established baselines, including SAC, PPO, DQN, and A2C.

Preference-based many-objective optimization faces two obstacles: an expanding space of trade-offs and heterogeneous, context-dependent human value structures. Towards this, we propose a Bayesian framework that learns a small set of latent preference archetypes rather than assuming a single fixed utility function, modelling them as components of a Dirichlet-process mixture with uncertainty over both archetypes and their weights. To query efficiently, we designing hybrid queries that target information about (i) mode identity and (ii) within-mode trade-offs. Under mild assumptions, we provide a simple regret guarantee for the resulting mixture-aware Bayesian optimization procedure. Empirically, our method outperforms standard baselines on synthetic and real-world many-objective benchmarks, and mixture-aware diagnostics reveal structure that regret alone fails to capture.

Recent progress in deep research systems has been impressive, but evaluation still lags behind real user needs. Existing benchmarks predominantly assess final reports using fixed rubrics, failing to evaluate the underlying research process. Most also offer limited multimodal coverage, rely on synthetic tasks that do not reflect real-world query complexity, and cannot be refreshed as knowledge evolves. To address these gaps, we introduce MiroEval, a benchmark and evaluation framework for deep research systems. The benchmark comprises 100 tasks (70 text-only, 30 multimodal), all grounded in real user needs and constructed via a dual-path pipeline that supports periodic updates, enabling a live and evolving setting. The proposed evaluation suite assesses deep research systems along three complementary dimensions: adaptive synthesis quality evaluation with task-specific rubrics, agentic factuality verification via active retrieval and reasoning over both web sources and multimodal attachments, and process-centric evaluation audits how the system searches, reasons, and refines throughout its investigation. Evaluation across 13 systems yields three principal findings: the three evaluation dimensions capture complementary aspects of system capability, with each revealing distinct strengths and weaknesses across systems; process quality serves as a reliable predictor of overall outcome while revealing weaknesses invisible to output-level metrics; and multimodal tasks pose substantially greater challenges, with most systems declining by 3 to 10 points. The MiroThinker series achieves the most balanced performance, with MiroThinker-H1 ranking the highest overall in both settings. Human verification and robustness results confirm the reliability of the benchmark and evaluation framework. MiroEval provides a holistic diagnostic tool for the next generation of deep research agents.

Diffusion Transformers (DiT) have established a new state-of-the-art in high-fidelity image synthesis; however, their massive computational complexity and memory requirements hinder local deployment on resource-constrained edge devices. In this paper, we introduce EdgeDiT, a family of hardware-efficient generative transformers specifically engineered for mobile Neural Processing Units (NPUs), such as the Qualcomm Hexagon and Apple Neural Engine (ANE). By leveraging a hardware-aware optimization framework, we systematically identify and prune structural redundancies within the DiT backbone that are particularly taxing for mobile data-flows. Our approach yields a series of lightweight models that achieve a 20-30% reduction in parameters, a 36-46% decrease in FLOPs, and a 1.65-fold reduction in on-device latency without sacrificing the scaling advantages or the expressive capacity of the original transformer architecture. Extensive benchmarking demonstrates that EdgeDiT offers a superior Pareto-optimal trade-off between Frechet Inception Distance (FID) and inference latency compared to both optimized mobile U-Nets and vanilla DiT variants. By enabling responsive, private, and offline generative AI directly on-device, EdgeDiT provides a scalable blueprint for transitioning large-scale foundation models from high-end GPUs to the palm of the user.

Machine Learning (ML)-based detectors are becoming essential to counter the proliferation of malware. However, common ML algorithms are not designed to cope with the dynamic nature of real-world settings, where both legitimate and malicious software evolve. This distribution drift causes models trained under static assumptions to degrade over time unless they are continuously updated. Regularly retraining these models, however, is expensive, since labeling new acquired data requires costly manual analysis by security experts. To reduce labeling costs and address distribution drift in malware detection, prior work explored active learning (AL) and semi-supervised learning (SSL) techniques. Yet, existing studies (i) are tightly coupled to specific detector architectures and restricted to a specific malware domain, resulting in non-uniform comparisons; and (ii) lack a consistent methodology for analyzing the distribution drift, despite the critical sensitivity of the malware domain to temporal changes. In this work, we bridge this gap by proposing a model-agnostic framework that evaluates an extensive set of AL and SSL techniques, isolated and combined, for Android and Windows malware detection. We show that these techniques, when combined, can reduce manual annotation costs by up to 90% across both domains while achieving comparable detection performance to full-labeling retraining. We also introduce a methodology for feature-level drift analysis that measures feature stability over time, showing its correlation with the detector performance. Overall, our study provides a detailed understanding of how AL and SSL behave under distribution drift and how they can be successfully combined, offering practical insights for the design of effective detectors over time.

Network performance modeling is a field that predates early computer networks and the beginning of the Internet. It aims to predict the traffic performance of packet flows in a given network. Its applications range from network planning and troubleshooting to feeding information to network controllers for configuration optimization. Traditional network performance modeling has relied heavily on Discrete Event Simulation (DES) and analytical methods grounded in mathematical theories such as Queuing Theory and Network Calculus. However, as of late, we have observed a paradigm shift, with attempts to obtain efficient Parallel DES, the surge of Machine Learning models, and their integration with other methodologies in hybrid approaches. This has resulted in a great variety of modeling approaches, each with its strengths and often tailored to specific scenarios or requirements. In this paper, we comprehensively survey the relevant network performance modeling approaches for wired networks over the last decades. With this understanding, we also define a taxonomy of approaches, summarizing our understanding of the state-of-the-art and how both technology and the concerns of the research community evolve over time. Finally, we also consider how these models are evaluated, how their different nature results in different evaluation requirements and goals, and how this may complicate their comparison.

Trustworthy clinical AI requires that performance gains reflect genuine evidence integration rather than surface-level artifacts. We evaluate 12 open-weight vision-language models (VLMs) on binary classification across two clinical neuroimaging cohorts, \textsc{FOR2107} (affective disorders) and \textsc{OASIS-3} (cognitive decline). Both datasets come with structural MRI data that carries no reliable individual-level diagnostic signal. Under these conditions, smaller VLMs exhibit gains of up to 58\% F1 upon introduction of neuroimaging context, with distilled models becoming competitive with counterparts an order of magnitude larger. A contrastive confidence analysis reveals that merely \emph{mentioning} MRI availability in the task prompt accounts for 70-80\% of this shift, independent of whether imaging data is present, a domain-specific instance of modality collapse we term the \emph{scaffold effect}. Expert evaluation reveals fabrication of neuroimaging-grounded justifications across all conditions, and preference alignment, while eliminating MRI-referencing behavior, collapses both conditions toward random baseline. Our findings demonstrate that surface evaluations are inadequate indicators of multimodal reasoning, with direct implications for the deployment of VLMs in clinical settings.

A central challenge in building continually improving agents is that training environments are typically static or manually constructed. This restricts continual learning and generalization beyond the training distribution. We address this with COvolve, a co-evolutionary framework that leverages large language models (LLMs) to generate both environments and agent policies, expressed as executable Python code. We model the interaction between environment and policy designers as a two-player zero-sum game, ensuring adversarial co-evolution in which environments expose policy weaknesses and policies adapt in response. This process induces an automated curriculum in which environments and policies co-evolve toward increasing complexity. To guarantee robustness and prevent forgetting as the curriculum progresses, we compute the mixed-strategy Nash equilibrium (MSNE) of the zero-sum game, thereby yielding a meta-policy. This MSNE meta-policy ensures that the agent does not forget to solve previously seen environments while learning to solve previously unseen ones. Experiments in urban driving, symbolic maze-solving, and geometric navigation showcase that COvolve produces progressively more complex environments. Our results demonstrate the potential of LLM-driven co-evolution to achieve open-ended learning without predefined task distributions or manual intervention.

Maritime surveillance missions, such as search and rescue and environmental monitoring, rely on the efficient allocation of sensing assets over vast and geometrically complex areas. Traditional Coverage Path Planning (CPP) approaches depend on decomposition techniques that struggle with irregular coastlines, islands, and exclusion zones, or require computationally expensive re-planning for every instance. We propose a Deep Reinforcement Learning (DRL) framework to solve CPP on hexagonal grid representations of irregular maritime areas. Unlike conventional methods, we formulate the problem as a neural combinatorial optimization task where a Transformer-based pointer policy autoregressively constructs coverage tours. To overcome the instability of value estimation in long-horizon routing problems, we implement a critic-free Group-Relative Policy Optimization (GRPO) scheme. This method estimates advantages through within-instance comparisons of sampled trajectories rather than relying on a value function. Experiments on 1,000 unseen synthetic maritime environments demonstrate that a trained policy achieves a 99.0% Hamiltonian success rate, more than double the best heuristic (46.0%), while producing paths 7% shorter and with 24% fewer heading changes than the closest baseline. All three inference modes (greedy, stochastic sampling, and sampling with 2-opt refinement) operate under 50~ms per instance on a laptop GPU, confirming feasibility for real-time on-board deployment.

Static timing analysis (STA) is crucial for Electronic Design Automation (EDA) flows but remains a computational bottleneck. While existing GPU-based STA engines are faster than CPU, they suffer from inefficiencies, particularly intra-warp load imbalance caused by irregular circuit graphs. This paper introduces Warp-STAR, a novel GPU-accelerated STA engine that eliminates this imbalance by orchestrating parallel computations at the warp level. This approach achieves a 2.4X speedup over previous state-of-the-art (SoTA) GPU-based STA. When integrated into a timing-driven global placement framework, Warp-STAR delivers a 1.7X speedup over SoTA frameworks. The method also proves effective for differentiable gradient analysis with minimal overhead.

We present the first systematic Membership Inference Attack (MIA) evaluation of Large Audio Language Models (LALMs). As audio encodes non-semantic information, it induces severe train and test distribution shifts and can lead to spurious MIA performance. Using a multi-modal blind baseline based on textual, spectral, and prosodic features, we demonstrate that common speech datasets exhibit near-perfect train/test separability (AUC approximately 1.0) even without model inference, and the standard MIA scores strongly correlate with these blind acoustic artifacts (correlation greater than 0.7). Using this blind baseline, we identify that distribution-matched datasets enable reliable MIA evaluation without distribution shift confounds. We benchmark multiple MIA methods and conduct modality disentanglement experiments on these datasets. The results reveal that LALM memorization is cross-modal, arising only from binding a speaker's vocal identity with its text. These findings establish a principled standard for auditing LALMs beyond spurious correlations.

Deep research agents autonomously conduct open-ended investigations, integrating complex information retrieval with multi-step reasoning across diverse sources to solve real-world problems. To sustain this capability on long-horizon tasks, reliable verification is critical during both training and inference. A major bottleneck in existing paradigms stems from the lack of explicit verification mechanisms in QA data synthesis, trajectory construction, and test-time scaling. Errors introduced at each stage propagate downstream and degrade the overall agent performance. To address this, we present Marco DeepResearch, a deep research agent optimized with a verification-centric framework design at three levels: \textbf{(1)~QA Data Synthesis:} We introduce verification mechanisms to graph-based and agent-based QA synthesis to control question difficulty while ensuring answers are unique and correct; \textbf{(2)~Trajectory Construction:} We design a verification-driven trajectory synthesis method that injects explicit verification patterns into training trajectories; and \textbf{(3)~Test-time scaling:} We use Marco DeepResearch itself as a verifier at inference time and effectively improve performance on challenging questions. Extensive experimental results demonstrate that our proposed Marco DeepResearch agent significantly outperforms 8B-scale deep research agents on most challenging benchmarks, such as BrowseComp and BrowseComp-ZH. Crucially, under a maximum budget of 600 tool calls, Marco DeepResearch even surpasses or approaches several 30B-scale agents, like Tongyi DeepResearch-30B.

When an agent can articulate why something works, we typically take this as evidence of genuine understanding. This presupposes that effective action and correct explanation covary, and that coherent explanation reliably signals both. I argue that this assumption fails for contemporary Large Language Models (LLMs). I introduce what I call the Bidirectional Coherence Paradox: competence and grounding not only dissociate but invert across epistemic conditions. In low-observability domains, LLMs often act successfully while misidentifying the mechanisms that produce their success. In high-observability domains, they frequently generate explanations that accurately track observable causal structure yet fail to translate those diagnoses into effective intervention. In both cases, explanatory coherence remains intact, obscuring the underlying dissociation. Drawing on experiments in compiler optimization and hyperparameter tuning, I develop the Epistemic Triangle, a model of how priors, signals, and domain knowledge interact under varying observability. The results suggest that neither behavioral success nor explanatory accuracy alone suffices for attributing understanding. I argue that evaluating artificial epistemic agents requires a tripartite framework -- coherence, grounding, and a proper basing relation linking explanation to action. The systematic separation of knowing-that and knowing-how in LLMs thus challenges assumptions inherited from both epistemology and current AI evaluation practice.

Neurodiverse learners often require reading supports, yet increasing scaffold richness can sometimes overload attention and working memory rather than improve comprehension. Grounded in the Construction-Integration model and a contingent scaffolding perspective, we examine how structural versus semantic scaffolds shape comprehension and reading experience in a supervised inclusive context. Using an adapted reading interface, we compared four modalities: unmodified text, sentence-segmented text, segmented text with pictograms, and segmented text with pictograms plus keyword labels. In a within-subject pilot with 14 primary-school learners with special educational needs and disabilities, we measured reading comprehension using standardized questions and collected brief child- and therapist-reported experience measures alongside open-ended feedback. Results highlight heterogeneous responses as some learners showed patterns consistent with benefits from segmentation and pictograms, while others showed patterns consistent with increased coordination costs when visual scaffolds were introduced. Experience ratings showed limited differences between modalities, with some apparent effects linked to clinical complexity, particularly for perceived ease of understanding. Open-ended feedback of the learners frequently requested simpler wording and additional visual supports. These findings suggest that no single scaffold is universally optimal, reinforcing the need for calibrated, adjustable scaffolding and provide design implications for human-AI co-regulation in supervised inclusive reading contexts.

With the advancement of large language models (LLMs) in their knowledge base and reasoning capabilities, their interactive modalities have evolved from pure text to multimodality and further to agentic tool use. Consequently, their applications have broadened from question answering to AI assistants and now to general-purpose agents. Deep research (DR) represents a prototypical vertical application for general-purpose agents, which represents an ideal approach for intelligent information processing and assisting humans in discovering and solving problems, with the goal of reaching or even surpassing the level of top human scientists. This paper provides a deep research of deep research. We articulate a clear and precise definition of deep research and unify perspectives from industry's deep research and academia's AI for Science (AI4S) within a developmental framework. We position LLMs and Stable Diffusion as the twin pillars of generative AI, and lay out a roadmap evolving from the Transformer to agents. We examine the progress of AI4S across various disciplines. We identify the predominant paradigms of human-AI interaction and prevailing system architectures, and discuss the major challenges and fundamental research issues that remain. AI supports scientific innovation, and science also can contribute to AI growth (Science for AI, S4AI). We hope this paper can help bridge the gap between the AI and AI4S communities.

Uncertainty estimation in multi-LLM systems remains largely single-model-centric: existing methods quantify uncertainty within each model but do not adequately capture semantic disagreement across models. To address this gap, we propose Collaborative Entropy (CoE), a unified information-theoretic metric for semantic uncertainty in multi-LLM collaboration. CoE is defined on a shared semantic cluster space and combines two components: intra-model semantic entropy and inter-model divergence to the ensemble mean. CoE is not a weighted ensemble predictor; it is a system-level uncertainty measure that characterizes collaborative confidence and disagreement. We analyze several core properties of CoE, including non-negativity, zero-value certainty under perfect semantic consensus, and the behavior of CoE when individual models collapse to delta distributions. These results clarify when reducing per-model uncertainty is sufficient and when residual inter-model disagreement remains. We also present a simple CoE-guided, training-free post-hoc coordination heuristic as a practical application of the metric. Experiments on \textit{TriviaQA} and \textit{SQuAD} with LLaMA-3.1-8B-Instruct, Qwen-2.5-7B-Instruct, and Mistral-7B-Instruct show that CoE provides stronger uncertainty estimation than standard entropy- and divergence-based baselines, with gains becoming larger as additional heterogeneous models are introduced. Overall, CoE offers a useful uncertainty-aware perspective on multi-LLM collaboration.

The accurate classification of brain tumors from MRI scans is essential for effective diagnosis and treatment planning. This paper presents a weighted ensemble learning approach that combines deep learning and traditional machine learning models to improve classification performance. The proposed system integrates multiple classifiers, including ResNet101, DenseNet121, Xception, CNN-MRI, and ResNet50 with edge-enhanced images, SVM, and KNN with HOG features. A weighted voting mechanism assigns higher influence to models with better individual accuracy, ensuring robust decision-making. Image processing techniques such as Balance Contrast Enhancement, K-means clustering, and Canny edge detection are applied to enhance feature extraction. Experimental evaluations on the Figshare and Kaggle MRI datasets demonstrate that the proposed method achieves state-of-the-art accuracy, outperforming existing models. These findings highlight the potential of ensemble-based learning for improving brain tumor classification, offering a reliable and scalable framework for medical image analysis.

Subjective judgments are part of several NLP datasets and recent work is increasingly prioritizing models whose outputs reflect this diversity of perspectives. Such responses allow us to shed light on minority voices, which are frequently marginalized or obscured by dominant perspectives. It remains a question whether our evaluation practices align with these models' objectives. This position paper proposes seven evaluation desiderata for subjectivity-sensitive models, rooted in how subjectivity is represented in NLP data and models. The desiderata are constructed in a top-down approach, keeping in mind the user-centric impact of such models. We scan the experimental setup of 60 papers and show that various aspects of subjectivity are still understudied: the distinction between ambiguous and polyphonic input, whether subjectivity is effectively expressed to the user, and a lack of interplay between different desiderata, amongst other gaps.

Efficient estimation of high-dimensional matrices-including covariance and precision matrices-is a cornerstone of modern multivariate statistics. Most existing studies have focused primarily on the theoretical properties of the estimators (e.g., consistency and sparsity), while largely overlooking the computational challenges inherent in high-dimensional settings. Motivated by recent advances in learning-based optimization method-which integrate data-driven structures with classical optimization algorithms-we explore high-dimensional matrix estimation assisted by machine learning. Specifically, for the optimization problem of high-dimensional matrix estimation, we first present a solution procedure based on the Linearized Alternating Direction Method of Multipliers (LADMM). We then introduce learnable parameters and model the proximal operators in the iterative scheme with neural networks, thereby improving estimation accuracy and accelerating convergence. Theoretically, we first prove the convergence of LADMM, and then establish the convergence, convergence rate, and monotonicity of its reparameterized counterpart; importantly, we show that the reparameterized LADMM enjoys a faster convergence rate. Notably, the proposed reparameterization theory and methodology are applicable to the estimation of both high-dimensional covariance and precision matrices. We validate the effectiveness of our method by comparing it with several classical optimization algorithms across different structures and dimensions of high-dimensional matrices.

LLM API calls are becoming a ubiquitous program construct, yet they create a boundary that no existing program analysis can cross: runtime values enter a natural-language prompt, undergo opaque processing inside the LLM, and re-emerge as code, SQL, JSON, or text that the program consumes. Every analysis that tracks data across function boundaries, including taint analysis, program slicing, dependency analysis, and change-impact analysis, relies on dataflow summaries of callee behavior. LLM calls have no such summaries, breaking all of these analyses at what we call the NL/PL boundary. We present the first information flow method to bridge this boundary. Grounded in quantitative information flow theory, our taxonomy defines 24 labels along two orthogonal dimensions: information preservation level (from lexically preserved to fully blocked) and output modality (natural language, structured format, executable artifact). We label 9,083 placeholder-output pairs from 4,154 real-world Python files and validate reliability with Cohen's $κ= 0.82$ and near-complete coverage (0.01\% unclassifiable). We demonstrate the taxonomy's utility on two downstream applications: (1)~a two-stage taint propagation pipeline combining taxonomy-based filtering with LLM verification achieves $F_1 = 0.923$ on 353 expert-annotated pairs, with cross-language validation on six real-world OpenClaw prompt injection cases further confirming effectiveness; (2)~taxonomy-informed backward slicing reduces slice size by a mean of 15\% in files containing non-propagating placeholders. Per-label analysis reveals that four blocked labels account for nearly all non-propagating cases, providing actionable filtering criteria for tool builders.

We present Kernel-Smith, a framework for high-performance GPU kernel and operator generation that combines a stable evaluation-driven evolutionary agent with an evolution-oriented post-training recipe. On the agent side, Kernel-Smith maintains a population of executable candidates and iteratively improves them using an archive of top-performing and diverse programs together with structured execution feedback on compilation, correctness, and speedup. To make this search reliable, we build backend-specific evaluation services for Triton on NVIDIA GPUs and Maca on MetaX GPUs. On the training side, we convert long-horizon evolution trajectories into step-centric supervision and reinforcement learning signals by retaining correctness-preserving, high-gain revisions, so that the model is optimized as a strong local improver inside the evolutionary loop rather than as a one-shot generator. Under a unified evolutionary protocol, Kernel-Smith-235B-RL achieves state-of-the-art overall performance on KernelBench with Nvidia Triton backend, attaining the best average speedup ratio and outperforming frontier proprietary models including Gemini-3.0-pro and Claude-4.6-opus. We further validate the framework on the MetaX MACA backend, where our Kernel-Smith-MACA-30B surpasses large-scale counterparts such as DeepSeek-V3.2-think and Qwen3-235B-2507-think, highlighting potential for seamless adaptation across heterogeneous platforms. Beyond benchmark results, the same workflow produces upstream contributions to production systems including SGLang and LMDeploy, demonstrating that LLM-driven kernel optimization can transfer from controlled evaluation to practical deployment.

Systematic literature reviews in the social sciences overwhelmingly follow arborescent logics -- hierarchical keyword filtering, linear screening, and taxonomic classification -- that suppress the lateral connections, ruptures, and emergent patterns characteristic of complex research landscapes. This research note presents the Rhizomatic Research Agent (V3), a multi-agent computational pipeline grounded in Deleuzian process-relational ontology, designed to conduct non-linear literature analysis through 12 specialized agents operating across a seven-phase architecture. The system was developed in response to the methodological groundwork established by (Narayan2023), who employed rhizomatic inquiry in her doctoral research on sustainable energy transitions but relied on manual, researcher-driven exploration. The Rhizomatic Research Agent operationalizes the six principles of the rhizome -- connection, heterogeneity, multiplicity, asignifying rupture, cartography, and decalcomania -- into an automated pipeline integrating large language model (LLM) orchestration, dual-source corpus ingestion from OpenAlex and arXiv, SciBERT semantic topography, and dynamic rupture detection protocols. Preliminary deployment demonstrates the system's capacity to surface cross-disciplinary convergences and structural research gaps that conventional review methods systematically overlook. The pipeline is open-source and extensible to any phenomenon zone where non-linear knowledge mapping is required.

The rapid adoption of data-driven methods in biomedicine has intensified concerns over privacy, governance, and regulation, limiting raw data sharing and hindering the assembly of representative cohorts for clinically relevant AI. This landscape necessitates practical, efficient privacy solutions, as cryptographic defenses often impose heavy overhead and differential privacy can degrade performance, leading to sub-optimal outcomes in real-world settings. Here, we present a lightweight federated learning method, INFL, based on Implicit Neural Representations that addresses these challenges. Our approach integrates plug-and-play, coordinate-conditioned modules into client models, embeds a secret key directly into the architecture, and supports seamless aggregation across heterogeneous sites. Across diverse biomedical omics tasks, including cohort-scale classification in bulk proteomics, regression for perturbation prediction in single-cell transcriptomics, and clustering in spatial transcriptomics and multi-omics with both public and private data, we demonstrate that INFL achieves strong, controllable privacy while maintaining utility, preserving the performance necessary for downstream scientific and clinical applications.

With the widespread adoption of autonomous vehicles and robotics, amodal completion, which reconstructs the occluded parts of people and objects in an image, has become increasingly crucial. Just as humans infer hidden regions based on prior experience and common sense, this task inherently requires physical knowledge about real-world entities. However, existing approaches either depend solely on the image generation ability of visual generative models, which lack such knowledge, or leverage it only during the segmentation stage, preventing it from explicitly guiding the completion process. To address this, we propose AmodalCG, a novel framework that harnesses the real-world knowledge of Multimodal Large Language Models (MLLMs) to guide amodal completion. Our framework first assesses the extent of occlusion to selectively invoke MLLM guidance only when the target object is heavily occluded. If guidance is required, the framework further incorporates MLLMs to reason about both the (1) extent and (2) content of the missing regions. Finally, a visual generative model integrates these guidance and iteratively refines imperfect completions that may arise from inaccurate MLLM guidance. Experimental results on various real-world images show impressive improvements compared to all existing works, suggesting MLLMs as a promising direction for addressing challenging amodal completion.

Accurate prediction of hydrogen sorption in fine-grained geological materials is essential for evaluating underground hydrogen storage capacity, assessing caprock integrity, and characterizing hydrogen migration in subsurface energy systems. Classical isotherm models perform well at the individual-sample level but fail when generalized across heterogeneous populations, with the coefficient of determination collapsing from 0.80-0.90 for single-sample fits to 0.09-0.38 for aggregated multi-sample datasets. We present a multi-scale physics-informed neural network framework that addresses this limitation by embedding classical adsorption theory and thermodynamic constraints directly into the learning process. The framework utilizes 1,987 hydrogen sorption isotherm measurements across clays, shales, coals, supplemented by 224 characteristic uptake measurements. A seven-category physics-informed feature engineering scheme generates 62 thermodynamically meaningful descriptors from raw material characterization data. The loss function enforces saturation limits, a monotonic pressure response, and Van't Hoff temperature dependence via penalty weighting, while a three-phase curriculum-based training strategy ensures stable integration of competing physical constraints. An architecture-diverse ensemble of ten members provides calibrated uncertainty quantification, with post-hoc temperature scaling achieving target prediction interval coverage. The optimized PINN achieves R2 = 0.9544, RMSE = 0.0484 mmol/g, and MAE = 0.0231 mmol/g on the held-out test set, with 98.6% monotonicity satisfaction and zero non-physical negative predictions. Physics-informed regularization yields a 10-15% cross-lithology generalization advantage over a well-tuned random forest under leave-one-lithology-out validation, confirming that thermodynamic constraints transfer meaningfully across geological boundaries.

Biomedical knowledge resources often either preserve evidence as unstructured text or compress it into flat triples that omit study design, provenance, and quantitative support. Here we present EvidenceNet, a framework and dataset for building disease-specific knowledge graphs from full-text biomedical literature. EvidenceNet uses a large language model (LLM)-assisted pipeline to extract experimentally grounded findings as structured evidence nodes, normalize biomedical entities, score evidence quality, and connect evidence records through typed semantic relations. We release two resources: EvidenceNet-HCC with 7,872 evidence records, 10,328 graph nodes, and 49,756 edges, and EvidenceNet-CRC with 6,622 records, 8,795 nodes, and 39,361 edges. Technical validation shows high component fidelity, including 98.3% field-level extraction accuracy, 100.0% high-confidence entity-link accuracy, 87.5% fusion integrity, and 90.0% semantic relation-type accuracy. In downstream evaluation, EvidenceNet improves internal and external retrieval-augmented question answering and retains structural signal for future link prediction and target prioritization. These results establish EvidenceNet as a disease-specific resource for evidence-aware biomedical reasoning and hypothesis generation.

We introduce a novel conditional stochastic interpolant framework for generative modeling of three-dimensional shapes. The method builds on a recent LDDMM-based registration approach to learn the conditional drift between geometries. By leveraging the resulting pull-back and push-forward operators, we extend this formulation beyond standard Cartesian grids to complex shapes and random variables defined on distinct domains. We present an application in the context of cardiovascular simulations, where aortic shapes are generated from an initial cohort of patients. The conditioning variable is a latent geometric representation defined by a set of centerline points and the radii of the corresponding inscribed spheres. This methodology facilitates both data augmentation for three-dimensional biomedical shapes, and the generation of random perturbations of controlled magnitude for a given shape. These capabilities are essential for quantifying the impact of domain uncertainties arising from medical image segmentation on the estimation of relevant biomarkers.

Graph condensation (GC) has become a vital strategy for scaling Graph Neural Networks by compressing massive datasets into small, synthetic node sets. While current GC methods effectively maintain predictive accuracy, they are primarily designed for utility and often ignore fairness constraints. Because these techniques are bias-blind, they frequently capture and even amplify demographic disparities found in the original data. This leads to synthetic proxies that are unsuitable for sensitive applications like credit scoring or social recommendations. To solve this problem, we introduce FairGC, a unified framework that embeds fairness directly into the graph distillation process. Our approach consists of three key components. First, a Distribution-Preserving Condensation module synchronizes the joint distributions of labels and sensitive attributes to stop bias from spreading. Second, a Spectral Encoding module uses Laplacian eigen-decomposition to preserve essential global structural patterns. Finally, a Fairness-Enhanced Neural Architecture employs multi-domain fusion and a label-smoothing curriculum to produce equitable predictions. Rigorous evaluations on four real-world datasets, show that FairGC provides a superior balance between accuracy and fairness. Our results confirm that FairGC significantly reduces disparity in Statistical Parity and Equal Opportunity compared to existing state-of-the-art condensation models. The codes are available at https://github.com/LuoRenqiang/FairGC.

Data literacy skills are fundamental in computer science education. However, understanding how data-driven systems work represents a paradigm shift from traditional rule-based programming. We conducted a systematic literature review of 84 studies to understand K-12 learners' engagement with data across disciplines and contexts. We propose the data paradigms framework that categorises learning activities along two dimensions: (i) logic (knowledge-based or data-driven systems), and (ii) explainability (transparent or opaque models). We further apply the notion of learning trajectories to visualize the pathways learners follow across these distinct paradigms. We detail four distinct trajectories as a provocation for researchers and educators to reflect on how the notion of data literacy varies depending on the learning context. We suggest these trajectories could be useful to those concerned with the design of data literacy learning environments within and beyond CS education.

In this paper, we present Federated Robust Curvature Optimization (FedRCO), a novel second-order optimization framework designed to improve convergence speed and reduce communication cost in Federated Learning systems under statistical heterogeneity. Existing second-order optimization methods are often computationally expensive and numerically unstable in distributed settings. In contrast, FedRCO addresses these challenges by integrating an efficient approximate curvature optimizer with a provable stability mechanism. Specifically, FedRCO incorporates three key components: (1) a Gradient Anomaly Monitor that detects and mitigates exploding gradients in real-time, (2) a Fail-Safe Resilience protocol that resets optimization states upon numerical instability, and (3) a Curvature-Preserving Adaptive Aggregation strategy that safely integrates global knowledge without erasing the local curvature geometry. Theoretical analysis shows that FedRCO can effectively mitigate instability and prevent unbounded updates while preserving optimization efficiency. Extensive experiments show that FedRCO achieves superior robustness against diverse non-IID scenarios while achieving higher accuracy and faster convergence than both state-of-the-art first-order and second-order methods.

Thyroid nodule classification using ultrasound imaging is essential for early diagnosis and clinical decision-making; however, despite promising performance on in-distribution data, existing deep learning methods often exhibit limited robustness and generalisation when deployed across different ultrasound devices or clinical environments. This limitation is mainly attributed to the pronounced heterogeneity of thyroid ultrasound images, which can lead models to capture spurious correlations rather than reliable diagnostic cues. To address this challenge, we propose PEMV-thyroid, a Prototype-Enhanced Multi-View learning framework that accounts for data heterogeneity by learning complementary representations from multiple feature perspectives and refining decision boundaries through a prototype-based correction mechanism with mixed prototype information. By integrating multi-view representations with prototype-level guidance, the proposed approach enables more stable representation learning under heterogeneous imaging conditions. Extensive experiments on multiple thyroid ultrasound datasets demonstrate that PEMV-thyroid consistently outperforms state-of-the-art methods, particularly in cross-device and cross-domain evaluation scenarios, leading to improved diagnostic accuracy and generalisation performance in real-world clinical settings. The source code is available at https://github.com/chenyangmeii/Prototype-Enhanced-Multi-View-Learning.

Self++ is a design blueprint for human-AI symbiosis in extended reality (XR) that preserves human authorship while still benefiting from increasingly capable AI agents. Because XR can shape both perceptual evidence and action, apparently 'helpful' assistance can drift into over-reliance, covert persuasion, and blurred responsibility. Self++ grounds interaction in two complementary theories: Self-Determination Theory (autonomy, competence, relatedness) and the Free Energy Principle (predictive stability under uncertainty). It operationalises these foundations through co-determination, treating the human and the AI as a coupled system that must keep intent and limits legible, tune support over time, and preserve the user's right to endorse, contest, and override. These requirements are summarised as the co-determination principles (T.A.N.): Transparency, Adaptivity, and Negotiability. Self++ organises augmentation into three concurrently activatable overlays spanning sensorimotor competence support (Self: competence overlay), deliberative autonomy support (Self+: autonomy overlay), and social and long-horizon relatedness and purpose support (Self++: relatedness and purpose overlay). Across the overlays, it specifies nine role patterns (Tutor, Skill Builder, Coach; Choice Architect, Advisor, Agentic Worker; Contextual Interpreter, Social Facilitator, Purpose Amplifier) that can be implemented as interaction patterns, not personas. The contribution is a role-based map for designing and evaluating XR-AI systems that grow capability without replacing judgment, enabling symbiotic agency in work, learning, and social life and resilient human development.

LLM-as-a-Judge has emerged as an effective and low-cost paradigm for evaluating text quality and factual correctness. Prior studies have shown substantial agreement between LLM judges and human experts, even on tasks that are difficult to assess automatically. In practice, researchers commonly employ fixed temperature configurations during the evaluation process-with values of 0.1 and 1.0 being the most prevalent choices-a convention that is largely empirical rather than principled. However, recent researches suggest that LLM performance exhibits non-trivial sensitivity to temperature settings, that lower temperatures do not universally yield optimal outcomes, and that such effects are highly task-dependent. This raises a critical research question: does temperature influence judge performance in LLM centric evaluation? To address this, we systematically investigate the relationship between temperature and judge performance through a series of controlled experiments, and further adopt a causal inference framework within our empirical statistical analysis to rigorously examine the direct causal effect of temperature on judge behavior, offering actionable engineering insights for the design of LLM-centric evaluation pipelines.

Vision-Language-Action (VLA) models achieve strong performance in robotic manipulation by leveraging pre-trained vision-language backbones. However, in downstream robotic settings, they are typically fine-tuned with limited data, leading to overfitting to specific instruction formulations and leaving robustness to paraphrased instructions underexplored. To study this gap, we introduce LIBERO-Para, a controlled benchmark that independently varies action expressions and object references for fine-grained analysis of linguistic generalization. Across seven VLA configurations (0.6B-7.5B), we observe consistent performance degradation of 22-52 pp under paraphrasing. This degradation is primarily driven by object-level lexical variation: even simple synonym substitutions cause large drops, indicating reliance on surface-level matching rather than semantic grounding. Moreover, 80-96% of failures arise from planning-level trajectory divergence rather than execution errors, showing that paraphrasing disrupts task identification. Binary success rate treats all paraphrases equally, obscuring whether models perform consistently across difficulty levels or rely on easier cases. To address this, we propose PRIDE, a metric that quantifies paraphrase difficulty using semantic and syntactic factors. Our benchmark and corresponding code are available at: https://github.com/cau-hai-lab/LIBERO-Para

Graph anomaly detection (GAD) aims to identify irregular nodes or structures in attributed graphs. Neighbor information, which reflects both structural connectivity and attribute consistency with surrounding nodes, is essential for distinguishing anomalies from normal patterns. Although recent graph neural network (GNN)-based methods incorporate such information through message passing, they often fail to explicitly model its effect or interaction with attributes, limiting detection performance. This work introduces NeiGAD, a novel plug-and-play module that captures neighbor information through spectral graph analysis. Theoretical insights demonstrate that eigenvectors of the adjacency matrix encode local neighbor interactions and progressively amplify anomaly signals. Based on this, NeiGAD selects a compact set of eigenvectors to construct efficient and discriminative representations. Experiments on eight real-world datasets show that NeiGAD consistently improves detection accuracy and outperforms state-of-the-art GAD methods. These results demonstrate the importance of explicit neighbor modeling and the effectiveness of spectral analysis in anomaly detection. Code is available at: https://github.com/huafeihuang/NeiGAD.

The rapid emergence of Large Language Models (LLMs) presents both opportunities and challenges for programming education. While students increasingly use generative AI tools, direct access often hinders the learning process by providing complete solutions rather than pedagogical hints. Concurrently, educators face significant workload and scalability challenges when providing timely, personalized feedback. This study investigates the capabilities of LLMs to safely and effectively assist educators in answering student questions within a CS1 programming course. To achieve this, we established a rigorous, reproducible evaluation process by curating a benchmark dataset of 170 authentic student questions from a learning management system, paired with ground-truth responses authored by subject matter experts. Because traditional text-matching metrics are insufficient for evaluating open-ended educational responses, we developed and validated a custom LLM-as-a-Judge metric optimized for assessing pedagogical accuracy. Our findings demonstrate that models, such as Gemini 3 flash, can surpass the quality baseline of typical educator responses, achieving high alignment with expert pedagogical standards. To mitigate persistent risks like hallucination and ensure alignment with course-specific context, we advocate for a "teacher-in-the-loop" implementation. Finally, we abstract our methodology into a task-agnostic evaluation framework, advocating for a shift in the development of educational LLM tools from ad-hoc, post-deployment testing to a quantifiable, pre-deployment validation process.

Distributed learning is widely used for training large models on large datasets by distributing parts of the model or dataset across multiple devices and aggregating the computed results for subsequent computations or parameter updates. Existing communication algorithms for distributed learning such as ring all-reduce result in heavy communication overhead between servers. Since communication in large-scale systems uses optical fibers, we propose an Optical In-Network-Computing (OptINC) architecture to offload the computation in servers onto the optical interconnects. To execute gradient averaging and quantization in the optical domain, we incorporate optical devices such as Mach-Zehnder-Interferometers (MZIs) into the interconnects. Such a de facto optical neural network (ONN) can effectively reduce the communication overhead in existing distributed training solutions. To reduce dataset complexity for training this neural network, a preprocessing algorithm implemented in the optical domain is also proposed. Hardware cost is lowered by approximating the weight matrices of the optical neural network with unitary and diagonal matrices, while the accuracy is maintained by a proposed hardware-aware training algorithm. The proposed solution was evaluated on real distributed learning tasks, including ResNet50 on CIFAR-100, and a LLaMA-based network on Wikipedia-1B. In both cases, the proposed framework can achieve comparable training accuracy to the ring all-reduce baseline, while eliminating communication overhead.

Kolmogorov-Arnold Networks (KAN) employ B-spline bases on a fixed grid, providing no intrinsic multi-scale decomposition for non-smooth function approximation. We introduce Fractal Interpolation KAN (FI-KAN), which incorporates learnable fractal interpolation function (FIF) bases from iterated function system (IFS) theory into KAN. Two variants are presented: Pure FI-KAN (Barnsley, 1986) replaces B-splines entirely with FIF bases; Hybrid FI-KAN (Navascues, 2005) retains the B-spline path and adds a learnable fractal correction. The IFS contraction parameters give each edge a differentiable fractal dimension that adapts to target regularity during training. On a Holder regularity benchmark ($α\in [0.2, 2.0]$), Hybrid FI-KAN outperforms KAN at every regularity level (1.3x to 33x). On fractal targets, FI-KAN achieves up to 6.3x MSE reduction over KAN, maintaining 4.7x advantage at 5 dB SNR. On non-smooth PDE solutions (scikit-fem), Hybrid FI-KAN achieves up to 79x improvement on rough-coefficient diffusion and 3.5x on L-shaped domain corner singularities. Pure FI-KAN's complementary behavior, dominating on rough targets while underperforming on smooth ones, provides controlled evidence that basis geometry must match target regularity. A fractal dimension regularizer provides interpretable complexity control whose learned values recover the true fractal dimension of each target. These results establish regularity-matched basis design as a principled strategy for neural function approximation.

Whittaker smoother is a widely adopted solution to pre-process satellite image time series. Yet, two key limitations remain: the smoothing parameter must be tuned individually for each pixel, and the standard formulation assumes homoscedastic noise, imposing uniform smoothing across the temporal dimension. This paper addresses both limitations by casting the Whittaker smoother as a differentiable neural layer, in which the smoothing parameter is inferred by a neural network. The framework is further extended to handle heteroscedastic noise through a time-varying regularization, allowing the degree of smoothing to adapt locally along the time series. To enable large-scale processing, a sparse, memory-efficient, and fully differentiable implementation is proposed, exploiting the symmetric banded structure of the underlying linear system via Cholesky factorization. Benchmarks on GPU demonstrate that this implementation substantially outperforms standard dense linear solvers, both in speed and memory consumption. The approach is validated on SITS acquired over the French metropolitan territory between 2016 and 2024. Results confirm the feasibility of large-scale heteroscedastic Whittaker smoothing, though reconstruction differences with the homoscedastic baseline remain limited, suggesting that the transformer architecture used for smoothing parameter estimation may lack the temporal acuity needed to capture abrupt noise variations such as singleday cloud contamination.

Edge AI systems increasingly rely on federated learning to train perception models in distributed, privacy-preserving, and resource-constrained environments. Yet, before training begins, practitioners often lack practical tools to estimate how difficult a federated learning task will be in terms of achievable accuracy and communication cost. This paper presents a classifier-agnostic, pre-deployment framework for estimating learning complexity in federated perception systems by jointly modeling intrinsic properties of the data and characteristics of the distributed environment. The proposed complexity metric integrates dataset attributes such as dimensionality, sparsity, and heterogeneity with factors related to the composition of participating clients. Using federated learning as a representative distributed training setting, we examine how learning difficulty varies across different federated configurations. Experiments on multiple variants of the MNIST dataset and CIFAR dataset show that the proposed metric strongly correlates with federated learning performance and the communication effort required to reach fixed accuracy targets. These findings suggest that complexity estimation can serve as a practical diagnostic tool for resource planning, dataset assessment, and feasibility evaluation in edge-deployed perception systems.

We consider robustness against data corruption in offline multi-agent reinforcement learning from human feedback (MARLHF) under a strong-contamination model: given a dataset $D$ of trajectory-preference tuples (each preference being an $n$-dimensional binary label vector representing each of the $n$ agents' preferences), an $ε$-fraction of the samples may be arbitrarily corrupted. We model the problem using the framework of linear Markov games. First, under a uniform coverage assumption - where every policy of interest is sufficiently represented in the clean (prior to corruption) data - we introduce a robust estimator that guarantees an $O(ε^{1 - o(1)})$ bound on the Nash equilibrium gap. Next, we move to the more challenging unilateral coverage setting, in which only a Nash equilibrium and its single-player deviations are covered. In this case, our proposed algorithm achieves an $O(\sqrtε)$ bound on the Nash gap. Both of these procedures, however, suffer from intractable computation. To address this, we relax our solution concept to coarse correlated equilibria (CCE). Under the same unilateral coverage regime, we derive a quasi-polynomial-time algorithm whose CCE gap scales as $O(\sqrtε)$. To the best of our knowledge, this is the first systematic treatment of adversarial data corruption in offline MARLHF.

Large Language Models (LLMs) remain heavily centered on English, with limited performance in low-resource languages. Existing adaptation approaches, such as continual pre-training, demand significant computational resources. In the case of instructed models, high-quality instruction data is also required, both of which are often inaccessible for low-resource language communities. Under these constraints, model merging offers a lightweight alternative, but its potential in low-resource contexts has not been systematically explored. In this work, we explore whether it is possible to transfer language knowledge to an instruction-tuned LLM by merging it with a language-specific base model, thereby eliminating the need of language-specific instructions and repeated fine-tuning processes whenever stronger instructed variants become available. Through experiments covering four Iberian languages (Basque, Catalan, Galician, and Spanish) and two model families, we show that merging enables effective instruction following behavior in new languages and even supports multilingual capability through the combination of multiple language-specific models. Our results indicate that model merging is a viable and efficient alternative to traditional adaptation methods for low-resource languages, achieving competitive performance while greatly reducing computational cost.

The paper proposes annotation guidelines for syntactic dependencies that span across speaker turns - including collaborative coconstructions proper, wh-question answers, and backchannels - in spoken language treebanks within the Universal Dependencies framework. Two representations are proposed: a speaker-based representation following the segmentation into speech turns, and a dependency-based representation with dependencies across speech turns. New propositions are also put forward to distinguish between reformulations and repairs, and to promote elements in unfinished phrases.

In the content-oblivious (CO) model (proposed by Censor-Hillel et al.), processes inhabit an asynchronous network and communicate only by exchanging pulses. A series of works has clarified the computational power of this model. In particular, it was shown that, when a leader is present and the network is 2-edge-connected, content-oblivious communication can simulate classical asynchronous message passing. Subsequent results extended this equivalence to leaderless oriented and unoriented rings, and, under non-uniform assumptions, to general 2-edge-connected networks. The simulator of Censor-Hillel et al. requires $O(n^3b+n^3\log n)$ pulses to emulate the send of a single $b$-bit message, making it impractical even on modest-size networks. We focus on message-efficient computation in CO networks. We study the fundamental problem of counting in ring topologies, both because knowing the exact network size is a basic prerequisite for many distributed tasks and because counting immediately implies a broad class of aggregation primitives. We give an algorithm that counts using $O(n^{1.5})$ pulses in anonymous rings with a leader, an $O(n\log^2 n)$ algorithm for counting in rings with IDs. Moreover, we show that any counting algorithm in CO requires $Ω(n\log n)$ pulses. Interestingly, in the course of this investigation, we design a simulator for classic message passing: in one simulated round, each process can send a $b$-bit message to each of its neighbors using only $O(b)$ pulses per process. The simulator extends to general 2-edge-connected networks, after a pre-processing step that requires $O(n^{8}\log n)$ pulses, where $n$ is the number of processes, allowing thus efficient simulation of asynchronous message passing in general 2-edge-connected networks.

Quantum algorithms require encoding classical vectors as quantum states, a step known as amplitude encoding. General-purpose state preparation routines accept any input vector of length $N = 2^m$ and produce circuits with $\bigO{2^m}$ gates. However, vectors arising in scientific and engineering applications often exhibit mathematical structure that admits far more efficient encoding. Recent theoretical work has established closed-form circuits for several structured vector classes, but without open-source implementations. We present PyEncode, an open-source Python library that implements this body of theory in a unified, immediately deployable framework. The library covers sparse, step, square (general interval), Walsh, geometric, and Fourier patterns, and supports weighted superpositions of pattern states via the linear combination of unitaries (LCU) protocol, enabling exact preparation of piecewise-structured vectors such as multi-interval Hamiltonians. PyEncode exposes a single function encode(VectorObj, N) that maps a typed parameter declaration directly to a verified Qiskit circuit, with no vector materialization and no approximation. Sparse, step, and Walsh vectors require only $\bigO{m}$ gates; geometric (exponential-decay) vectors require $\bigO{m}$ gates with zero two-qubit gates; square (general interval) vectors require $\bigO{m^2}$ gates via a QFT-based constant adder, with $\bigO{m}$ special cases; Fourier (sinusoidal) vectors require $\bigO{m^2}$ gates via the inverse Quantum Fourier Transform -- all exponentially fewer than the $\bigO{2^m}$ cost of general state preparation. LCU combines $r$ component circuits whose total gate cost is the sum of individual component costs, with success probability $p \in (0,1]$ determined analytically. The library is available at https://github.com/UW-ERSL/PyEncode.

Categorical perception (CP) -- enhanced discriminability at category boundaries -- is among the most studied phenomena in perceptual psychology. This paper reports that analogous geometric warping occurs in the hidden-state representations of large language models (LLMs) processing Arabic numerals. Using representational similarity analysis across six models from five architecture families, the study finds that a CP-additive model (log-distance plus a boundary boost) fits the representational geometry better than a purely continuous model at 100% of primary layers in every model tested. The effect is specific to structurally defined boundaries (digit-count transitions at 10 and 100), absent at non-boundary control positions, and absent in the temperature domain where linguistic categories (hot/cold) lack a tokenisation discontinuity. Two qualitatively distinct signatures emerge: "classic CP" (Gemma, Qwen), where models both categorise explicitly and show geometric warping, and "structural CP" (Llama, Mistral, Phi), where geometry warps at the boundary but models cannot report the category distinction. This dissociation is stable across boundaries and is a property of the architecture, not the stimulus. Structural input-format discontinuities are sufficient to produce categorical perception geometry in LLMs, independently of explicit semantic category knowledge.

KAN-PCA is an autoencoder that uses a KAN as encoder and a linear map as decoder. It generalizes classical PCA by replacing linear projections with learned B-spline functions on each edge. The motivation is to capture more variance than classical PCA, which becomes inefficient during market crises when the linear assumption breaks down and correlations between assets change dramatically. We prove that if the spline activations are forced to be linear, KAN-PCA yields exactly the same results as classical PCA, establishing PCA as a special case. Experiments on 20 S&P 500 stocks (2015-2024) show that KAN-PCA achieves a reconstruction R^2 of 66.57%, compared to 62.99% for classical PCA with the same 3 factors, while matching PCA out-of-sample after correcting for data leakage in the training procedure.

The domain of metaheuristic optimization has become vibrant due to a flood of new algorithms using a new nature-inspired metaphor but lacking clear methodological novelty. The Criticism behind the development of these algorithms has reached such an extent that the critics started to assert that all novel algorithms are only copies of already developed ones. In this study, we try to show that the situation is not so black and white. Therefore, we define a strong equivalence theorem for estimating the similarity between two nature-inspired metaheuristics, according to which two algorithms are equivalent if, and only if, the cosine similarity of their phenotypic and genotypic feature vectors, characterizing their behavior by searching for the optimal solutions, is above some threshold. On the theorem basis, a framework is developed for identifying the equivalence between nature-inspired metaheuristics. Extensive experimental work using the framework has shown that searching for conditions to achieve the high similarity of the more well-known nature-inspired metaheuristics is hard, or even not possible to achieve, in the limited computational environments.

Orthogonalized-update optimizers such as Muon improve training of matrix-valued parameters, but existing extensions mostly act either after orthogonalization by rescaling updates or before it with heavier whitening-based preconditioners. We introduce {\method}, a lightweight family of pre-orthogonalization equilibration schemes for Muon in three forms: two-sided row/column normalization (RC), row normalization (R), and column normalization (C). These variants rebalance the momentum matrix before finite-step Newton--Schulz using row/column squared-norm statistics and only $\mathcal{O}(m+n)$ auxiliary state. We show that finite-step orthogonalization is governed by input spectral properties, especially stable rank and condition number, and that row/column normalization is a zeroth-order whitening surrogate that removes marginal scale mismatch. For the hidden matrix weights targeted by {\method}, the row-normalized variant R is the natural default and preserves the $\widetilde{\mathcal{O}}(T^{-1/4})$ stationarity guarantee of Muon-type methods. In LLaMA2 pretraining on C4, the default R variant consistently outperforms Muon on 130M and 350M models, yielding faster convergence and lower validation perplexity.

Time series forecasting is vital across many domains, yet existing models struggle with fixed-length inputs and inadequate multi-scale modeling. We propose MR-CDM, a framework combining hierarchical multi-resolution trend decomposition, an adaptive embedding mechanism for variable-length inputs, and a multi-scale conditional diffusion process. Evaluations on four real-world datasets demonstrate that MR-CDM significantly outperforms state-of-the-art baselines (e.g., CSDI, Informer), reducing MAE and RMSE by approximately 6-10 to a certain degree.

Drivers' visual attention provides critical cues for anticipating latent hazards and directly shapes decision-making and control maneuvers, where its absence can compromise traffic safety. To emulate drivers' perception patterns and advance visual attention prediction for intelligent vehicles, we propose DiffAttn, a diffusion-based framework that formulates this task as a conditional diffusion-denoising process, enabling more accurate modeling of drivers' attention. To capture both local and global scene features, we adopt Swin Transformer as encoder and design a decoder that combines a Feature Fusion Pyramid for cross-layer interaction with dense, multi-scale conditional diffusion to jointly enhance denoising learning and model fine-grained local and global scene contexts. Additionally, a large language model (LLM) layer is incorporated to enhance top-down semantic reasoning and improve sensitivity to safety-critical cues. Extensive experiments on four public datasets demonstrate that DiffAttn achieves state-of-the-art (SoTA) performance, surpassing most video-based, top-down-feature-driven, and LLM-enhanced baselines. Our framework further supports interpretable driver-centric scene understanding and has the potential to improve in-cabin human-machine interaction, risk perception, and drivers' state measurement in intelligent vehicles.

Single-shot neural decoders commit to answers without iterative refinement, while chain-of-thought methods introduce discrete intermediate steps but lack a scalar measure of reasoning progress. We propose Energy-Based Reasoning via Structured Latent Planning (EBRM), which models reasoning as gradient-based optimization of a multi-step latent trajectory $z_{1:T}$ under a learned energy function $E(h_x, z)$. The energy decomposes into per-step compatibility, transition consistency, and trajectory smoothness terms. Training combines supervised encoder-decoder learning with contrastive energy shaping using hard negatives, while inference performs gradient descent or Langevin dynamics over $z$ and decodes from $z_T$. We identify a critical failure mode: on CNF logic satisfaction, latent planning reduces accuracy from $\approx 95\%$ to $\approx 56\%$. This degradation arises from a distribution mismatch, where the decoder is trained on encoder outputs $h_x$ but evaluated on planner outputs $z_T$ that drift into unseen latent regions. We analyze this behavior through per-step decoding, latent drift tracking, and gradient decomposition. To address it, we propose dual-path decoder training and latent anchoring. We further introduce a six-part ablation protocol covering component contributions, trajectory length, planner dynamics, initialization, decoder training distribution, and anchor weight. Experiments on three synthetic tasks show that energy decreases monotonically and induces structured latent trajectories on graph and logic tasks, while remaining flat on arithmetic ($r = 0.073$), indicating a negative result. Code is available at https://github.com/dkjo8/ebr-via-structured-latent-planning.

Block-based programming environments like Scratch have become widely adopted in Computer Science Education, but the mouse-based drag-and-drop interface can challenge users with disabilities. While prior work has provided solutions supporting children with visual impairment, these solutions tend to focus on making content perceivable and do not address the physical interaction barriers faced by users with motor disabilities. To bridge this gap, we introduce MeowCrophone, an approach that uses voice control to allow editing code in Scratch. MeowCrophone supports clicking elements, placing blocks, and navigating the workspace via a multi-modal voice user interface that uses numerical overlays and label reading to bypass physical input entirely. As imperfect speech recognition is common in classrooms and for children with dysarthria, MeowCrophone employs a multi-stage matching pipeline using regular expressions, phonetic matching, and a custom grammar. Evaluation shows that while free speech recognition systems achieved a baseline success rate of only 46.4%, MeowCrophone's pipeline improved results to 82.8% overall, with simple commands reaching 96.9% accuracy. This demonstrates that robust voice control can make Scratch accessible to users for whom visual aids are insufficient.

In-network computing techniques, exemplified by NVLink Sharp (NVLS), offer a promising approach to addressing the communication bottlenecks in LLM inference by offloading collective operations, such as All-Reduce, to switches. However, the accelerator-centric architecture of NVLS suffers from two fundamental limitations: 1) it relies on GPU load instructions to trigger reduction operations, which means that the data reduced in the switch must be additionally transferred back to the initiating GPU rather than being broadcast directly, thereby introducing unnecessary communication overhead; 2) due to its architectural constraints, NVLS cannot offload operators that are not decomposable into memory-semantic instructions, such as the in-network quantization (INQ) proposed in this work. As a result, All-Reduce in NVLS must operate at FP16/BF16 precision, leading to substantial bandwidth waste.To address these limitations, we propose SCIN, the first switch-centric in-network architecture for shared-memory networks of AI accelerators, enabling both low-latency and high-bandwidth All-Reduce. Specifically, we introduce an in-switch accelerator (ISA) capable of initiating memory-semantic operations for in-network processing, together with a co-designed communication fabric that incurs negligible protocol overhead. By eliminating redundant data movement, SCIN delivers lower All-Reduce latency than NVLS. Moreover, by integrating a quantization module into the ISA, SCIN enables INQ for All-Reduce, reducing its precision to 8 bits and nearly doubling bandwidth with negligible accuracy loss. We also present a prototype of SCIN on a multi-FPGA system to demonstrate its feasibility and effectiveness. Experimental results show that our design accelerates All-Reduce by up to 8.7x for small messages and 3.8x for large messages, leading up to 1.74x faster TTFT and 1.34x faster TPOT on LLaMA-2 models.

Accurate and efficient perception is essential for autonomous driving, where segmentation tasks such as drivable-area and lane segmentation provide critical cues for motion planning and control. However, achieving high segmentation accuracy while maintaining real-time performance on low-cost hardware remains a challenging problem. To address this issue, we introduce TwinMixing, a lightweight multi-task segmentation model designed explicitly for drivable-area and lane segmentation. The proposed network features a shared encoder and task-specific decoders, enabling both feature sharing and task specialization. Within the encoder, we propose an Efficient Pyramid Mixing (EPM) module that enhances multi-scale feature extraction through a combination of grouped convolutions, depthwise dilated convolutions and channel shuffle operations, effectively expanding the receptive field while minimizing computational cost. Each decoder adopts a Dual-Branch Upsampling (DBU) Block composed of a learnable transposed convolution-based Fine detailed branch and a parameter-free bilinear interpolation-based Coarse grained branch, achieving detailed yet spatially consistent feature reconstruction. Extensive experiments on the BDD100K dataset validate the effectiveness of TwinMixing across three configurations - tiny, base, and large. Among them, the base configuration achieves the best trade-off between accuracy and computational efficiency, reaching 92.0% mIoU for drivable-area segmentation and 32.3% IoU for lane segmentation with only 0.43M parameters and 3.95 GFLOPs. Moreover, TwinMixing consistently outperforms existing segmentation models on the same tasks, as illustrated in Fig. 1. Thanks to its compact and modular design, TwinMixing demonstrates strong potential for real-time deployment in autonomous driving and embedded perception systems. The source code: https://github.com/Jun0se7en/TwinMixing.

Bridges are critical components of national infrastructure and smart cities. Therefore, smart bridge monitoring is essential for ensuring public safety and preventing catastrophic failures or accidents. Traditional bridge monitoring methods rely heavily on human visual inspections, which are time-consuming and prone to subjectivity and error. This paper proposes an artificial intelligence (AI)-driven anomaly detection approach for smart bridge monitoring. Specifically, a simple machine learning (ML) model is developed using real-time sensor data collected by the iBridge sensor devices installed on a bridge in Norway. The proposed model is evaluated against different ML models. Experimental results demonstrate that the density-based spatial clustering of applications with noise (DBSCAN)-based model outperforms in accurately detecting the anomalous events (bridge accident). These findings indicate that the proposed model is well-suited for smart bridge monitoring and can enhance public safety by enabling the timely detection of unforeseen incidents.

Transformers for language modeling usually rely on deterministic internal computation, with uncertainty expressed mainly at the output layer. We introduce variational neurons into Transformer feed-forward computation so that uncertainty becomes part of the internal computation itself. Concretely, we replace deterministic feed-forward units with local variational units based on EVE while preserving the overall Transformer backbone. We evaluate this design in compact next-token language-modeling settings. We compare deterministic and variational variants with both predictive and probabilistic criteria. Alongside negative log-likelihood, perplexity and accuracy, we analyze calibration, conditional variance, mutual information and latent-usage statistics. The resulting picture is clear. Variational neurons integrate stably into Transformers, preserve strong predictive performance and produce informative uncertainty signals. The experiments also show that task quality, useful depth and internal stability are distinct properties. These results establish variational Transformers as a practical form of uncertainty-aware language modeling. They show that Transformers can predict with an explicit internal structure of uncertainty, which supports stronger probabilistic evaluation and a more informative analysis of model behavior.

Decarbonization in buildings calls for advanced control strategies that coordinate on-site renewables, grid electricity, and thermal demand. Literature approaches typically rely on demand side management strategies or on active energy storage, like batteries. However, the first solution often neglects carbon-aware objectives, and could lead to grid overload issues, while batteries entail environmental, end-of-life, and cost concerns. To overcome these limitations, we propose an optimal, carbon-aware optimization strategy that exploits the building's thermal mass as a passive storage, avoiding dedicated batteries. Specifically, when a surplus of renewable energy is available, our strategy computes the optimal share of surplus to store by temporarily adjusting the indoor temperature setpoint within comfort bounds. Thus, by explicitly accounting for forecasts of building energy consumption, solar production, and time-varying grid carbon intensity, our strategy enables emissions-aware load shifting while maintaining comfort. We evaluate the approach by simulating three TRNSYS models of the same system with different thermal mass. In all cases, the results show consistent reductions in grid electricity consumption with respect to a baseline that does not leverage surplus renewable generation. These findings highlight the potential of thermal-mass-based control for building decarbonization.

The design of Large Language Models and generative artificial intelligence has been shown to be "unfair" to less-spoken languages and to deepen the digital language divide. Critical sociolinguistic work has also argued that these technologies are not only made possible by prior socio-historical processes of linguistic standardisation, often grounded in European nationalist and colonial projects, but also exacerbate epistemologies of language as "monolithic, monolingual, syntactically standardized systems of meaning". In our paper, we draw on earlier work on the intersections of technology and language policy and bring our respective expertise in critical sociolinguistics and computational linguistics to bear on an interrogation of these arguments. We take two different complexes of non-standard linguistic varieties in our respective repertoires--South Tyrolean dialects, which are widely used in informal communication in South Tyrol, Italy, as well as varieties of Kurdish--as starting points to an interdisciplinary exploration of the intersections between GenAI and linguistic variation and standardisation. We discuss both how LLMs can be made to deal with nonstandard language from a technical perspective, and whether, when or how this can contribute to "democratic and decolonial digital and machine learning strategies", which has direct policy implications.

Aspect-Based Sentiment Analysis (ABSA) is fundamentally challenged by representation entanglement, where aspect semantics and sentiment polarities are often conflated in real-valued embedding spaces. Furthermore, standard contrastive learning suffers from false-negative collisions, severely degrading performance on high-frequency aspects. In this paper, we propose a novel framework featuring a Zero-Initialized Residual Complex Projection (ZRCP) and an Anti-collision Masked Angle Loss,inspired by quantum projection and entanglement ideas. Our approach projects textual features into a complex semantic space, systematically utilizing the phase to disentangle sentiment polarities while allowing the amplitude to encode the semantic intensity and lexical richness of subjective descriptions. To tackle the collision bottleneck, we introduce an anti-collision mask that elegantly preserves intra-polarity aspect cohesion while expanding the inter-polarity discriminative margin by over 50%. Experimental results demonstrate that our framework achieves a state-of-the-art Macro-F1 score of 0.8851. Deep geometric analyses further reveal that explicitly penalizing the complex amplitude catastrophically over-regularizes subjective representations, proving that our unconstrained-amplitude and phase-driven objective is crucial for robust, fine-grained sentiment disentanglement.

Reinforcement learning from verifiable rewards (RLVR) has significantly advanced the reasoning capabilities of large language models. However, standard Group Relative Policy Optimization (GRPO) typically assigns a uniform, sequence-level advantage to all tokens, thereby overlooking the intrinsic information heterogeneity along reasoning chains. We show that this coarse-grained credit assignment leads to premature entropy collapse and encourages the model to generate redundant, low-quality reasoning paths. Through systematic empirical analysis, we identify Critical Decision Pivots (CDPs): transient high-entropy states where the policy's trajectory is most sensitive to perturbations. These pivots represent the "forks in the road" where effective multi-path exploration is most crucial yet often suppressed by uniform advantage signals. Building on these insights, we propose Entropy-Regulated Policy Optimization (ERPO), which transitions the optimization focus from coarse sequences to fine-grained token dynamics. ERPO introduces three synergistic components: (i) Entropy-aware Gating, which adaptively amplifies exploration at CDPs to facilitate diverse path discovery; (ii) Bucket-based Implicit Normalization, which mitigates difficulty bias by aligning token progress windows; and (iii) Result-anchored Advantage Synthesis, which re-weights token-level signals via outcome-driven anchors. Extensive experiments on competitive mathematical benchmarks (e.g., MATH, AIME) demonstrate that ERPO significantly outperforms GRPO. Notably, ERPO not only boosts reasoning accuracy but also yields significantly more concise and robust derivation paths, establishing a new efficiency-accuracy frontier for large reasoning models.

With the rise of renewable energy sources and their high variability in generation, the management of power grids becomes increasingly complex and computationally demanding. Conventional AC-power-flow simulations, which use the Newton-Raphson (NR) method, suffer from poor scalability, making them impractical for emerging use cases such as joint transmission-distribution modeling and global grid analysis. At the same time, purely data-driven surrogate models lack physical guarantees and may violate fundamental constraints. In this work, we propose Differentiable Power-Flow (DPF), a reformulation of the AC power-flow problem as a differentiable simulation. DPF enables end-to-end gradient propagation from the physical power mismatches to the underlying simulation parameters, thereby allowing these parameters to be identified efficiently using gradient-based optimization. We demonstrate that DPF provides a scalable alternative to NR by leveraging GPU acceleration, sparse tensor representations, and batching capabilities available in modern machine-learning frameworks such as PyTorch. DPF is especially suited as a tool for time-series analyses due to its efficient reuse of previous solutions, for N-1 contingency-analyses due to its ability to process cases in batches, and as a screening tool by leveraging its speed and early stopping capability. The code is available in the authors' code repository.

We revisit the standard perturbation-based approach of Abernethy et al. (2008) in the context of unconstrained Bandit Linear Optimization (uBLO). We show the surprising result that in the unconstrained setting, this approach effectively reduces Bandit Linear Optimization (BLO) to a standard Online Linear Optimization (OLO) problem. Our framework improves on prior work in several ways. First, we derive expected-regret guarantees when our perturbation scheme is combined with comparator-adaptive OLO algorithms, leading to new insights about the impact of different adversarial models on the resulting comparator-adaptive rates. We also extend our analysis to dynamic regret, obtaining the optimal $\sqrt{P_T}$ path-length dependencies without prior knowledge of $P_T$. We then develop the first high-probability guarantees for both static and dynamic regret in uBLO. Finally, we discuss lower bounds on the static regret, and prove the folklore $Ω(\sqrt{dT})$ rate for adversarial linear bandits on the unit Euclidean ball, which is of independent interest.

Static regret to a single expert is often the wrong target for strictly online prediction under non-stationarity, where the best expert may switch repeatedly over time. We study Policy-Controlled Generalized Share (PCGS), a general strictly online framework in which the generalized-share recursion is fixed while the post-loss update controls are allowed to vary adaptively. Its principal instantiation in this paper is PCGS-TF, which uses a causal Transformer as an update controller: after round t finishes and the loss vector is observed, the Transformer outputs the controls that map w_t to w_{t+1} without altering the already committed decision w_t. Under admissible post-loss update controls, we obtain a pathwise weighted regret guarantee for general time-varying learning rates, and a standard dynamic-regret guarantee against any expert path with at most S switches under the constant-learning-rate specialization. Empirically, on a controlled synthetic suite with exact dynamic-programming switching-oracle evaluation, PCGS-TF attains the lowest mean dynamic regret in all seven non-stationary families, with its advantage increasing for larger expert pools. On a reproduced household-electricity benchmark, PCGS-TF also achieves the lowest normalized dynamic regret for S = 5, 10, and 20.

Pluralistic alignment is essential for adapting large language models (LLMs) to the diverse preferences of individuals and minority groups. However, existing approaches often mix stable personal traits with episode-specific factors, limiting their ability to generalize across episodes. To address this challenge, we introduce EpiPersona, a framework for explicit persona-episode coupling. EpiPersona first projects noisy preference feedback into a low-dimensional persona space, where similar personas are aggregated into shared discrete codes. This process separates enduring personal characteristics from situational signals without relying on predefined preference dimensions. The inferred persona representation is then coupled with the current episode, enabling episode-aware preference prediction. Extensive experiments show that EpiPersona consistently outperforms the baselines. It achieves notable performance gains in hard episodic-shift scenarios, while remaining effective with sparse preference data.

As AI becomes embedded in customer-facing systems, ethical scrutiny has largely focused on models, data, and governance. Far less attention has been paid to how AI is experienced through user-facing design. This commentary argues that many AI front-ends implicitly assume an 'ideal user body and mind', and that this becomes visible and ethically consequential when examined through the experiences of differently abled users. We explore this through retail AI front-ends for customer engagement - i.e., virtual assistants, virtual try-on systems, and hyper-personalised recommendations. Despite intuitive and inclusive framing, these systems embed interaction assumptions that marginalise users with vision, hearing, motor, cognitive, speech and sensory differences, as well as age-related variation in digital literacy and interaction norms. Drawing on practice-led insights, we argue that these failures persist not primarily due to technical limits, but due to the commercial, organisational, and procurement contexts in which AI front-ends are designed and deployed, where accessibility is rarely contractual. We propose front-end assurance as a practical complement to AI governance, aligning claims of intelligence and multimodality with the diversity of real users.

The clinical burden of spleen-stomach disorders is substantial. While large language models (LLMs) offer new potential for medical applications, they face three major challenges in the context of integrative Chinese and Western medicine (ICWM): a lack of high-quality data, the absence of models capable of effectively integrating the reasoning logic of traditional Chinese medicine (TCM) syndrome differentiation with that of Western medical (WM) disease diagnosis, and the shortage of a standardized evaluation benchmark. To address these interrelated challenges, we propose DongYuan, an ICWM spleen-stomach diagnostic framework. Specifically, three ICWM datasets (SSDF-Syndrome, SSDF-Dialogue, and SSDF-PD) were curated to fill the gap in high-quality data for spleen-stomach disorders. We then developed SSDF-Core, a core diagnostic LLM that acquires robust ICWM reasoning capabilities through a two-stage training regimen of supervised fine-tuning. tuning (SFT) and direct preference optimization (DPO), and complemented it with SSDF-Navigator, a pluggable consultation navigation model designed to optimize clinical inquiry strategies. Additionally, we established SSDF-Bench, a comprehensive evaluation benchmark focused on ICWM diagnosis of spleen-stomach disorders. Experimental results demonstrate that SSDF-Core significantly outperforms 12 mainstream baselines on SSDF-Bench. DongYuan lays a solid methodological foundation and provides practical technical references for the future development of intelligent ICWM diagnostic systems.

Adders are fundamental building blocks in modern digital systems, and their performance, power, and area (PPA) directly impact system efficiency. Contemporary adders typically use parallel-prefix architectures with established PPA trade-offs, but these often fail to deliver globally optimal PPA for specific design goals. Prior work lacks netlist-/cell-level awareness, and general synthesis heuristics are not adder-specific, resulting in suboptimal PPA. To address this, we propose AXON, an automated netlist optimization framework for adders. It performs design space exploration from architectural to netlist level, integrating prefix topology search with standard-cell-aware mapping via a hierarchical approach to quickly converge to near-optimal PPA solutions. We also introduce a hybrid ultra-high-speed adder combining parallel-prefix and Ling architectures to shorten the critical path. Experiments on TSMC 28nm library show AXON improves delay, area-delay product, and energy-delay product by up to 10.3%, 12.6%, and 32.1% respectively, compared to commercial synthesis tools.

Multimodal Large Language Models have demonstrated powerful cross-modal understanding and reasoning capabilities in general domains. However, in the electromagnetic (EM) domain, they still face challenges such as data scarcity and insufficient integration of domain knowledge. This paper proposes PReD, the first foundation model for the EM domain that covers the intelligent closed-loop of "perception, recognition, decision-making." We constructed a high-quality multitask EM dataset, PReD-1.3M, and an evaluation benchmark, PReD-Bench. The dataset encompasses multi-perspective representations such as raw time-domain waveform, frequency-domain spectrograms, and constellation diagrams, covering typical features of communication and radar signals. It supports a range of core tasks, including signal detection, modulation recognition, parameter estimation, protocol recognition, radio frequency fingerprint recognition, and anti-jamming decision-making. PReD adopts a multi-stage training strategy that unifies multiple tasks for EM signals. It achieves closed-loop optimization from end-to-end signal understanding to language-driven reasoning and decision-making, significantly enhancing EM domain expertise while maintaining general multimodal capabilities. Experimental results show that PReD achieves state-of-the-art performance on PReD-Bench constructed from both open-source and self-collected signal datasets. These results collectively validate the feasibility and potential of vision-aligned foundation models in advancing the understanding and reasoning of EM signals.

Data-driven weather models have advanced global medium-range forecasting, yet high-resolution regional prediction remains challenging due to unresolved multiscale interactions between large-scale dynamics and small-scale processes such as terrain-induced circulations and coastal effects. This paper presents a global-regional coupling framework for kilometer-scale regional weather forecasting that synergistically couples a pretrained Transformer-based global model with a high-resolution regional network via a novel bidirectional coupling module, ScaleMixer. ScaleMixer dynamically identifies meteorologically critical regions through adaptive key-position sampling and enables cross-scale feature interaction through dedicated attention mechanisms. The framework produces forecasts at $0.05^\circ$ ($\sim 5 \mathrm{km}$ ) and 1-hour resolution over China, significantly outperforming operational NWP and AI baselines on both gridded reanalysis data and real-time weather station observations. It exhibits exceptional skill in capturing fine-grained phenomena such as orographic wind patterns and Foehn warming, demonstrating effective global-scale coherence with high-resolution fidelity. The code is available at https://anonymous.4open.science/r/ScaleMixer-6B66.

This study presents a fully automated methodology for early prediction studies in clinical settings, leveraging information extracted from unstructured discharge reports. The proposed pipeline uses discharge reports to support the three main steps of early prediction: cohort selection, dataset generation, and outcome labeling. By processing discharge reports with natural language processing techniques, we can efficiently identify relevant patient cohorts, enrich structured datasets with additional clinical variables, and generate high-quality labels without manual intervention. This approach addresses the frequent issue of missing or incomplete data in codified electronic health records (EHR), capturing clinically relevant information that is often underrepresented. We evaluate the methodology in the context of predicting atrial fibrillation (AF) progression, showing that predictive models trained on datasets enriched with discharge report information achieve higher accuracy and correlation with true outcomes compared to models trained solely on structured EHR data, while also surpassing traditional clinical scores. These results demonstrate that automating the integration of unstructured clinical text can streamline early prediction studies, improve data quality, and enhance the reliability of predictive models for clinical decision-making.

Equipping LLM agents with real-world tools can substantially improve productivity. However, granting agents autonomy over tool use also transfers the associated privileges to both the agent and the underlying LLM. Improper privilege usage may lead to serious consequences, including information leakage and infrastructure damage. While several benchmarks have been built to study agents' security, they often rely on pre-coded tools and restricted interaction patterns. Such crafted environments differ substantially from the real-world, making it hard to assess agents' security capabilities in critical privilege control and usage. Therefore, we propose GrantBox, a security evaluation sandbox for analyzing agent privilege usage. GrantBox automatically integrates real-world tools and allows LLM agents to invoke genuine privileges, enabling the evaluation of privilege usage under prompt injection attacks. Our results indicate that while LLMs exhibit basic security awareness and can block some direct attacks, they remain vulnerable to more sophisticated attacks, resulting in an average attack success rate of 84.80% in carefully crafted scenarios.

App store reviews provide a constant flow of real user feedback that can help improve software requirements. However, these reviews are often messy, informal, and difficult to analyze manually at scale. Although automated techniques exist, many do not perform well when replicated and often fail to produce clean, backlog-ready user stories for agile projects. In this study, we evaluate how well large language models (LLMs) such as GPT-3.5 Turbo, Gemini 2.0 Flash, and Mistral 7B Instruct can generate usable user stories directly from raw app reviews. Using the Mini-BAR dataset of 1,000+ health app reviews, we tested zero-shot, one-shot, and two-shot prompting methods. We evaluated the generated user stories using both human judgment (via the RUST framework) and a RoBERTa classifier fine-tuned on UStAI to assess their overall quality. Our results show that LLMs can match or even outperform humans in writing fluent, well-formatted user stories, especially when few-shot prompts are used. However, they still struggle to produce independent and unique user stories, which are essential for building a strong agile backlog. Overall, our findings show how LLMs can reliably turn unstructured app reviews into actionable software requirements, providing developers with clear guidance to turn user feedback into meaningful improvements.

Graphs with diverse structural characteristics play a central role in modelling and optimization tasks. The ability to generate different types of graphs that exhibit shared properties is likewise essential for algorithm selection and configuration. However, constructing graphs that preserve high-level properties across a broad range of graph classes remains a challenging problem. We present a novel evolutionary approach to evolve graphs based on the Laplacian graph spectra descriptor. This descriptor can be used as part of a fitness function to evaluate graphs according to their desired high-level properties. Our evolutionary algorithm evolves graphs towards this descriptor in order to obtain graphs having properties that are consistent with it but are different from each other in terms of non-spectral graph metrics, such as path length, clustering coefficient and betweenness centrality. Our experimental results show that our approach is successful for different classes of graphs and a wide range of Laplacian graph spectra.

Domain Generalized Semantic Segmentation (DGSS) aims to maintain robust performance across unseen target domains. Vision Foundation Models (VFMs) offer rich multi-domain knowledge that can enhance generalization. However, strategies for actively exploiting the rich subspace structures within VFMs remain under-explored, with many existing methods focusing primarily on preserving pre-trained knowledge. Furthermore, their LoRA components often suffer from limited representational diversity and inefficient parameter utilization. We propose RecycleLoRA, which addresses both challenges by employing Rank-Revealing QR Decomposition (RRQR) to systematically exploit VFM's subspace structures and enhance LoRA's representational richness. Our main adapter leverages minor subspace directions identified by RRQR to learn diverse and independent features, achieving competitive performance even when used alone. We further introduce a sub adapter that carefully refines major directions with minimal adjustments, providing complementary improvements to the main adapter's strong baseline performance. This design enables the dual adapters to learn distinct representations without requiring additional regularization losses. Our systematic exploitation of pre-trained subspace structures through RRQR-based initialization leads to superior domain generalization performance. RecycleLoRA achieves state-of-the-art performance on both synthetic-to-real generalization and real-to-real generalization tasks without complex architectures or additional inference latency.

In this paper, we investigate the capabilities of in-air 3D SONAR sensors for the monitoring of road surface conditions. Concretely, we consider two applications: Road material classification and Road damage detection and classification. While such tasks can be performed with other sensor modalities, such as camera sensors and LiDAR sensors, these sensor modalities tend to fail in harsh sensing conditions, such as heavy rain, smoke or fog. By using a sensing modality that is robust to such interference, we enable the creation of opportunistic sensing applications, where vehicles performing other tasks (garbage collection, mail delivery, etc.) can also be used to monitor the condition of the road. For these tasks, we use a single dataset, in which different types of damages are annotated, with labels including the material of the road surface. In the material classification task, we differentiate between three different road materials: Asphalt, Concrete and Element roads. In the damage detection and classification task, we determine if there is damage, and what type of damage (independent of material type), without localizing the damage. We are succesful in determining the road surface type from SONAR sensor data, with F1 scores approaching 90% on the test set, but find that for the detection of damages performace lags, with F1 score around 75%. From this, we conclude that SONAR sensing is a promising modality to include in opportunistic sensing-based pavement management systems, but that further research is needed to reach the desired accuracy.

Recent test-time reasoning methods improve performance by generating more candidate chains or searching over larger reasoning trees, but they typically lack explicit control over when to expand, what to prune, how to repair, and when to abstain. We introduce CoT2-Meta, a training-free metacognitive reasoning framework that combines object-level chain-of-thought generation with meta-level control over partial reasoning trajectories. The framework integrates four components: strategy-conditioned thought generation, tree-structured search, an online process oracle for step-level reasoning evaluation, and a meta-controller that allocates computation through expansion, pruning, repair, stopping, and fallback decisions. Under matched inference budgets, CoT2-Meta consistently outperforms strong single-path, sampling-based, and search-based baselines, including ReST-MCTS. On the default backbone, it achieves 92.8 EM on MATH, 90.4 accuracy on GPQA, 98.65 EM on GSM8K, 75.8 accuracy on BBEH, 85.6 accuracy on MMMU-Pro, and 48.8 accuracy on HLE, with gains over the strongest non-CoT2-Meta baseline of +3.6, +5.2, +1.15, +2.0, +4.3, and +4.3 points, respectively. Beyond these core results, the framework remains effective across a broader 15-benchmark suite spanning knowledge and QA, multi-hop reasoning, coding, and out-of-distribution evaluation. Additional analyses show better compute scaling, improved calibration, stronger selective prediction, targeted repair behavior, and consistent gains across backbone families. These results suggest that explicit metacognitive control is a practical design principle for reliable and compute-efficient test-time reasoning systems.

We introduce Multilingual Document Parsing Benchmark, the first benchmark for multilingual digital and photographed document parsing. Document parsing has made remarkable strides, yet almost exclusively on clean, digital, well-formatted pages in a handful of dominant languages. No systematic benchmark exists to evaluate how models perform on digital and photographed documents across diverse scripts and low-resource languages. MDPBench comprises 3,400 document images spanning 17 languages, diverse scripts, and varied photographic conditions, with high-quality annotations produced through a rigorous pipeline of expert model labeling, manual correction, and human verification. To ensure fair comparison and prevent data leakage, we maintain separate public and private evaluation splits. Our comprehensive evaluation of both open-source and closed-source models uncovers a striking finding: while closed-source models (notably Gemini3-Pro) prove relatively robust, open-source alternatives suffer dramatic performance collapse, particularly on non-Latin scripts and real-world photographed documents, with an average drop of 17.8% on photographed documents and 14.0% on non-Latin scripts. These results reveal significant performance imbalances across languages and conditions, and point to concrete directions for building more inclusive, deployment-ready parsing systems. Source available at https://github.com/Yuliang-Liu/MultimodalOCR.

Although Graph Neural Networks (GNNs) have shown promise for smart contract vulnerability detection, they still face significant limitations. Homogeneous graph models fail to capture the interplay between control flow and data dependencies, while heterogeneous graph approaches often lack deep semantic understanding, leaving them susceptible to adversarial attacks. Moreover, most black-box models fail to provide explainable evidence, hindering trust in professional audits. To address these challenges, we propose ORACAL (Observable RAG-enhanced Analysis with CausAL reasoning), a heterogeneous multimodal graph learning framework that integrates Control Flow Graph (CFG), Data Flow Graph (DFG), and Call Graph (CG). ORACAL selectively enriches critical subgraphs with expert-level security context from Retrieval-Augmented Generation (RAG) and Large Language Models (LLMs), and employs a causal attention mechanism to disentangle true vulnerability indicators from spurious correlations. For transparency, the framework adopts PGExplainer to generate subgraph-level explanations identifying vulnerability triggering paths. Experiments on large-scale datasets demonstrate that ORACAL achieves state-of-the-art performance, outperforming MANDO-HGT, MTVHunter, GNN-SC, and SCVHunter by up to 39.6 percentage points, with a peak Macro F1 of 91.28% on the primary benchmark. ORACAL maintains strong generalization on out-of-distribution datasets with 91.8% on CGT Weakness and 77.1% on DAppScan. In explainability evaluation, PGExplainer achieves 32.51% Mean Intersection over Union (MIoU) against manually annotated vulnerability triggering paths. Under adversarial attacks, ORACAL limits performance degradation to approximately 2.35% F1 decrease with an Attack Success Rate (ASR) of only 3%, surpassing SCVHunter and MANDO-HGT which exhibit ASRs ranging from 10.91% to 18.73%.

Constitutional AI (CAI) aligns language models with explicitly stated normative principles, offering a transparent alternative to implicit alignment through human feedback alone. However, because constitutions are authored by specific groups of people, the resulting models may reflect particular cultural perspectives. We investigate this question by evaluating Anthropic's Claude Sonnet on 55 World Values Survey items, selected for high cross-cultural variance across six value domains and administered as both direct survey questions and naturalistic advice-seeking scenarios. Comparing Claude's responses to country-level data from 90 nations, we find that Claude's value profile most closely resembles those of Northern European and Anglophone countries, but on a majority of items extends beyond the range of all surveyed populations. When users provide cultural context, Claude adjusts its rhetorical framing but not its substantive value positions, with effect sizes indistinguishable from zero across all twelve tested countries. An ablation removing the system prompt increases refusals but does not alter the values expressed when responses are given, and replication on a smaller model (Claude Haiku) confirms the same cultural profile across model sizes. These findings suggest that when a constitution is authored within the same cultural tradition that dominates the training data, constitutional alignment may codify existing cultural biases rather than correct them--producing a value floor that surface-level interventions cannot meaningfully shift. We discuss the compounding nature of this risk and the need for globally representative constitution-authoring processes.

Large Language Models (LLMs) are now capable of resolving real-world GitHub issues. However, current approaches overapproximate the code context and suffer from two compounding problems: the prohibitive cost of processing massive inputs, and low effectiveness as noise floods the context window and distracts the model from the bug-fixing signal. Existing compression techniques fail to resolve this tension: generic compressors compromise the semantic integrity of code, while code-specific tools lack awareness of code structure and task context to preserve essential patch ingredients. To address this, we propose a novel framework consisting of two components. First, Oracle-guided Code Distillation (OCD), a context distillation algorithm that combines genetic search and delta debugging to systematically reduce code contexts to their minimal sufficient subsequence - retaining only the ingredients required for a successful fix. We use this distilled data to fine-tune SWEzze, a lightweight model that learns to compress code context at inference time, filtering noise and combating distraction while preserving fix ingredients. Evaluated on SWE-bench Verified across three frontier LLMs, SWEzze maintains a stable compression rate of about 6 times across models, reduces the total token budget by 51.8%-71.3% relative to the uncompressed setting, improves issue resolution rates by 5.0%-9.2%, and delivers the best overall balance among effectiveness, compression ratio, and latency compared with state-of-the-art context compression baselines.

Livestock growth prediction is essential for optimising farm management and improving the efficiency and sustainability of livestock production, yet it remains underexplored due to limited large-scale datasets and privacy concerns surrounding farm-level data. Existing biophysical models rely on fixed formulations, while most machine learning approaches are trained on small, isolated datasets, limiting their robustness and generalisability. To address these challenges, we propose LivestockFL, the first federated learning framework specifically designed for livestock growth prediction. LivestockFL enables collaborative model training across distributed farms without sharing raw data, thereby preserving data privacy while alleviating data sparsity, particularly for farms with limited historical records. The framework employs a neural architecture based on a Gated Recurrent Unit combined with a multilayer perceptron to model temporal growth patterns from historical weight records and auxiliary features. We further introduce LivestockPFL, a novel personalised federated learning framework that extends the above federated learning framework with a personalized prediction head trained on each farm's local data, producing farm-specific predictors. Experiments on a real-world dataset demonstrate the effectiveness and practicality of the proposed approaches.

Digital health research has advanced dynamic graph-based disease models, topological learning on simplicial complexes, and multimodal mixture-of-experts architectures, but these strands remain largely disconnected. We propose Graph Vector Field (GVF), a framework that models health risk as a vector-valued field on time-varying simplicial complexes, coupling discrete differential-geometric operators with modality-structured mixture-of-experts. Risk is represented as a vector-valued cochain whose evolution is parameterised with Hodge Laplacians and discrete exterior calculus operators, yielding a Helmholtz-Hodge decomposition into potential-driven (exact), circulation-like (coexact), and topologically constrained (harmonic) components linked to interpretable propagation, cyclic, and persistent risk mechanisms. Multimodal inputs from wearable sensors, behavioural/environmental context, and clinical/genomic data are incorporated through a bundle-structured mixture-of-experts in which modality-specific latent spaces are attached as fibres to the base complex. This separates modality-specific from shared contributions and offers a principled route toward modality-level identifiability. GVF integrates geometric dynamical systems, higher-order topology (enforced indirectly via geometric regularisation and Hodge decomposition), and structured multimodal fusion into a single framework for interpretable, modality-resolved risk modelling. This paper develops the mathematical foundations, architectural design, and formal guarantees; empirical validation is the subject of ongoing work.

Cross-attention is the primary interface through which text conditions latent diffusion models, yet its step-wise multi-resolution dynamics remain under-characterized, limiting principled training-free control. We cast diffusion cross-attention as a spatiotemporal signal on the latent grid by summarizing token-softmax weights into token-agnostic concentration maps and tracking their radially binned Fourier power over denoising. Across prompts and seeds, encoder cross-attention exhibits a consistent coarse-to-fine spectral progression, yielding a stable time-frequency fingerprint of token competition. Building on this structure, we introduce Attention Frequency Modulation (AFM), a plug-and-play inference-time intervention that edits token-wise pre-softmax cross-attention logits in the Fourier domain: low- and high-frequency bands are reweighted with a progress-aligned schedule and can be adaptively gated by token-allocation entropy, before the token softmax. AFM provides a continuous handle to bias the spatial scale of token-competition patterns without retraining, prompt editing, or parameter updates. Experiments on Stable Diffusion show that AFM reliably redistributes attention spectra and produces substantial visual edits while largely preserving semantic alignment. Finally, we find that entropy mainly acts as an adaptive gain on the same frequency-based edit rather than an independent control axis.

The global Lipschitz constant of a neural network governs both adversarial robustness and generalization. Conventional approaches to ``certified training" typically follow a train-then-verify paradigm: they train a network and then attempt to bound its Lipschitz constant. Because the efficient ``trivial bound" (the product of the layerwise Lipschitz constants) is exponentially loose for arbitrary networks, these approaches must rely on computationally expensive techniques such as semidefinite programming, mixed-integer programming, or branch-and-bound. We propose a different paradigm: rather than designing complex verifiers for arbitrary networks, we design networks to be verifiable by the fast trivial bound. We show that directly penalizing the trivial bound during training forces it to become tight, thereby effectively regularizing the true Lipschitz constant. To achieve this, we identify three structural obstructions to a tight trivial bound (dead neurons, bias terms, and ill-conditioned weights) and introduce architectural mitigations, including a novel notion of norm-saturating polyactivations and bias-free sinusoidal layers. Our approach avoids the runtime complexity of advanced verification while achieving strong results: we train robust networks on MNIST with Lipschitz bounds that are small (orders of magnitude lower than comparable works) and tight (within 10% of the ground truth). The experimental results validate the theoretical guarantees, support the proposed mechanisms, and extend empirically to diverse activations and non-Euclidean norms.

The digitisation of historical documents has traditionally been conceived as a process limited to character-level transcription, producing flat text that lacks the structural and semantic information necessary for substantive computational analysis. We present VERITAS (Vision-Enhanced Reading, Interpretation, and Transcription of Archival Sources), a modular, model-agnostic framework that reconceptualises digitisation as an integrated workflow encompassing transcription, layout analysis, and semantic enrichment. The pipeline is organised into four stages - Preprocessing, Extraction, Refinement, and Enrichment - and employs a schema-driven architecture that allows researchers to declaratively specify their extraction objectives. We evaluate VERITAS on the critical edition of Bernardino Corio's Storia di Milano, a Renaissance chronicle of over 1,600 pages. Results demonstrate that the pipeline achieves a 67.6% relative reduction in word error rate compared to a commercial OCR baseline, with a threefold reduction in end-to-end processing time when accounting for manual correction. We further illustrate the downstream utility of the pipeline's output by querying the transcribed corpus through a retrieval-augmented generation system, demonstrating its capacity to support historical inquiry.

Parliamentary proceedings represent a rich yet challenging resource for computational analysis, particularly when preserved only as scanned historical documents. Existing efforts to transcribe Italian parliamentary speeches have relied on traditional Optical Character Recognition pipelines, resulting in transcription errors and limited semantic annotation. In this paper, we propose a pipeline based on Vision-Language Models for the automatic transcription, semantic segmentation, and entity linking of Italian parliamentary speeches. The pipeline employs a specialised OCR model to extract text while preserving reading order, followed by a large-scale Vision-Language Model that performs transcription refinement, element classification, and speaker identification by jointly reasoning over visual layout and textual content. Extracted speakers are then linked to the Chamber of Deputies knowledge base through SPARQL queries and a multi-strategy fuzzy matching procedure. Evaluation against an established benchmark demonstrates substantial improvements both in transcription quality and speaker tagging.

Agentic Reinforcement Learning (RL) enables LLMs to solve complex tasks by alternating between a data-collection rollout phase and a policy training phase. During rollout, the agent generates trajectories, i.e., multi-step interactions between LLMs and external tools. Yet, frequent tool calls induce long-tailed trajectory generation that bottlenecks rollouts. This stems from step-centric designs that ignore trajectory context, triggering three system problems for long-tail trajectory generation: queueing delays, interference overhead, and inflated per-token time. We propose Heddle, a trajectory-centric system to optimize the when, where, and how of agentic rollout execution. Heddle integrates three core mechanisms: trajectory-level scheduling using runtime prediction and progressive priority to minimize cumulative queueing; trajectory-aware placement via presorted dynamic programming and opportunistic migration during idle tool call intervals to minimize interference; and trajectory-adaptive resource manager that dynamically tunes model parallelism to accelerate the per-token time of long-tail trajectories while maintaining high throughput for short trajectories. Evaluations across diverse agentic RL workloads demonstrate that Heddle effectively neutralizes the long-tail bottleneck, achieving up to 2.5$\times$ higher end-to-end rollout throughput compared to state-of-the-art baselines.

Dataset Condensation (DC) is a data-efficient learning paradigm that synthesizes small yet informative datasets, enabling models to match the performance of full-data training. However, recent work exposes a critical vulnerability of DC to backdoor attacks, where malicious patterns (\textit{e.g.}, triggers) are implanted into the condensation dataset, inducing targeted misclassification on specific inputs. Existing attacks always prioritize attack effectiveness and model utility, overlooking the crucial dimension of stealthiness. To bridge this gap, we propose InkDrop, which enhances the imperceptibility of malicious manipulation without degrading attack effectiveness and model utility. InkDrop leverages the inherent uncertainty near model decision boundaries, where minor input perturbations can induce semantic shifts, to construct a stealthy and effective backdoor attack. Specifically, InkDrop first selects candidate samples near the target decision boundary that exhibit latent semantic affinity to the target class. It then learns instance-dependent perturbations constrained by perceptual and spatial consistency, embedding targeted malicious behavior into the condensed dataset. Extensive experiments across diverse datasets validate the overall effectiveness of InkDrop, demonstrating its ability to integrate adversarial intent into condensed datasets while preserving model utility and minimizing detectability. Our code is available at https://github.com/lvdongyi/InkDrop.

Voice design from natural language aims to generate speaker timbres directly from free-form textual descriptions, allowing users to create voices tailored to specific roles, personalities, and emotions. Such controllable voice creation benefits a wide range of downstream applications-including storytelling, game dubbing, role-play agents, and conversational assistants, making it a significant task for modern Text-to-Speech models. However, existing models are largely trained on carefully recorded studio data, which produces speech that is clean and well-articulated, yet lacks the lived-in qualities of real human voices. To address these limitations, we present MOSS-VoiceGenerator, an open-source instruction-driven voice generation model that creates new timbres directly from natural language prompts. Motivated by the hypothesis that exposure to real-world acoustic variation produces more perceptually natural voices, we train on large-scale expressive speech data sourced from cinematic content. Subjective preference studies demonstrate its superiority in overall performance, instruction-following, and naturalness compared to other voice design models.

Generating coherent and communicative visual sequences, such as image sequences and videos, remains a significant challenge for current multimodal systems. Despite advances in visual quality and the integration of world knowledge, existing models still struggle to maintain logical flow, often resulting in disjointed actions, fragmented narratives, and unclear storylines. We attribute these issues to the lack of attention to visual logic, a critical yet underexplored dimension of visual sequence generation that we define as the perceptual and causal coherence among characters, actions, and scenes over time. To bridge this gap, we propose a logic-aware multi-image story visualization framework, LogiStory. The framework is built around the central innovation of explicitly modeling visual logic in story visualization. To realize this idea, we design a multi-agent system that grounds roles, extracts causal chains, and verifies story-level consistency, transforming narrative coherence from an implicit byproduct of image generation into an explicit modeling objective. This design effectively bridges structured story planning with visual generation, enhancing both narrative clarity and visual quality in story visualization. Furthermore, to evaluate the generation capacity, we construct LogicTale, a benchmark comprising richly annotated stories, emphasizing causal reasoning, and visual logic interpretability. We establish comprehensive automatic and human evaluation protocols designed to measure both visual logic and perceptual quality. Experiments demonstrate that our approach significantly improves the narrative logic of generated visual stories. This work provides a foundational step towards modeling and enforcing visual logic in general image sequence and video generation tasks.

We enhance optical network availability and reliability through a lightweight transformer model that predicts optical fiber amplifier lifetime from condition-based monitoring data, enabling real-time, edge-level predictive maintenance and advancing deployable AI for autonomous network operation.

Training reinforcement learning (RL) agents to control fluid dynamics systems is computationally expensive due to the high cost of direct numerical simulations (DNS) of the governing equations. Surrogate models offer a promising alternative by approximating the dynamics at a fraction of the computational cost, but their feasibility as training environments for RL is limited by distribution shifts, as policies induce state distributions not covered by the surrogate training data. In this work, we investigate the use of Linear Recurrent Autoencoder Networks (LRANs) for accelerating RL-based control of 2D Rayleigh-Bénard convection. We evaluate two training strategies: a surrogate trained on precomputed data generated with random actions, and a policy-aware surrogate trained iteratively using data collected from an evolving policy. Our results show that while surrogate-only training leads to reduced control performance, combining surrogates with DNS in a pretraining scheme recovers state-of-the-art performance while reducing training time by more than 40%. We demonstrate that policy-aware training mitigates the effects of distribution shift, enabling more accurate predictions in policy-relevant regions of the state space.

Reconstructing high-resolution turbulent flow fields from severely under-resolved observations is a fundamental inverse problem in computational fluid dynamics and scientific machine learning. Classical interpolation methods fail to recover missing fine-scale structures, while existing deep learning approaches rely on convolutional architectures that lack the spectral and multiscale inductive biases necessary for physically faithful reconstruction at large upscaling factors. We introduce the Spectrally-Informed Multi-Resolution Neural Operator (SIMR-NO), a hierarchical operator learning framework that factorizes the ill-posed inverse mapping across intermediate spatial resolutions, combines deterministic interpolation priors with spectrally gated Fourier residual corrections at each stage, and incorporates local refinement modules to recover fine-scale spatial features beyond the truncated Fourier basis. The proposed method is evaluated on Kolmogorov-forced two-dimensional turbulence, where $128\times128$ vorticity fields are reconstructed from extremely coarse $8\times8$ observations representing a $16\times$ downsampling factor. Across 201 independent test realizations, SIMR-NO achieves a mean relative $\ell_2$ error of $26.04\%$ with the lowest error variance among all methods, reducing reconstruction error by $31.7\%$ over FNO, $26.0\%$ over EDSR, and $9.3\%$ over LapSRN. Beyond pointwise accuracy, SIMR-NO is the only method that faithfully reproduces the ground-truth energy and enstrophy spectra across the full resolved wavenumber range, demonstrating physically consistent super-resolution of turbulent flow fields.

Grounding has become a fundamental capability of vision-language models (VLMs). Most existing VLMs point by generating coordinates as part of their text output, which requires learning a complicated coordinate system and results in a high token count. Instead, we propose a more intuitive pointing mechanism that directly selects the visual tokens that contain the target concept. Our model generates a special pointing token that cross-attends to the input image or video tokens and selects the appropriate one. To make this model more fine-grained, we follow these pointing tokens with an additional special token that selects a fine-grained subpatch within the initially selected region, and then a third token that specifies a location within that subpatch. We further show that performance improves by generating points sequentially in a consistent order, encoding the relative position of the previously selected point, and including a special no-more-points class when selecting visual tokens. Using this method, we set a new state-of-the-art on image pointing (70.7% on PointBench), set a new state-of-the-art among fully open models on GUI pointing (61.1% on ScreenSpotPro), and improve video pointing (59.1% human preference win rate vs. a text coordinate baseline) and tracking (+6.3% gain on Molmo2Track). We additionally show that our method achieves much higher sample efficiency and discuss the qualitative differences that emerge from this design change.

Digital testing has emerged as a key paradigm for the development and verification of autonomous maritime navigation systems, yet the availability of realistic and diverse safety-critical encounter scenarios remains limited. Existing approaches either rely on handcrafted templates, which lack realism, or extract cases directly from historical data, which cannot systematically expand rare high-risk situations. This paper proposes a data-driven framework that converts large-scale Automatic Identification System (AIS) trajectories into structured safety-critical encounter scenarios. The framework combines generative trajectory modeling with automated encounter pairing and temporal parameterization to enable scalable scenario construction while preserving real traffic characteristics. To enhance trajectory realism and robustness under noisy AIS observations, a multi-scale temporal variational autoencoder is introduced to capture vessel motion dynamics across different temporal resolutions. Experiments on real-world maritime traffic flows demonstrate that the proposed method improves trajectory fidelity and smoothness, maintains statistical consistency with observed data, and enables the generation of diverse safety-critical encounter scenarios beyond those directly recorded. The resulting framework provides a practical pathway for building scenario libraries to support digital testing, benchmarking, and safety assessment of autonomous navigation and intelligent maritime traffic management systems. Code is available at https://anonymous.4open.science/r/traj-gen-anonymous-review.

Generative agent simulations operate at two scales: individual personas for character interaction, and population models for collective behavior analysis and intervention testing. We propose a third scale: meso-level simulation - interaction with group-level representations that retain grounding in rich individual experience. To enable this, we present Synonymix, a pipeline that constructs a "unigraph" from multiple life story personas via graph-based abstraction and merging, producing a queryable collective representation that can be explored for sensemaking or sampled for synthetic persona generation. Evaluating synthetic agents on General Social Survey items, we demonstrate behavioral signal preservation beyond demographic baselines (p<0.001, r=0.59) with demonstrable privacy guarantee (max source contribution <13%). We invite discussion on interaction modalities enabled by meso-level simulations, and whether "high-fidelity" personas can ever capture the texture of lived experience.

This work presents a novel tensor network algorithm for solving Quadratic Unconstrained Binary Optimization (QUBO) problems, Quadratic Unconstrained Discrete Optimization (QUDO) problems, and Tensor Quadratic Unconstrained Discrete Optimization (T-QUDO) problems. The proposed algorithm is based on the MeLoCoToN methodology, which solves combinatorial optimization problems by employing superposition, imaginary time evolution, and projective measurements. Additionally, two different approaches are presented to solve QUBO and QUDO problems with k-neighbors interactions in a lineal chain, one based on 4-order tensor contraction and the other based on matrix-vector multiplication, including sparse computation and a new technique called "Waterfall". Furthermore, the performance of both implementations is compared with a quadratic optimization solver to demonstrate the performance of the method, showing advantages in several problem instances.

We prove that under five minimal axioms -- multi-dimensional quality, finite evaluation, effective optimization, resource finiteness, and combinatorial interaction -- any optimized AI agent will systematically under-invest effort in quality dimensions not covered by its evaluation system. This result establishes reward hacking as a structural equilibrium, not a correctable bug, and holds regardless of the specific alignment method (RLHF, DPO, Constitutional AI, or others) or evaluation architecture employed. Our framework instantiates the multi-task principal-agent model of Holmstrom and Milgrom (1991) in the AI alignment setting, but exploits a structural feature unique to AI systems -- the known, differentiable architecture of reward models -- to derive a computable distortion index that predicts both the direction and severity of hacking on each quality dimension prior to deployment. We further prove that the transition from closed reasoning to agentic systems causes evaluation coverage to decline toward zero as tool count grows -- because quality dimensions expand combinatorially while evaluation costs grow at most linearly per tool -- so that hacking severity increases structurally and without bound. Our results unify the explanation of sycophancy, length gaming, and specification gaming under a single theoretical structure and yield an actionable vulnerability assessment procedure. We further conjecture -- with partial formal analysis -- the existence of a capability threshold beyond which agents transition from gaming within the evaluation system (Goodhart regime) to actively degrading the evaluation system itself (Campbell regime), providing the first economic formalization of Bostrom's (2014) "treacherous turn."

While Large Language Models (LLMs) have demonstrated remarkable fluency in educational dialogues, most generative tutors primarily operate through intuitive, single-pass generation. This reliance on fast thinking precludes a dedicated reasoning workspace, forcing multiple diagnostic and strategic signals to be processed in a conflated manner. As a result, learner cognitive diagnosis, affective perception, and pedagogical decision-making become tightly entangled, which limits the tutoring system's capacity for deliberate instructional adaptation. We propose SLOW, a theory-informed tutoring framework that supports deliberate learner-state reasoning within a transparent decision workspace. Inspired by dual-process accounts of human tutoring, SLOW explicitly separates learner-state inference from instructional action selection. The framework integrates causal evidence parsing from learner language, fuzzy cognitive diagnosis with counterfactual stability analysis, and prospective affective reasoning to anticipate how instructional choices may influence learners' emotional trajectories. These signals are jointly considered to guide pedagogically and affectively aligned tutoring strategies. Evaluation using hybrid human-AI judgments demonstrates significant improvements in personalization, emotional sensitivity, and clarity. Ablation studies further confirm the necessity of each module, showcasing how SLOW enables interpretable and reliable intelligent tutoring through a visualized decision-making process. This work advances the interpretability and educational validity of LLM-based adaptive instruction.

Modern software systems heavily rely on various libraries, which require understanding the API semantics in static analysis. However, summarizing API semantics remains challenging due to complex implementations or unavailable library code. This paper presents DAInfer+, a novel approach for inferring API specifications from library documentation. We employ Natural Language Processing (NLP) to interpret informal semantic information provided by the documentation, which enables us to reduce the specification inference to an optimization problem. Specifically, we investigate the effectiveness of sentence embedding models and Large Language Models (LLMs) in deriving memory operation abstractions from API descriptions. These abstractions are used to retrieve data-flow and aliasing relations to generate comprehensive API specifications. To solve the optimization problem efficiently, we propose neurosymbolic optimization, yielding precise data-flow and aliasing specifications. Our evaluation of popular Java libraries shows that zero-shot sentence embedding models outperform few-shot prompted LLMs in robustness, capturing fine-grained semantic nuances more effectively. While our initial attempts using two-stage LLM prompting yielded promising results, we found that the embedding-based approach proved superior. Specifically, these models achieve over 82% recall and 85% precision for data-flow inference and 88% recall and 79% precision for alias relations, all within seconds. These results demonstrate the practical value of DAInfer+ in library-aware static analysis.

Deep learning (DL) models have achieved strong performance in an intelligence healthcare setting, yet most existing approaches operate as black boxes and ignore the physical processes that govern tumor growth, limiting interpretability, robustness, and clinical trust. To address this limitation, we propose PhysNet, a physics-embedded DL framework that integrates tumor growth dynamics directly into the feature learning process of a convolutional neural network (CNN). Unlike conventional physics-informed methods that impose physical constraints only at the output level, PhysNet embeds a reaction diffusion model of tumor growth within intermediate feature representations of a ResNet backbone. The architecture jointly performs multi-class tumor classification while learning a latent tumor density field, its temporal evolution, and biologically meaningful physical parameters, including tumor diffusion and growth rates, through end-to-end training. This design is necessary because purely data-driven models, even when highly accurate or ensemble-based, cannot guarantee physically consistent predictions or provide insight into tumor behavior. Experimental results on a large brain MRI dataset demonstrate that PhysNet outperforms multiple state-of-the-art DL baselines, including MobileNetV2, VGG16, VGG19, and ensemble models, achieving superior classification accuracy and F1-score. In addition to improved performance, PhysNet produces interpretable latent representations and learned bio-physical parameters that align with established medical knowledge, highlighting physics-embedded representation learning as a practical pathway toward more trustworthy and clinically meaningful medical AI systems.

In this paper, we introduce GhostWriteBench, a dataset for LLM authorship attribution. It comprises long-form texts (50K+ words per book) generated by frontier LLMs, and is designed to test generalisation across multiple out-of-distribution (OOD) dimensions, including domain and unseen LLM author. We also propose TRACE -- a novel fingerprinting method that is interpretable and lightweight -- that works for both open- and closed-source models. TRACE creates the fingerprint by capturing token-level transition patterns (e.g., word rank) estimated by another lightweight language model. Experiments on GhostWriteBench demonstrate that TRACE achieves state-of-the-art performance, remains robust in OOD settings, and works well in limited training data scenarios.

Preference-based reinforcement learning can learn effective reward functions from comparisons, but its scalability is constrained by the high cost of oracle feedback. Lightweight vision-language embedding (VLE) models provide a cheaper alternative, but their noisy outputs limit their effectiveness as standalone reward generators. To address this challenge, we propose ROVED, a hybrid framework that combines VLE-based supervision with targeted oracle feedback. Our method uses the VLE to generate segment-level preferences and defers to an oracle only for samples with high uncertainty, identified through a filtering mechanism. In addition, we introduce a parameter-efficient fine-tuning method that adapts the VLE with the obtained oracle feedback in order to improve the model over time in a synergistic fashion. This ensures the retention of the scalability of embeddings and the accuracy of oracles, while avoiding their inefficiencies. Across multiple robotic manipulation tasks, ROVED matches or surpasses prior preference-based methods while reducing oracle queries by up to 80%. Remarkably, the adapted VLE generalizes across tasks, yielding cumulative annotation savings of up to 90%, highlighting the practicality of combining scalable embeddings with precise oracle supervision for preference-based RL.

The performance of large language model (LLM) systems depends not only on model weights, but also on their harness: the code that determines what information to store, retrieve, and present to the model. Yet harnesses are still designed largely by hand, and existing text optimizers are poorly matched to this setting because they compress feedback too aggressively. We introduce Meta-Harness, an outer-loop system that searches over harness code for LLM applications. It uses an agentic proposer that accesses the source code, scores, and execution traces of all prior candidates through a filesystem. On online text classification, Meta-Harness improves over a state-of-the-art context management system by 7.7 points while using 4x fewer context tokens. On retrieval-augmented math reasoning, a single discovered harness improves accuracy on 200 IMO-level problems by 4.7 points on average across five held-out models. On agentic coding, discovered harnesses surpass the best hand-engineered baselines on TerminalBench-2. Together, these results show that richer access to prior experience can enable automated harness engineering.

Dogfight is a tactical behavior of cooperation between fighters. Inspired by this, this paper proposes a novel metaphor-free metaheuristic algorithm called Dogfight Search (DoS). Unlike traditional algorithms, DoS draws algorithmic framework from the inspiration, but its search mechanism is constructed based on the displacement integration equations in kinematics. Through experimental validation on CEC2017 and CEC2022 benchmark test functions, 10 real-world constrained optimization problems and mountainous terrain path planning tasks, DoS significantly outperforms 7 advanced competitors in overall performance and ranks first in the Friedman ranking. Furthermore, this paper compares the performance of DoS with 3 SOTA algorithms on the CEC2017 and CEC2022 benchmark test functions. The results show that DoS continues to maintain its lead, demonstrating strong competitiveness. The source code of DoS is available at https://ww2.mathworks.cn/matlabcentral/fileexchange/183519-dogfight-search.

Monte Carlo particle transport problems play a vital role in scientific computing, but solving them on exiting von Neumann architectures suffers from random branching and irregular memory access, causing computing inefficiency due to a fundamental mismatch between stochastic algorithms and deterministic hardware. To bridge this gap, we propose MCPT-Solver, a spin-based hardware true random number generator (TRNG) with tunable output probability enabled by a Bayesian inference network architecture. It is dedicated for efficiently solving stochastic applications including Monte Carlo particle transport problems. First, we leverage the stochastic switching property of spin devices to provide a high-quality entropy source for the TRNG and achieve high generating throughput and process-voltage-temperature tolerance through optimized control logic and write mechanism designs. Next, we propose a hardware Bayesian inference network to enable probability-tunable random number outputs. Finally, we present a system-level simulation framework to evaluate MCPT-Solver. Experimental results show that MCPT-Solver achieves a mean squared error of 7.6e-6 for solving transport problems while demonstrating a dramatic acceleration effect over general-purpose processors. Additionally, the MCPT-Solver's throughput reaches 185 Mb/s with an area of 27.8 um2/bit and energy consumption of 8.6 pJ/bit, making it the first spin-based TRNG that offers both process-voltage-temperature tolerance and adjustable probability.

Deep learning training is non-deterministic: identical code with different random seeds produces models that agree on aggregate metrics but disagree on individual predictions, with per-class AUC swings exceeding 20 percentage points on rare clinical classes. We present a framework for verified bit-identical training that eliminates three sources of randomness: weight initialization (via structured orthogonal basis functions), batch ordering (via golden ratio scheduling), and non-deterministic GPU operations (via architecture selection and custom autograd). The pipeline produces MD5-verified identical trained weights across independent runs. On PTB-XL ECG rhythm classification, structured initialization significantly exceeds Kaiming across two architectures (n=20; Conformer p = 0.016, Baseline p < 0.001), reducing aggregate variance by 2-3x and reducing per-class variability on rare rhythms by up to 7.5x (TRIGU range: 4.1pp vs 30.9pp under Kaiming, independently confirmed by 3-fold CV). A four-basis comparison at n=20 shows all structured orthogonal bases produce equivalent performance (Friedman p=0.48), establishing that the contribution is deterministic structured initialization itself, not any particular basis function. Cross-domain validation on seven MedMNIST benchmarks (n=20, all p > 0.14) confirms no performance penalty on standard tasks; per-class analysis on imbalanced tasks (ChestMNIST, RetinaMNIST) shows the same variance reduction on rare classes observed in ECG. Cross-dataset evaluation on three external ECG databases confirms zero-shot generalization (>0.93 AFIB AUC).

As Large Language Models (LLMs) achieve increasingly sophisticated performance on complex reasoning tasks, current architectures serve as critical proxies for the internal heuristics of frontier models. Characterizing emergent reasoning is vital for long-term interpretability and safety. Furthermore, understanding how prompting modulates these processes is essential, as natural language will likely be the primary interface for interacting with AGI systems. In this work, we use a custom variant of Genetic Pareto (GEPA) to systematically optimize prompts for scientific reasoning tasks, and analyze how prompting can affect reasoning behavior. We investigate the structural patterns and logical heuristics inherent in GEPA-optimized prompts, and evaluate their transferability and brittleness. Our findings reveal that gains in scientific reasoning often correspond to model-specific heuristics that fail to generalize across systems, which we call "local" logic. By framing prompt optimization as a tool for model interpretability, we argue that mapping these preferred reasoning structures for LLMs is an important prerequisite for effectively collaborating with superhuman intelligence.

Diffusion maps (DMAP) are often used as a dimensionality-reduction tool, but more precisely they provide a spectral representation of the intrinsic geometry rather than a complete charting method. To illustrate this distinction, we study a Swiss roll with known isometric coordinates and compare DMAP, Isomap, and UMAP across latent dimensions. For each representation, we fit an oracle affine readout to the ground-truth chart and measure reconstruction error. Isomap most efficiently recovers the low-dimensional chart, UMAP provides an intermediate tradeoff, and DMAP becomes accurate only after combining multiple diffusion modes. Thus the correct chart lies in the span of diffusion coordinates, but standard DMAP do not by themselves identify the appropriate combination.

This paper presents new resources and baselines for Dependency Parsing in Pomak, an endangered Eastern South Slavic language with substantial dialectal variation and no widely adopted standard. We focus on the variety spoken in Turkey (Uzunköprü) and ask how well a dependency parser trained on the existing Pomak Universal Dependencies treebank, which was built primarily from the variety that is spoken in Greece, transfers across dialects. We run two experimental phases. First, we train a parser on the Greek-variety UD data and evaluate zero-shot transfer to Turkish-variety Pomak, quantifying the impact of phonological and morphosyntactic differences. Second, we introduce a new manually annotated Turkish-variety Pomak corpus of 650 sentences and show that, despite its small size, targeted fine-tuning substantially improves accuracy; performance is further boosted by cross-variety transfer learning that combines the two dialects.

The convergence of low-altitude economies, embodied intelligence, and air-ground cooperative systems creates growing demand for simulation infrastructure capable of jointly modeling aerial and ground agents within a single physically coherent environment. Existing open-source platforms remain domain-segregated: driving simulators lack aerial dynamics, while multirotor simulators lack realistic ground scenes. Bridge-based co-simulation introduces synchronization overhead and cannot guarantee strict spatial-temporal consistency. We present CARLA-Air, an open-source infrastructure that unifies high-fidelity urban driving and physics-accurate multirotor flight within a single Unreal Engine process. The platform preserves both CARLA and AirSim native Python APIs and ROS 2 interfaces, enabling zero-modification code reuse. Within a shared physics tick and rendering pipeline, CARLA-Air delivers photorealistic environments with rule-compliant traffic, socially-aware pedestrians, and aerodynamically consistent UAV dynamics, synchronously capturing up to 18 sensor modalities across all platforms at each tick. The platform supports representative air-ground embodied intelligence workloads spanning cooperation, embodied navigation and vision-language action, multi-modal perception and dataset construction, and reinforcement-learning-based policy training. An extensible asset pipeline allows integration of custom robot platforms into the shared world. By inheriting AirSim's aerial capabilities -- whose upstream development has been archived -- CARLA-Air ensures this widely adopted flight stack continues to evolve within a modern infrastructure. Released with prebuilt binaries and full source: https://github.com/louiszengCN/CarlaAir

Optical character recognition remains critical infrastructure for document digitization, yet state-of-the-art performance is often restricted to well-resourced institutions by prohibitive computational barriers. End-to-end transformer architectures achieve strong accuracy but demand hundreds of GPU hours for domain adaptation, limiting accessibility for practitioners and digital humanities scholars. We present a modular detection-and-correction framework that achieves near-SOTA accuracy with single-GPU training. Our approach decouples lightweight visual character detection (domain-agnostic) from domain-specific linguistic correction using pretrained sequence models including T5, ByT5, and BART. By training the correctors entirely on synthetic noise, we enable annotation-free domain adaptation without requiring labeled target images. Evaluating across modern clean handwriting, cursive script, and historical documents, we identify a critical "Pareto frontier" in architecture selection: T5-Base excels on modern text with standard vocabulary, whereas ByT5-Base dominates on historical documents by reconstructing archaic spellings at the byte level. Our results demonstrate that this decoupled paradigm matches end-to-end transformer accuracy while reducing compute by approximately 95%, establishing a viable, resource-efficient alternative to monolithic OCR architectures.

Scientific figure multiple-choice question answering (MCQA) requires models to reason over diverse visual evidence, ranging from charts and multipanel figures to microscopy and biomedical images. However, this setting suffers from a distinctive bias: answer choices themselves can act as priors, steering multimodal models toward scientifically plausible options even when the figure supports a different answer. We investigate this failure mode through a simple question: what if decoding explicitly discounts what the model would prefer from text alone, so as to favor figure-grounded evidence? To this end, we propose SCICON, a training-free decoding method that scores each candidate by subtracting a text-only option score from its image-conditioned counterpart. Unlike prior contrastive decoding approaches that mitigate hallucinations by contrasting original inputs with distorted images or perturbed instructions, SCICON directly targets the choice-induced prior encoded in candidate text. Across three scientific figure QA benchmarks and three model backbones, SCICON consistently improves accuracy over standard decoding baselines. These results show that decoding against choice-induced priors is an effective and simple way to improve figure-grounded reasoning in scientific MCQA.

Deep learning models for drug-like molecules and proteins overwhelmingly reuse transformer architectures designed for natural language, yet whether molecular sequences benefit from different designs has not been systematically tested. We deploy autonomous architecture search via an agent across three sequence types (SMILES, protein, and English text as control), running 3,106 experiments on a single GPU. For SMILES, architecture search is counterproductive: tuning learning rates and schedules alone outperforms the full search (p = 0.001). For natural language, architecture changes drive 81% of improvement (p = 0.009). Proteins fall between the two. Surprisingly, although the agent discovers distinct architectures per domain (p = 0.004), every innovation transfers across all three domains with <1% degradation, indicating that the differences reflect search-path dependence rather than fundamental biological requirements. We release a decision framework and open-source toolkit for molecular modeling teams to choose between autonomous architecture search and simple hyperparameter tuning.

We present a stage-decomposed analysis of prompt injection attacks against five frontier LLM agents. Prior work measures task-level attack success rate (ASR); we localize the pipeline stage at which each model's defense activates. We instrument every run with a cryptographic canary token (SECRET-[A-F0-9]{8}) tracked through four kill-chain stages -- Exposed, Persisted, Relayed, Executed -- across four attack surfaces and five defense conditions (764 total runs, 428 no-defense attacked). Our central finding is that model safety is determined not by whether adversarial content is seen, but by whether it is propagated across pipeline stages. Concretely: (1) in our evaluation, exposure is 100% for all five models -- the safety gap is entirely downstream; (2) Claude strips injections at write_memory summarization (0/164 ASR), while GPT-4o-mini propagates canaries without loss (53% ASR, 95% CI: 41--65%); (3) DeepSeek exhibits 0% ASR on memory surfaces and 100% ASR on tool-stream surfaces from the same model -- a complete reversal across injection channels; (4) all four active defense conditions (write_filter, pi_detector, spotlighting, and their combination) produce 100% ASR due to threat-model surface mismatch; (5) a Claude relay node decontaminates downstream agents -- 0/40 canaries survived into shared memory.

Heterogeneous Multi-Embodied Agent Systems involve coordinating multiple embodied agents with diverse capabilities to accomplish tasks in dynamic environments. This process requires the collection, generation, and consumption of massive, heterogeneous data, which primarily falls into three categories: static knowledge regarding the agents, tasks, and environments; multimodal training datasets tailored for various AI models; and high-frequency sensor streams. However, existing frameworks lack a unified data management infrastructure to support the real-world deployment of such systems. To address this gap, we present \textbf{HeteroHub}, a data-centric framework that integrates static metadata, task-aligned training corpora, and real-time data streams. The framework supports task-aware model training, context-sensitive execution, and closed-loop control driven by real-world feedback. In our demonstration, HeteroHub successfully coordinates multiple embodied AI agents to execute complex tasks, illustrating how a robust data management framework can enable scalable, maintainable, and evolvable embodied AI systems.

Statistical heterogeneity in Federated Learning (FL) often leads to negative transfer, where a single global model fails to serve diverse client distributions. Personalized federated learning (pFL) aims to address this by tailoring models to individual clients. However, under most existing pFL approaches, clients integrate peer client contributions uniformly, which ignores the reality that not all peers are likely to be equally beneficial. Additionally, the potential for personalization at the instance level remains largely unexplored, even though the reliability of different peer models often varies across individual samples within the same client. We introduce FedDES (Federated Dynamic Ensemble Selection), a decentralized pFL framework that achieves instance-level personalization through dynamic ensemble selection. Central to our approach is a Graph Neural Network (GNN) meta-learner trained on a heterogeneous graph modeling interactions between data samples and candidate classifiers. For each test query, the GNN dynamically selects and weights peer client models, forming an ensemble of the most competent classifiers while effectively suppressing contributions from those that are irrelevant or potentially harmful for performance. Experiments on CIFAR-10 and real-world ICU healthcare data demonstrate that FedDES outperforms state-of-the-art pFL baselines in non-IID settings, offering robust protection against negative transfer.

Atomic decomposition -- breaking a candidate answer into claims before verifying each against a reference -- is a widely adopted design for LLM-based reference-grounded judges. However, atomic prompts are typically richer and longer, making it unclear whether any advantage comes from decomposition or from richer prompting. We study this for benchmark-style completeness-sensitive reference-support classification: classifying a candidate as fully supported, partially supported, or unsupported relative to a supplied reference. We compare a self-decomposing atomic judge (single-prompt decompose-and-verify) against a prompt-controlled holistic judge with the same inputs and a similarly detailed rubric. On 200 source examples per dataset across TruthfulQA, ASQA, and QAMPARI, with four model families, source-level paired tests, cluster bootstrap, and aggregation across three pre-frozen prompt variants per design family, we find the holistic judge matches or exceeds the atomic judge on two of three benchmarks: ASQA and QAMPARI favor holistic across all four families (statistically reliable in three of four), while TruthfulQA shows a small atomic edge. The holistic advantage is concentrated in partially\_supported cases -- incompleteness detection. A sensitivity check against human annotations confirms the ranking under both benchmark-completeness and human factual-correctness standards. Our finding is specific to the self-decomposing single-prompt pattern on three QA-style benchmarks with 200 source examples each; multi-stage atomic pipelines and non-QA tasks remain untested. Among perturbations examined, reference-quality degradation produced the largest accuracy drops for both judge families.

Reverse engineering tools remain monolithic and imperative compared to the advancement of modern compiler architectures: analyses are tied to a single mutable representation, making them difficult to extend or refine, and forcing premature choices between soundness and precision. We observe that decompilation is the reverse of compilation and can be structured as a sequence of modular passes, each performing a granular and clearly defined interpretation of the binary at a progressively higher level of abstraction. We formalize this as provenance-guided superset decompilation (PGSD), a framework that monotonically derives facts about the binary into a relation store. Instead of committing early to a single interpretation, the pipeline retains ambiguous interpretations as parallel candidates with provenance, deferring resolution until the final selection phase. Manifold implements PGSD as a declarative reverse engineering framework that lifts Linux ELF binaries to C99 through a granular intermediate representation in ~35K lines of Rust and Datalog. On GNU coreutils, Manifold's output quality matches Ghidra, IDA Pro, angr, and RetDec on multiple metrics while producing fewer compiler errors, and generalizes across compilers and optimization levels.

RDMA link failures can render connections temporarily unavailable, causing both performance degradation and significant recovery overhead. To tolerate such failures, production datacenters assign each primary link with a standby link and, upon failure, uniformly retransmit all in-flight RDMA request over the backup path. However, we observe that such blanket retransmission is unnecessary. In-flight requests can be split into pre-failure and post-failure categories depending on whether the responder has already executed. Retransmitting post-failure requests is not only redundant (consuming bandwidth), but also incorrect for non-idempotent operations, where duplicate execution can violate application semantics. We present Varuna, a failure-type-aware RDMA recovery mechanism that enables correct retransmission and us-level failover. Varuna piggybacks a lightweight completion log on every RDMA operation; after a link failure, this log deterministically reveals which in-flight requests were executed (post-failure) and which were lost (pre-failure). Varuna then retransmits only the pre-failure subset and fetches/recovers the return values for post-failure requests. Evaluated using synthetic microbenchmarks and end-to-end RDMA TPC-C transactions, Varuna incurs only 0.6-10% steady-state latency overhead in realistic applications, eliminates 65% of recovery retransmission time, preserves transactional consistency, and introduces zero connectivity rebuild overhead and negligible memory overhead during RDMA failover.

Individualized head-related impulse responses (HRIRs) enable binaural rendering, but dense per-listener measurements are costly. We address HRIR spatial up-sampling from sparse per-listener measurements: given a few measured HRIRs for a listener, predict HRIRs at unmeasured target directions. Prior learning methods often work in the frequency domain, rely on minimum-phase assumptions or separate timing models, and use a fixed direction grid, which can degrade temporal fidelity and spatial continuity. We propose BiFormer3D, a time-domain, grid-free binaural Transformer for reconstructing HRIRs at arbitrary directions from sparse inputs. It uses sinusoidal spatial features, a Conv1D refinement module, and auxiliary interaural time difference (ITD) and interaural level difference (ILD) heads. On SONICOM, it improves normalized mean squared error (NMSE), cosine distance, and ITD/ILD errors over prior methods; ablations validate modules and show minimum-phase pre-processing is unnecessary.

Diffusion models have emerged as a powerful framework for generative tasks in deep learning. They decompose generative modeling into two computational primitives: deterministic neural-network evaluation and stochastic sampling. Current implementations usually place most computation in the neural network, but diffusion as a framework allows a broader range of choices for the stochastic transition kernel. Here, we generalize the stochastic sampling component by replacing independent noise injection with Markov chain Monte Carlo (MCMC) dynamics that incorporate known interaction structure. Standard independent diffusion is recovered as a special case when couplings are set to zero. By explicitly incorporating Ising couplings into the diffusion dynamics, the noising and denoising processes exploit spatial correlations representative of the target system. The resulting framework maps naturally onto probabilistic computers (p-computers) built from probabilistic bits (p-bits), which provide orders-of-magnitude advantages in sampling throughput and energy efficiency over GPUs. We demonstrate the approach on equilibrium states of the 2D ferromagnetic Ising model and the 3D Edwards-Anderson spin glass, showing that correlated diffusion produces samples in closer agreement with MCMC reference distributions than independent diffusion. More broadly, the framework shows that p-computers can enable new classes of diffusion algorithms that exploit structured probabilistic sampling for generative modeling.

Interactive documents help readers engage with complex ideas through dynamic visualization, interactive animations, and exploratory interfaces. However, creating such documents remains costly, as it requires both domain expertise and web development skills. Recent Large Language Model (LLM)-based agents can automate content creation, but directly applying them to interactive document generation often produces outputs that are difficult to control. To address this, we present ViviDoc, to the best of our knowledge the first work to systematically address interactive document generation. ViviDoc introduces a multi-agent pipeline (Planner, Styler, Executor, Evaluator). To make the generation process controllable, we provide three levels of human control: (1) the Document Specification (DocSpec) with SRTC Interaction Specifications (State, Render, Transition, Constraint) for structured planning, (2) a content-aware Style Palette for customizing writing and interaction styles, and (3) chat-based editing for iterative refinement. We also construct ViviBench, a benchmark of 101 topics derived from real-world interactive documents across 11 domains, along with a taxonomy of 8 interaction types and a 4-dimensional automated evaluation framework validated against human ratings (Pearson r > 0.84). Experiments show that ViviDoc achieves the highest content richness and interaction quality in both automated and human evaluation. A 12-person user study confirms that the system is easy to use, provides effective control over the generation process, and produces documents that satisfy users.

The high cost and accessibility problem associated with large datasets hinder the development of large-scale visual recognition systems. Dataset Distillation addresses these problems by synthesizing compact surrogate datasets for efficient training, storage, transfer, and privacy preservation. The existing state-of-the-art diffusion-based dataset distillation methods face three issues: lack of theoretical justification, poor efficiency in scaling to high data volumes, and failure in data-free scenarios. To address these issues, we establish a theoretical framework that justifies the use of diffusion models by proving the equivalence between dataset distillation and distribution matching, and reveals an inherent efficiency limit in the dataset distillation paradigm. We then propose a Dataset Concentration (DsCo) framework that uses a diffusion-based Noise-Optimization (NOpt) method to synthesize a small yet representative set of samples, and optionally augments the synthetic data via "Doping", which mixes selected samples from the original dataset with the synthetic samples to overcome the efficiency limit of dataset distillation. DsCo is applicable in both data-accessible and data-free scenarios, achieving SOTA performances for low data volumes, and it extends well to high data volumes, where it nearly reduces the dataset size by half with no performance degradation.

Federated learning (FL) enables distributed clients to collaboratively train a global model using local private data. Nevertheless, recent studies show that conventional FL algorithms still exhibit deficiencies in privacy protection, and the server lacks a reliable and stable aggregation rule for updating the global model. This situation creates opportunities for adversaries: on the one hand, they may eavesdrop on uploaded gradients or model parameters, potentially leaking benign clients' private data; on the other hand, they may compromise clients to launch poisoning attacks that corrupt the global model. To balance accuracy and security, we propose FedFG, a robust FL framework based on flow-matching generation that simultaneously preserves client privacy and resists sophisticated poisoning attacks. On the client side, each local network is decoupled into a private feature extractor and a public classifier. Each client is further equipped with a flow-matching generator that replaces the extractor when interacting with the server, thereby protecting private features while learning an approximation of the underlying data distribution. Complementing the client-side design, the server employs a client-update verification scheme and a novel robust aggregation mechanism driven by synthetic samples produced by the flow-matching generator. Experiments on MNIST, FMNIST, and CIFAR-10 demonstrate that, compared with prior work, our approach adapts to multiple attack strategies and achieves higher accuracy while maintaining strong privacy protection.

Vision-language models (VLMs) achieve strong performance on many benchmarks, yet a basic reliability question remains underexplored: when visual evidence conflicts with commonsense, do models follow what is shown or what commonsense suggests? A characteristic failure in this setting is that the model overrides visual evidence and outputs the commonsense alternative. We term this phenomenon \textbf{commonsense-driven hallucination} (CDH). To evaluate it, we introduce \textbf{CDH-Bench}, a benchmark designed to create explicit \textbf{visual evidence--commonsense conflicts}. CDH-Bench covers three dimensions: \textit{counting anomalies}, \textit{relational anomalies}, and \textit{attribute anomalies}. We evaluate frontier VLMs under \textit{binary Question Answering (QA)} and \textit{multiple-choice QA}, and report metrics including \textit{Counterfactual Accuracy} (CF-Acc), \textit{Commonsense Accuracy} (CS-Acc), \textit{Counterfactual Accuracy Drop} (CFAD), \textit{Commonsense Collapse Rate} (CCR), and \textit{Relative Prior Dependency} (RPD). Results show that even strong models remain vulnerable to prior-driven normalization under visual evidence--commonsense conflict. CDH-Bench provides a controlled diagnostic of visual fidelity under visual evidence--commonsense conflict.

Automatic speech recognition (ASR) has advanced rapidly in recent years, driven by large-scale pretrained models and end-to-end architectures such as SLAM-ASR. A key component of SLAM-ASR systems is the Whisper speech encoder, which provides robust acoustic representations. While model pruning has been explored for the full Whisper encoder-decoder architecture, its impact within the SLAM-ASR setting remains under-investigated. In this work, we analyze the effects of layer pruning in the Whisper encoder when used as the acoustic backbone of SLAM-ASR. We further examine the extent to which LoRA-based fine-tuning can recover performance degradation caused by pruning. Experiments conducted across three Whisper variants (Small, Medium, Large-v2), three languages representing distinct resource levels (Danish, Dutch, English), and over 200 training runs demonstrate that pruning two encoder layers causes only 2-4% WER degradation, and that combining this pruning with LoRA adaptation consistently outperforms the unpruned baseline while reducing total parameters by 7-14%. Moreover, our error analysis reveals that LoRA primarily compensates through the language model's linguistic priors, reducing total word errors by 11-21% for Dutch and English, with substitutions and deletions showing the largest reductions. However, for low-resource Danish, the reduction is smaller (4-7%), and LoRA introduces increased insertion errors, indicating that compensation effectiveness depends on the LLM's pre-existing language proficiency and available training data.

Reinforcement learning has become central to improving large reasoning models, but its success still relies heavily on verifiable rewards or labeled supervision. This limits its applicability to open ended domains where correctness is ambiguous and cannot be verified. Moreover, reasoning trajectories remain largely unconstrained, and optimization towards final answer can favor early exploitation over generalization. In this work, we ask whether general reasoning ability can be improved by teaching models how to think (the structure of reasoning) rather than what to produce (the outcome of reasoning) and extend traditional RLVR to open ended settings. We introduce structure aware reinforcement learning (SARL), a label free framework that constructs a per response Reasoning Map from intermediate thinking steps and rewards its small world topology, inspired by complex networks and the functional organization of the human brain. SARL encourages reasoning trajectories that are both locally coherent and globally efficient, shifting supervision from destination to path. Our experiments on Qwen3-4B show SARL surpasses ground truth based RL and prior label free RL baselines, achieving the best average gain of 9.1% under PPO and 11.6% under GRPO on math tasks and 34.6% under PPO and 30.4% under GRPO on open ended tasks. Beyond good performance, SARL also exhibits lower KL divergence, higher policy entropy, indicating a more stable and exploratory training and generalized reasoning ability.

Recent years have witnessed the widespread adoption of reinforcement learning (RL), from solving real-time games to fine-tuning large language models using human preference data significantly improving alignment with user expectations. However, as model complexity grows exponentially, the interpretability of these systems becomes increasingly challenging. While numerous explainability methods have been developed for computer vision and natural language processing to elucidate both local and global reasoning patterns, their application to RL remains limited. Direct extensions of these methods often struggle to maintain the delicate balance between interpretability and performance within RL settings. Prototype-Wrapper Networks (PW-Nets) have recently shown promise in bridging this gap by enhancing explainability in RL domains without sacrificing the efficiency of the original black-box models. However, these methods typically require manually defined reference prototypes, which often necessitate expert domain knowledge. In this work, we propose a method that removes this dependency by automatically selecting optimal prototypes from the available data. Preliminary experiments on standard Gym environments demonstrate that our approach matches the performance of existing PW-Nets, while remaining competitive with the original black-box models.

Distributed learning has gained significant attention due to its advantages in scalability, privacy, and fault tolerance.In this paradigm, multiple agents collaboratively train a global model by exchanging parameters only with their neighbors. However, a key vulnerability of existing distributed learning approaches is their implicit assumption that all agents behave honestly during gradient updates. In real-world scenarios, this assumption often breaks down, as selfish or strategic agents may be incentivized to manipulate gradients for personal gain, ultimately compromising the final learning outcome. In this work, we propose a fully distributed payment mechanism that, for the first time, guarantees both truthful behaviors and accurate convergence in distributed stochastic gradient descent. This represents a significant advancement, as it overcomes two major limitations of existing truthfulness mechanisms for collaborative learning:(1) reliance on a centralized server for payment collection, and (2) sacrificing convergence accuracy to guarantee truthfulness. In addition to characterizing the convergence rate under general convex and strongly convex conditions, we also prove that our approach guarantees the cumulative gain that an agent can obtain through strategic behavior remains finite, even as the number of iterations approaches infinity--a property unattainable by most existing truthfulness mechanisms. Our experimental results on standard machine learning tasks, evaluated on benchmark datasets, confirm the effectiveness of the proposed approach.

Although Large Vision Language Models (LVLMs) have demonstrated impressive multimodal reasoning capabilities, their scalability and deployment are constrained by massive computational requirements. In particular, the massive amount of visual tokens from high-resolution input data aggravates the situation due to the quadratic complexity of attention mechanisms. To address these issues, the research community has developed several optimization frameworks. This paper presents a comprehensive survey of the current state-of-the-art techniques for accelerating LVLM inference. We introduce a systematic taxonomy that categorizes existing optimization frameworks into four primary dimensions: visual token compression, memory management and serving, efficient architectural design, and advanced decoding strategies. Furthermore, we critically examine the limitations of these current methodologies and identify critical open problems to inspire future research directions in efficient multimodal systems.

Analogical reasoning tests a fundamental aspect of human cognition: mapping the relation from one pair of objects to another. Existing evaluations of this ability in multimodal large language models (MLLMs) overlook the ability to compose rules from multiple sources, a critical component of higher-order intelligence. To close this gap, we introduce CARV (Compositional Analogical Reasoning in Vision), a novel task together with a 5,500-sample dataset as the first diagnostic benchmark. We extend the analogy from a single pair to multiple pairs, which requires MLLMs to extract symbolic rules from each pair and compose new transformations. Evaluation on the state-of-the-art MLLMs reveals a striking performance gap: even Gemini-2.5 Pro achieving only 40.4% accuracy, far below human-level performance of 100%. Diagnostic analysis shows two consistent failure modes: (1) decomposing visual changes into symbolic rules, and (2) maintaining robustness under diverse or complex settings, highlighting the limitations of current MLLMs on this task.

We introduce AI-Kolmogorov, a novel framework for Symbolic Density Estimation (SymDE). Symbolic regression (SR) has been effectively used to produce interpretable models in standard regression settings but its applicability to density estimation tasks has largely been unexplored. To address the SymDE task we introduce a multi-stage pipeline: (i) problem decomposition through clustering and/or probabilistic graphical model structure learning; (ii) nonparametric density estimation; (iii) support estimation; and finally (iv) SR on the density estimate. We demonstrate the efficacy of AI-Kolmogorov on synthetic mixture models, multivariate normal distributions, and three exotic distributions, two of which are motivated by applications in high-energy physics. We show that AI-Kolmogorov can discover underlying distributions or otherwise provide valuable insight into the mathematical expressions describing them.

Protein interaction modeling is central to protein design, which has been transformed by machine learning with applications in drug discovery and beyond. In this landscape, structure-based de novo binder design is cast as either conditional generative modeling or sequence optimization via structure predictors ("hallucination"). We argue that this is a false dichotomy and propose Proteina-Complexa, a novel fully atomistic binder generation method unifying both paradigms. We extend recent flow-based latent protein generation architectures and leverage the domain-domain interactions of monomeric computationally predicted protein structures to construct Teddymer, a new large-scale dataset of synthetic binder-target pairs for pretraining. Combined with high-quality experimental multimers, this enables training a strong base model. We then perform inference-time optimization with this generative prior, unifying the strengths of previously distinct generative and hallucination methods. Proteina-Complexa sets a new state of the art in computational binder design benchmarks: it delivers markedly higher in-silico success rates than existing generative approaches, and our novel test-time optimization strategies greatly outperform previous hallucination methods under normalized compute budgets. We also demonstrate interface hydrogen bond optimization, fold class-guided binder generation, and extensions to small molecule targets and enzyme design tasks, again surpassing prior methods. Code, models and new data will be publicly released.

Large Language Models (LLMs) are widely applied across various domains due to their powerful text generation capabilities. While LLM-generated texts often resemble human-written ones, their misuse can lead to significant societal risks. Detecting such texts is an essential technique for mitigating LLM misuse, and many detection methods have shown promising results across different datasets. However, real-world scenarios often involve out-of-domain inputs or adversarial samples, which can affect the performance of detection methods to varying degrees. Furthermore, most existing research has focused on English texts, with limited work addressing Chinese text detection. In this study, we propose EnsemJudge, a robust framework for detecting Chinese LLM-generated text by incorporating tailored strategies and ensemble voting mechanisms. We trained and evaluated our system on a carefully constructed Chinese dataset provided by NLPCC2025 Shared Task 1. Our approach outperformed all baseline methods and achieved first place in the task, demonstrating its effectiveness and reliability in Chinese LLM-generated text detection. Our code is available at https://github.com/johnsonwangzs/MGT-Mini.

As industry and academia continue to advance spaceborne computing and communication capabilities, the formation of cloud-native space clusters (CNSCs) has become an increasingly evident trend. This evolution progressively exposes the resource management challenges associated with coordinating fragmented and heterogeneous onboard resources while supporting large-scale and diverse space applications. However, directly transplanting mature terrestrial cloud-native cluster operating system paradigms into space is ineffective due to the fragmentation of spaceborne computing resources and satellite mobility, which collectively impose substantial challenges on resource awareness and orchestration. This article presents YUHENG-OS, a cloud-native space cluster operating system tailored for CNSCs. YUHENG-OS provides unified abstraction, awareness, and orchestration of heterogeneous spaceborne infrastructure, enabling cluster-wide task deployment and scheduling across distributed satellites. We introduce a four-layer system architecture and three key enabling technologies: modeling of heterogeneous resource demands for space tasks, fragmented heterogeneous resource awareness under network constraints, and matching of differentiated tasks with multidimensional heterogeneous resources under temporal dependency constraints. Evaluation results show that, compared with representative terrestrial cloud-native cluster operating systems exemplified by Kubernetes, YUHENG-OS achieves a substantially higher task completion ratio, with improvements of up to 98%. This advantage is primarily attributed to its ability to reduce resource awareness delay by 71%.

Japanese scene text poses challenges that multilingual benchmarks often fail to capture, including mixed scripts, frequent vertical writing, and a character inventory far larger than the Latin alphabet. Although Japanese is included in several multilingual benchmarks, these resources do not adequately capture the language-specific complexities. Meanwhile, existing Japanese visual text datasets have primarily focused on scanned documents, leaving in-the-wild scene text underexplored. To fill this gap, we introduce JaWildText, a diagnostic benchmark for evaluating vision-language models (VLMs) on Japanese scene text understanding. JaWildText contains 3,241 instances from 2,961 images newly captured in Japan, with 1.12 million annotated characters spanning 3,643 unique character types. It comprises three complementary tasks that vary in visual organization, output format, and writing style: (i) Dense Scene Text Visual Question Answering (STVQA), which requires reasoning over multiple pieces of visual text evidence; (ii) Receipt Key Information Extraction (KIE), which tests layout-aware structured extraction from mobile-captured receipts; and (iii) Handwriting OCR, which evaluates page-level transcription across various media and writing directions. We evaluate 14 open-weight VLMs and find that the best model achieves an average score of 0.64 across the three tasks. Error analyses show recognition remains the dominant bottleneck, especially for kanji. JaWildText enables fine-grained, script-aware diagnosis of Japanese scene text capabilities, and will be released with evaluation code.

Most of modern neural machine translation (NMT) models are based on an encoder-decoder framework with an attention mechanism. While they perform well on standard datasets, they can have trouble in translation of long inputs that are rare or unseen during training. Incorporating target syntax is one approach to dealing with such length-related problems. We propose a novel syntactic decoder that generates a target-language dependency tree in a top-down, left-to-right order. Experiments show that the proposed top-down string-to-tree decoding generalizes better than conventional sequence-to-sequence decoding in translating long inputs that are not observed in the training data.

Nonlinear Partial Differential Equations (PDEs) are ubiquitous in mathematical physics and engineering. Although Physics-Informed Neural Networks (PINNs) have emerged as a powerful tool for solving PDE problems, they typically struggle to identify multiple distinct solutions, since they are designed to find one solution at a time. To address this limitation, we introduce Deflation-PINNs, a novel framework that integrates a deflation loss with an architecture based on PINNs and Deep Operator Networks (DeepONets). By incorporating a deflation term into the loss function, our method systematically forces the Deflation-PINN to seek and converge upon distinct finitely many solution branches. We provide theoretical evidence on the convergence of our model and demonstrate the efficacy of Deflation-PINNs through numerical experiments on the Landau-de Gennes model of liquid crystals, a system renowned for its complex energy landscape and multiple equilibrium states. Our results show that Deflation-PINNs can successfully identify and characterize multiple distinct crystal structures.