arXiv Daily Digest - 2026-05-05

Power oversubscription is increasingly central to datacenter operation as power density grows, making it necessary to dynamically allocate limited power budgets across devices based on real-time demand. Existing approaches typically assume flat power domains, whereas in practice power distribution is hierarchical and allocation decisions must additionally respect tenant-level contractual constraints. We present nvPAX, a constrained-optimization policy that computes feasible power allocations at every control step via a three-phase hybrid QP/LP procedure. Phase I allocates power with minimum deviation from each device's power request, while respecting job priorities. Phase II fairly distributes excess power among active devices. Phase III fairly distributes any remaining power to idle devices. The rationale behind the three phases is to allow power oversubscription while maximizing datacenter utilization. On a trace-driven large-scale simulation using GPU power telemetry from a production datacenter, nvPAX runs with a mean wall-clock time of 264.69 ms per allocation interval and achieves a mean satisfaction ratio of 98.92%, outperforming static equal-share allocation and providing robustness beyond greedy proportional allocation in the presence of non-uniform hierarchical bottlenecks.

Modern hardware compilers increasingly rely on rich intermediate representations (IRs) to preserve optimization-relevant semantics before generating RTL code. However, one important optimization is still largely deferred to backend tools: pipeline optimization. In common RTL flows, registers are inserted by frontend heuristics or hardware designers and later adjusted by backend retiming after the design has been lowered to a much lower-level netlist representation. At that point, much of the operator-level structure originally exposed by the compiler IR has already been weakened or lost, limiting opportunities for global, compiler-level pipeline optimization. This paper presents PipeRTL, an IR-level pipeline optimization framework for hardware compilers, instantiated in CIRCT. PipeRTL makes the legality of register relocation explicit in the IR, uses a learned timing predictor to approximate downstream delay behavior, and formulates timing-aware register relocation as a global min-cost flow problem under timing constraints. Evaluation on open-source designs under a commercial backend synthesis flow shows that PipeRTL improves downstream implementation quality on average, reducing critical-path delay, power, and area across the evaluated benchmarks, while also providing a stronger starting point for backend retiming. These results indicate that exposing pipeline optimization as an explicit compiler pass can deliver backend-meaningful gains by improving the sequential structure presented to later stages and the resulting downstream implementation quality.

Nonlinear coupled systems are ubiquitous in science and engineering. The analysis and modeling of such systems is challenging due to their high dimensionality and complex interactions among subsystems. In recent years, operator-theoretic methods based on the Koopman operator have attracted attention as a powerful tool for analyzing and modeling nonlinear dynamical systems. Extended dynamic mode decomposition (EDMD) is one of the most popular methods to approximate the Koopman operator. However, EDMD is a purely data-driven method, and it could be unstable and inaccurate for coupled systems under limited data availability. In this paper, we propose a method to learn the Koopman operator for coupled systems using the differential equations governing each subsystem. We also demonstrate its effectiveness through numerical experiments on coupled oscillator systems.

Contrastive learning (CL) reduces annotation cost via auto-derived supervisory signals. Since large-scale in-house CL datasets are infeasible, reliance on third-party or internet data is common. Recent studies show CL models are vulnerable to data-poisoning backdoor attacks, but their generalization and robustness are underexplored. We systematically evaluate existing data-poisoning backdoor attacks on CL, revealing limitations: poor dataset adaptability, low success rates, limited portability, and restrictive assumptions (e.g., downstream task knowledge). Interestingly, trigger samples exhibit distinguishable statistical divergence from clean samples, which inspires repurposing it as a watermark for dataset IP protection. Direct repurposing is challenging due to low success rates; we overcome this by statistical verification using a unified density metric. We further propose a multi-level watermarking scheme adapting to feature-level, soft-label, or hard-label outputs in CL. Experiments show some backdoor attacks can be repurposed as effective watermarks with trade-offs among fidelity, verifiability, and robustness. This work demonstrates weak backdoor effects become reliable signals for dataset IP protection in challenging CL settings.

We address the challenge of remote control where one or more actors, lacking direct reward access, are steered by a controller over a communication-constrained channel. The controller learns an optimal policy from observed rewards and communicates action guidance to the actors, which becomes demanding for large or continuous action spaces. To achieve rate-efficient communication throughout this interactive learning and control process, we introduce a novel framework leveraging remote generation. Instead of transmitting full action specifications, the controller sends minimal information, enabling the actors to locally generate actions by sampling from the controller's evolving target policy. This guided sampling is facilitated by an importance sampling approach. Concurrently, the actors use the received guidance as supervised learning data to learn the controller's policy. This actor-side learning improves their local sampling capabilities, progressively reducing future communication needs. Our solution, Guided Remote Action Sampling Policy (GRASP), demonstrates significant communication reduction, achieving an average 12-fold data reduction across all experiments (50-fold for continuous action spaces) compared to direct action transmission, and a 41-fold reduction compared to reward transmission.

Reinforcement Learning from Human Feedback has become the standard paradigm for language model alignment, where reward models directly determine alignment effectiveness. In this work, we focus on how to evaluate the generalizability of reward models. By "generalizability", we mean the ability of RMs to correctly rank responses to align with diverse user preferences. However, existing reward model benchmarks are typically designed around a universal preference, failing to assess this generalization. To address this critical gap, we introduce RMGAP, a benchmark comprising 1,097 instances across Chat, Writing, Reasoning, and Safety domains. Since different users exhibit diverse preferences for the same task, we first generate four distinct responses with different linguistic profiles for each collected prompt. However, the original prompt set lacks the specificity to convey different preferences. We therefore construct tailored prompts by contrasting these candidates and designing scenarios in which one response becomes the uniquely appropriate choice. Moreover, we observe that users often express the same preference using different phrasings, and thus extend each prompt with two paraphrased variants. Our evaluation of 24 state-of-the-art RMs reveals their substantial limitations: even the best RM achieves only 49.27% Best-of-N accuracy, highlighting considerable room for improvement in reward model generalization. Related data and code are available at https://github.com/nanzhi84/RMGAP.

Brain MRI foundation models learn rich representations of anatomy, but interpreting what clinical information they encode remains an open problem. Standard sparse autoencoders (SAEs) suffer from severe feature collapse in deep transformer layers, and in Alzheimer's disease (AD) research, aging confounds nearly every clinical variable, making naive annotation unreliable. We propose GeoSAE, a geometry-guided SAE framework that uses the foundation model's learned manifold structure to prevent feature collapse and annotates each surviving feature via age-deconfounded partial correlations. Applied to ~14k T1-weighted MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Australian Imaging biomarkers and Lifestyle (AIBL) datasets, GeoSAE identifies a compact, fully interpretable feature set that predicts mild cognitive impairment (MCI)-to-AD conversion (AUC 0.746) using only 2% of the embedding dimensions, while comorbidity-annotated features achieve only chance-level performance. The identified features replicate across cohorts without retraining (r=0.97) and localize to neuroanatomically distinct regions consistent with Braak staging. This shows that geometry-guided SAEs can extract interpretable, biomarkers from frozen brain MRI foundation models.

Bandwidth-constrained robotic and surveillance systems often rely on a single compressed video stream to support both continuous scene awareness and downstream machine perception. In practice, this creates a mismatch: low-bitrate video can preserve motion and coarse context, but often loses the fine local detail needed for reliable object recognition and decision-making. Motivated by a hybrid architecture in which low-resolution video supports dynamic scene understanding while eventdriven high-detail regions of interest (ROIs) support close-up identification and analytics, this paper formalizes a two-channel visual telemetry scheme in which a continuous low-bitrate video stream is augmented by selectively transmitted high-detail still ROIs. This first paper does not attempt to prove the superiority of a new still-image codec. Instead, it establishes the hybrid transmission paradigm itself using a practical and reproducible codec stack: x265/HEVC for the base video stream and JPEG stills for ROI refinement. We formulate the problem as bitrate-constrained information selection for robotic vision and define an experimental protocol in which video-only and hybrid schemes are compared under matched total communication budgets. The study is designed around UAV-oriented datasets, two practical bitrate regimes, several ROI triggering policies, and object-level classification refinement on selectively transmitted ROI stills. The resulting paper lays the methodological foundation for a second-stage investigation of JPEG AI as the semantic still-image channel within the same hybrid architecture.

Recently, Reinforcement Learning from Verifiable Rewards (RLVR) has been established as a highly effective technique for augmenting the math reasoning skills of Large Language Models (LLMs) based on a single instance. Current state-of-the-art 1-shot RLVR models adopt heuristics for selecting instances, mostly based on historical variance in rewards, which we find to be inherently misleading as a measure of transferability value. In this paper, we propose a Selector-Guided Autonomous Curriculum (SGAC) approach, which employs a learnable selector model on a multi-dimensional feature space consisting of success probability, reward variance, output disagreement (entropy), and semantic difficulty level, instead of the static reward variance heuristic. In our empirical evaluation on pools of candidate problems, we observed that output disagreement, rather than reward variance, is the strongest predictor of reasoning gains in subsequent iterations. Leveraging this finding, we develop an autonomous curriculum algorithm for dynamically siphoning candidate problems from a large pool, ranking them by the learned selector, and running micro-bursts of 1-shot GRPO. Our framework is evaluated using the Hendrycks MATH benchmark, with the Qwen2.5-Math-1.5B model serving as the baseline. Our framework obtains an accuracy of 68.0\% on the hold-out dataset, which is better than the accuracy obtained from the state-of-the-art model, 64.0\%, as well as the 1-shot RLVR checkpoint proposed by Wang et al., which achieved an accuracy of 66.0\%. The results confirm that entropy-based intelligent data curation leads to strict reasoning improvement over static training methods, particularly in severely limited data conditions.

Large Language Models (LLMs) are increasingly being used to support scientific discovery. In chemistry, tasks such as reaction prediction and structure elucidation require reasoning about the structures of molecules. As such, LLM-based systems for chemistry must interact reliably with molecular structures. Most previous studies of LLMs in chemistry have used SMILES strings or IUPAC names as molecular representations; however, the suitability of these formats has not been systematically assessed. In this work, we introduce MolJSON, a novel molecular representation for LLMs, and systematically compare it with five common chemical formats. We evaluated each representation with GPT-5-nano, GPT-5-mini, GPT-5, and Claude Haiku 4.5 using a set of 78,045 questions spanning translation, shortest path, and constrained generation reasoning tasks. We observed substantial variation across representations in the ability of LLMs to interpret and generate molecular graphs, with MolJSON consistently outperforming existing formats. On translation tasks, GPT-5 achieved 71.0% accuracy when converting IUPAC names to MolJSON, compared with 43.7% when converting the same inputs to SMILES. For constrained generation, GPT-5 reached 95.3% accuracy generating MolJSON, compared with 76.3% for IUPAC and 64.0% for SMILES. As an input format for shortest-path reasoning, GPT-5 successfully answered 98.5% of questions with MolJSON, compared with 92.2% for SMILES and 82.7% for IUPAC, whilst also using fewer reasoning tokens. We observed systematic errors associated with atom count and ring complexity for SMILES strings and IUPAC names, whereas MolJSON was more robust to these failure modes. Our results show that the choice of molecular representation has a material impact on LLM performance, and that explicit molecular graph schemas, such as MolJSON, are a promising direction for LLM-based systems in chemistry.

Diffusion models (DMs) excel on dense continuous data, but are not designed for sparse continuous data. They do not model exact zeros that represent the deliberate absence of a signal. As a result, they erase sparsity patterns and perform unnecessary computation on mostly zero entries. With Sparsity-Exploiting Diffusion (SED), we model only non-zero values, preserving sparsity. SED delivers computational savings while maintaining or improving generation quality by skipping zeros during training and inference. Across physics and biology benchmarks, SED matches or surpasses conventional DMs and domain-specific baselines, while vision experiments provide intuitive insights into the limitations of dense DMs and the benefits of SED.

Gestational Diabetes Mellitus (GDM) is a high-prevalence pregnancy complication that requires accurate early risk stratification to reduce maternal and fetal morbidity. However, real-world clinical deployment of machine learning is hindered by two coupled constraints: (i) label scarcity, where a large fraction of electronic health records (EHR) lack confirmed diagnostic labels, and (ii) data privacy, which prevents sharing patient-level data across hospitals. This paper proposes FedTGNN-SS, a privacy-preserving federated semi-supervised framework for clinical tabular EHR. Each hospital builds a local k-nearest-neighbor patient similarity graph and trains a topology-adaptive GNN encoder. To robustly exploit unlabeled records, FedTGNN-SS combines (1) prototype-guided pseudo-labeling with neighborhood agreement, (2) adaptive graph refinement that periodically updates the k-NN graph using learned embeddings, (3) clinical-aware consistency augmentation applied only to continuous variables, and (4) privacy-safe prototype sharing that exchanges only class-level centroids. Across three diabetes-related datasets (GDM: N = 3,525; Pima: N = 768; Early Stage: N = 520) under 10\%-80\% missing labels per silo, FedTGNN-SS achieves 56 significant wins ($p < 0.05$) against 11 federated baselines and attains strong AUROC under extreme scarcity (Pima: 0.8037 at 80\% missing, Early Stage: 0.9634 at 80\% missing).

Unified audio-visual generation is rapidly gaining industrial and creative relevance, enabling applications in virtual production and interactive media. However, when moving from general audio-video synthesis to music-dance co-generation, the task becomes substantially harder: musical rhythm, phrasing, and accents must drive choreographic motion at fine temporal resolution, and such rhythmic coupling is not captured by unimodal metrics or generic audiovisual consistency scores used in current evaluation practice. We introduce TMD-Bench, a benchmark for text-driven music-dance co-generation that assesses systems across unimodal generation quality, instruction adherence, and cross-modal rhythmic alignment. The benchmark integrates computable physical metrics with perceptual multimodal judgments, and is supported by a curated rhythm-aligned music-dance dataset and a fine-grained Music Captioner for structured music semantics. TMD-Bench further reveals that (i) modern commercial audio-visual models, such as Veo 3 and Sora 2, produce high-quality music and video, while rhythmic coupling remains less consistently optimized and leaves room for improvement, and (ii) our unified baseline RhyJAM trained on rhythm-aligned data achieves competitive beat-level synchronization while maintaining competitive unimodal fidelity. This presents prospects for building next-generation music-dance models that explicitly optimize rhythmic and kinetic coherence.

A key challenge in multi-agent reinforcement learning (MARL) lies in designing learning signals that effectively promote coordination among agents. Designing such signals necessitates the ability to quantify the true, long-term causal influence between agents. To address this, we introduce Multi-step Advantage-Gated Interventional Causal MARL (MAGIC), a framework that extracts multi-step causal influences between agents and selectively converts them into intrinsic rewards. MAGIC uses causal intervention with conditional mutual information to quantify long-horizon agent influence, and introduces an advantage-based gating mechanism to ensure exploration is directed toward beneficial, goal-aligned behaviors. Experiments across multiple standard MARL benchmarks and task families, including MPE and SMAC/SMACv2, demonstrate that MAGIC outperforms state-of-the-art methods by a significant margin, achieving an improvement of at least 10.1% in the main evaluation metric.

This paper presents a Koopman-based framework for early outbreak detection and intervention selection in a multi-agent epidemic simulation. Agents exhibit mobility patterns, heterogeneous susceptibility, immunity-dependent viral load progression, and local transmission through co-location. The goal of the simulation is to study near-critical epidemic regimes in which small changes in exposure or timing can alter the final outcome. Aggregate daily observables from early trajectory windows are encoded into a low-dimensional Koopman latent space whose approximately linear evolution supports short-horizon forecasting and outbreak risk estimation. These representations are combined with a random forest classifier trained to predict whether the final attack rate exceeds a major outbreak threshold. Experiments near the system tipping points show strong early warning performance, with Koopman-derived features contributing to class separation. Counterfactual analysis further shows that minimal interventions, such as keeping a single selected agent at home for one day, can reduce attack rates and, often, shift the trajectory below the outbreak threshold.

Quantum software development has largely focused on algorithms, with limited attention to software architecture. As computing moves toward hybrid quantum-classical systems, this gap limits scalability, reusability, and engineering rigor. This study introduces a component-based quantum software architecture framework (QSAF) for hybrid quantum-classical software systems, enabling developers to transition from circuit-level design to system-level reasoning. We identified 34 reusable quantum circuit primitives across seven functional categories and reinterpreted them as architectural components with explicit interfaces and design-relevant constraints. These components are further characterized using non-functional dimensions such as circuit depth, error sensitivity, and information flow, enabling a structured analysis of design trade-offs. The proposed QSAF framework establishes a multi-level abstraction hierarchy linking quantum gates, circuit primitives, algorithmic structures, and hybrid system architectures. Through this approach, common workflows, particularly hybrid quantum-classical workflows such as variational quantum algorithms, can be systematically decomposed, compared, and optimized. By making the architectural structure and trade-offs explicit, this study provides a foundation for quantum software engineering, supporting modular design, reuse, and informed architectural decision-making in quantum application development.

Integration of Answer Set Programming (ASP) with neural networks has emerged as a promising tool in Neuro-symbolic AI. While existing approaches extend the capabilities of ASP to real world domains, their reasoning pipelines depend on classical solvers, which is a bottleneck for scalability. To tackle this problem, we propose a new method to compute stable models, called decision-propagation (DProp), which alternates falsity decisions and truth propagations. Successful DProp computations are shown to capture the stable model semantics. We then develop Neural DProp (NDProp), a differentiable extension of DProp with neural computation for decisions and fuzzy evaluation for propagations. We evaluate the capabilities of NDProp for learning decision heuristics as well as neuro-symbolic integration, and compare it with existing neuro-symbolic approaches. The results show that NDProp can learn to efficiently compute stable models, and it improves accuracy and scalability on neuro-symbolic benchmarks.

Confidence calibration has been dominated by the Expected Calibration Error (ECE), a linear metric that counts calibration offset equally regardless of the confidence level at which it occurs. We show that ECE can remain small even under arbitrarily large overconfidence risk, so we propose Calibrated Size Ratio (CSR) instead, an interpretable metric that equals 1 under perfect calibration, from which we derive the risk probability $P_{\mathrm{risk}}$ that quantifies the statistical evidence for overconfidence. We further argue that overconfidence risk assessment must be complemented by a measure of discriminative value: whether the assigned confidences actively distinguish correct from incorrect predictions. We show that confidence-weighted accuracy $\mathrm{cwA}$ is the natural such complement, and that confidence-weighting extends to all standard classification metrics. In particular, we prove that the confidence-weighted AUC (cwAUC) captures the information about calibration while the classical AUC cannot. We validate the proposed indicators on several synthetic confidence distributions under multiple controlled calibration profiles and on fifteen real datasets with and without post-hoc calibration. Experiments demonstrate that CSR achieves near-perfect sensitivity and specificity across all tested conditions.

Efficient radar resource allocation is a fundamental yet computationally challenging problem, as optimal solutions typically require iterative optimization with high complexity. Motivated by the need for real-time scheduling, robust generalization, and low data dependency, this paper proposes a novel paradigm that leverages large language model (LLM)-guided evolutionary search (AlphaEvolve) to autonomously discover a closed-form power allocation solution for multi-target tracking. The approach encodes high-dimensional radar states into physically inspired features, then evolves a compact and interpretable scoring function, which is transformed to feasible power allocations via a deterministic constraint-satisfying transformation. Extensive experiments demonstrate that the discovered closed-form solution achieves near-optimal tracking accuracy (average relative performance loss of only $1.51\%$), reliable generalization across diverse scenarios and target counts, and over three orders of magnitude speedup compared to conventional iterative solvers. These results highlight the potential of LLM-guided symbolic search to revolutionize not only radar resource management but also broader classes of engineering optimization problems.

As artificial intelligence (AI) models quickly spread and become more advanced, they are requiring an ever-increasing amount of data and compute capability, leading to a significant energy cost. Training and inference of AI models including the large language models (LLMs) and deep neural networks (DNNs) are contributing to a large carbon footprint owing to the massive amount of memory they consume in data centers. In this article, we present a generalized framework that quantifies these energy costs incurred to the environment. This framework provides a foundational quantification of AI's ecological footprint, facilitating the development of sustainable architectural strategies for future models.

A common design pattern in high-quality music generation is to handle structure and fidelity in different representation spaces: a generator first models high-level structure, followed by diffusion-based or neural decoding stages that reconstruct fine details. In this work, we explore an alternative view: both may be progressively modeled within a single deep acoustic-token hierarchy. To study this, we build a 64-layer residual vector quantization (RVQ) acoustic representation and propose a two-stage coarse-to-fine generation framework. A backbone model first generates coarse acoustic tokens for the full track, and a super-resolution model then completes finer tokens within the same acoustic token space. The super-resolution stage works at full-track scale and refines tokens layer by layer while running in parallel over time, leading to a fixed 62-step inference process. To jointly improve lyric alignment and fine-detail reconstruction, we further introduce hybrid-attention training: the alignment objective uses causal attention, while layer-wise refinement uses full attention. A key finding is that text--vocal alignment can emerge within pure acoustic-token language modeling, without requiring a separate semantic token stage. Moreover, initializing the super-resolution model from the trained backbone significantly improves convergence and final quality. Taken together, our results suggest that high-quality music generation can be effectively pursued without separating structure and fidelity into heterogeneous representation spaces. Instead, both can be progressively modeled within a unified acoustic-token hierarchy, pointing toward a simpler and more unified path to high-quality music generation.

Constructing controllable visual data is a major bottleneck for image editing and multimodal understanding. Useful supervision is rarely produced by a single rendering pass; instead it emerges through iterative generation, inspection, correction, filtering, and export. We present DataEvolver, a closed-loop visual data engine that organizes this process around explicit goals, persistent artifacts, bounded corrective actions, and acceptance decisions. DataEvolver supports multiple artifact types, including RGB images, masks, depth maps, normal maps, meshes, poses, trajectories, and review traces. In the current release, the system operates through two coupled loops: generation-time self-correction within each sample and validation-time self-expansion across dataset rounds. We validate the framework on an image-level object-rotation setting. With a fixed Qwen-Edit LoRA probe, our final Ours+DualGate model outperforms both the unadapted base model and a public multi-angle LoRA on SpatialEdit and a held-out evaluation set. Ablations show a consistent improvement path from scene-aware generation to feedback-driven correction and dual-gated validation. Beyond the released rotation data, our main contribution is a reusable framework for building visual datasets through explicit goal tracking, review, correction, and acceptance loops.

Autonomous navigation and obstacle avoidance remain a core challenge of modern Unmanned Aerial Vehicles (UAVs). While traditional control methods struggle with the complexity and variability of the environment, reinforcement learning (RL) enables UAVs to learn adaptive behaviors through interaction with the environment. Existing research with RL prioritizes the mission success at the expense of mission time and safety of UAVs. This study integrates Potential Based Reward Shaping (PBRS) with Control Lyapunov Functions (CLF) and Control Barrier Functions (CBF) to simultaneously optimize mission time and ensure formal safety guarantees. An RL model is trained in a generalized simple environment, then used in complex scenarios incorporating a CLF-CBF-QP filter without further training. Experimental results in simulated environments demonstrate a significant reduction in mission time and outstanding performance in complex environment.

Procedural Content Generation (PCG) enables game content to be created algorithmically without direct manual level-design effort, but it introduces a serious evaluation problem: generated content may become unbalanced, blocked, repetitive, or technically unsolvable. This paper presents Momentum, an endless-runner game that integrates runtime terrain generation, environment object spawning, and autonomous agent-based evaluation into a single gameplay loop. Ground tiles and environmental objects are generated dynamically as the player advances, object placement follows a constraint-driven mechanism inspired by Wave Function Collapse (WFC), and the runtime navigation surface is rebuilt asynchronously to remain consistent with the streamed environment. Two autonomous evaluation agents move ahead of the player and inspect the generated path: an aerial scanner that examines the corridor geometrically, and a ground-traversal agent that validates the same region from a navigational perspective. The evaluation pipeline combines ray casting, volumetric physics sweeps, obstacle-layer filtering, and structured crash reporting to identify problematic generated scenarios before they reach the player. The work demonstrates how generation and validation can be unified within the same runtime loop, rather than treating evaluation as a separate offline pass. Around this implementation, the paper formulates a measurable evaluation framework along the canonical PCG axes of playability, diversity, controllability, and runtime performance, derives a structural saturation bound on the spawner from its own placement constraints, and quantifies the per-segment scanning cost of the agents from first principles.

Adversarial imitation learning (AIL), a prominent approach in imitation learning, has achieved significant practical success powered by neural network approximation. However, existing theoretical analyses of AIL are primarily confined to simplified settings, such as tabular and linear function approximation, and involve complex algorithmic designs that impede practical implementation. This creates a substantial gap between theory and practice. This paper bridges this gap by exploring the theoretical underpinnings of online AIL with general function approximation. We introduce a novel framework called optimization-based AIL (OPT-AIL), which performs online optimization for reward learning coupled with optimism-regularized optimization for policy learning. Within this framework, we develop two concrete methods: model-free OPT-AIL and model-based OPT-AIL. Our theoretical analysis demonstrates that both variants achieve polynomial expert sample complexity and interaction complexity for learning near-expert policies. To the best of our knowledge, they represent the first provably efficient AIL methods under general function approximation. From a practical standpoint, OPT-AIL requires only the approximate optimization of two objectives, thereby facilitating practical implementation. Empirical studies demonstrate that OPT-AIL outperforms previous state-of-the-art deep AIL methods across several challenging tasks.

An evolution of Wireless Communications towards 5G and beyond provides improved user experience in terms of quality of services. Understanding and estimating Channel information plays crucial role in providing better user experience. Traditional methods of channel estimation involves periodically sending pilots (known signals), estimating channel and send back estimated channel information to the BS which increases computational complexity and communication complexity. Hence, we focus on data driven approach for channel estimation. This work can be deployed as Digital twin in 5G and beyond wireless networks. In this work, we explore a channel estimation mechanism at 7GHz frequency band for a given user location. This work involves data generation using Ray tracing mechanism and Machine learning model training that contains feature variables such as transmitter location, user location and target variable as channel coefficient . We explored Linear Regression, Support Vector Regression and Decision Tree Regression. We found via simulations that Linear Regression performs (with MAE of $\mathbf{7.5155\times10^{-5}}$ and RMSE of $\mathbf{9.2861\times10^{-5}}$) better than Support Vector Regression and Decision Tree Regression.

Addressing the diverse fault morphologies, complex dependencies, and time-varying operational states in microservice distributed systems, this paper proposes a distributed fault discrimination model based on temporal graph neural networks. This model characterizes the microservice operation process as a dynamic graph sequence evolving, and performs joint representation learning of temporal modeling and structural interactions within a unified framework. First, service-level multi-source observation signals are aligned and characterized to construct node feature sequences and their corresponding time-dependent dependencies. Then, a temporal coding module is introduced to extract the dynamic evolution representation of service states, and at each time step, attention-based structured message passing is used to characterize dependency interactions and propagation associations, forming a structure-enhanced temporal node representation. Furthermore, a dual readout mechanism is employed to aggregate the node and temporal dimensions, obtaining a system-level global representation and outputting the fault category distribution. Finally, supervised learning objectives are used to optimize model parameters, enabling the model to learn stable discrimination evidence under complex interactions and multi-source noise conditions. Comparative experimental results show that the proposed method achieves superior performance on multiple evaluation metrics, validating the effectiveness of jointly modeling temporal evolution and dependency structures in improving the distributed fault discrimination capability of microservices.

We consider the problem of two-sample testing in a semi-supervised setting with abundant unlabeled covariate data. Standard two-sample tests neglect covariate information, which has the potential to significantly boost performance. However, incorporating covariates potentially breaks the exchangeability assumption under the null, which further complicates a calibration procedure. To address these issues, we propose a semi-supervised method that produces a test statistic with asymptotic normality, while effectively integrating additional information from covariates. Our test is straightforward to calibrate due to the asymptotic normality under the null and achieves asymptotic power that is often much higher than existing kernel tests without covariates. Furthermore, we formally show that the proposed method is consistent in power against fixed and local alternatives. Simulations confirm the practical and theoretical strengths of our approach.

Vision-Language-Action (VLA) models have emerged as a powerful paradigm for embodied intelligence, enabling robots to perform tasks based on natural language instructions and current visual input. However, existing VLA models struggle with long-horizon tasks due to compounding errors. Prior methods decompose tasks into subtasks of fixed granularity, which cannot adapt to the varying complexity of execution states, limiting their robustness in long-horizon tasks. To overcome this, we introduce Anticipation Model, which adaptively and recursively generates future subgoals. This model continuously adapts as the task unfolds, adjusting future subgoals in response to evolving dynamics, facilitating more reliable planning paths. Building on this concept, we propose Anticipation-VLA, a hierarchical VLA model that leverages the anticipation model to generate actionable subgoals that guide VLA policy execution. We implement Anticipation-VLA with finetuning a Unified Multimodal Model (UMM) for high-level subgoal generation and a goal-conditioned VLA policy for low-level action execution. Experiments in both simulated and real-world robotic tasks demonstrate the effectiveness of Anticipation-VLA, highlighting the importance of adaptive and recursive subgoal generation for robust policy execution.

An auditor instructs an AI assistant: "open each file individually using the Read tool -- no scripts, no agents." The AI replies "Yes" -- then issues a single batched call summarizing all fifty files at once. We call this the Compliance Gap: a third, orthogonal axis of AI honesty distinct from factual truthfulness and rhetorical substance. Three questions: does this verbal-behavioral disconnect exist (existence); can any text-only observer recover it (detectability); what infrastructure does AI deployment need (remedy)? Some 75 benchmarks (IFEval, SWE-bench, BFCL, COMPASS, SpecEval) measure outcome fidelity; none measures process fidelity. Theorem 1 shows the gap is structurally inevitable under RL that rewards text without observing behavior. Theorem 2, via the Data Processing Inequality, shows it is undetectable from text alone -- by any human or LLM observer, present or future. Thirteen experiments and 2,031 sessions on six frontier models confirm both predictions. Under default framing, all six exhibit instruction compliance rates of 0% -- Claude Sonnet 4 verbally agrees ten out of ten times then bypasses in all ten. The gap is selective: 97% compliance where rationale is rewarded (audit trails), 0-4% where it is not (file reading, privacy masking); removing delegation tools raises compliance to 75% (Cohen's d = 2.47), confirming environmental affordance rather than weight-encoded failure. Nine blinded human raters achieve Fleiss' kappa = 0.130 and correctly identify zero of fifteen compliant sessions, exactly as Theorem 2 predicts. Where humans show 47% intention-behavior gaps in psychology and 96.5pp gaps in surgical audits, RLHF-trained models approach 100% under default conditions -- a regime warranting its own measurement infrastructure. We release BS-Bench: the first open benchmark for process compliance, with seven tool-call-log audit metrics and a public leaderboard.

The increasing prevalence of software vulnerabilities highlights the need for effective Automatic Vulnerability Repair (AVR) tools. While LLM-based approaches are promising, they struggle to incorporate structured security knowledge from sources like CWE and NVD. Current methods either use this information superficially by concatenating the CWE-ID into the input prompt, yielding negligible benefits, or rely on few-shot learning with rigid, non-generalizable examples, which limits their effectiveness in real-world scenarios. To address this gap, we propose VulKey, an LLM-based AVR framework that leverages a hierarchical abstraction of expert knowledge to guide patch generation. Our novel three-level abstraction formulates repair strategies in terms of CWE type, syntactic actions, and semantic key elements. This approach captures the essence of a security fix with greater generality than concrete examples and more semantic richness than traditional syntax-based templates, overcoming the coverage limitations of prior methods. VulKey is implemented as a two-stage pipeline: first, expert knowledge matching predicts an appropriate repair pattern for the vulnerability; second, repair code generation uses a pattern-guided, fine-tuned LLM to produce secure patches. On the real-world C/C++ dataset PrimeVul, VulKey achieves 31.5% repair accuracy, surpassing the best baseline by 7.6% and outperforming leading tools such as VulMaster and GPT-5. Moreover, VulKey demonstrates cross-language and cross-model generalizability, with state-of-the-art performance on the Java benchmark Vul4J. These results underscore the importance of structured expert knowledge in advancing AVR effectiveness. Our work demonstrates that explicitly modeling and integrating expert security knowledge through hierarchical patterns is a crucial step toward building more effective and reliable AVR tools.

Multimodal large language models (MLLMs) have revolutionized the landscape of AI, demonstrating impressive capabilities in tackling complex vision and audio-language tasks. However, a critical challenge remains: these models often suffer from hallucinations, generating outputs that diverge from the provided perceptual inputs. This tendency stems from an inherent imbalance in modality utilization during inference, where the dominance of textual tokens undermines the potential of perceptual inputs. As a result, the model frequently resorts to textual language priors at the expense of grounded evidence. To tackle this issue, we propose Learning Inference-time Modality Enhancement (LIME), a training-free framework designed to bolster multimodal grounding by explicitly enhancing modality usage during decoding. LIME leverages Layer-wise Relevance Propagation (LRP) to quantify token-level contributions and defines a relevance-based objective that promotes increased reliance on perceptual inputs. This objective is enforced through inference-time updates to the model's key-value representations, without modifying model parameters or requiring additional training data. We evaluate LIME across multiple multimodal benchmarks in both vision and audio domains, demonstrating consistent reductions in hallucinations and enhanced grounding while preserving generation quality. Further analysis shows that LIME increases modality contribution and produces more localized and semantically aligned relevance patterns.

Mediation analysis has traditionally focused on outcome-level summary contrasts, such as mean effects, which may obscure substantial distributional changes induced by complex and nonlinear causal mechanisms. We propose Distributional Causal Mediation Analysis (DCMA), a generative learning framework for identifying and estimating treatment effects on entire outcome distributions transmitted through multiple mediators. DCMA learns conditional generative models for the mediators and the outcome, recovering the relevant conditional distributions from observational data. Leveraging the identification formulas, it reconstructs interventional outcome distributions via Monte Carlo forward simulation by noise resampling, enabling the capture of both classical summary effects and rich distributional contrasts such as energy distance and the Wasserstein distance. Analytical error bounds are derived to decompose how estimation errors in the learned conditional models propagate to the reconstructed interventional outcome distributions. The empirical effectiveness of DCMA is demonstrated through numerical experiments and real-world data applications.

Large multimodal model-based Multi-Agent Systems (MASs) enable collaborative complex problem solving through specialized agents. However, MASs are vulnerable to infectious jailbreak, where compromising a single agent can spread to others, leading to widespread compromise. Existing defenses counter this by training a more contagious cure factor, biasing agents to retrieve it over virus adversarial examples (VirAEs). However, this homogenizes agent responses, providing only superficial suppression rather than true recovery. We revisit these defenses, which operate globally via a shared cure factor, while infectious jailbreak arise from localized interaction behaviors. This mismatch limits their effectiveness. To address this, we propose a training-free Foresight-Guided Local Purification (FLP) framework, where each agent reasons over future interactions to track behavioral evolution and eliminate infections. Specifically, each agent simulates future behavioral trajectories over subsequent chat rounds. To reflect diversity in MASs, we introduce a multi-persona simulation strategy for robust prediction across interaction contexts. We then use response diversity as a diagnostic signal to detect infection by analyzing inconsistencies across persona-based predictions at both retrieval-result and semantic levels. For infected agents, we apply localized purification: recent infections are mitigated via immediate album rollback, while long-term infections are handled using Recursive Binary Diagnosis (RBD), which recursively partitions the image album and applies the same diagnosis strategy to localize and eliminate VirAEs. Experiments show that FLP reduces the maximum cumulative infection rate from over 95% to below 5.47%. Moreover, retrieval and semantic metrics closely match benign baselines, indicating effective preservation of interaction diversity.

Existing auto-bidding algorithms in digital advertising often treat the value of an ad opportunity as the revenue obtained when an ad is shown and/or clicked, and bid accordingly. This can lead to wasteful spending because the true value is the marginal gain from paid exposure: even without winning a sponsored slot, an advertiser may still earn revenue via an organic search result (e.g., on Google or Amazon). Motivated by recent work, we model ad value as a treatment effect--the outcome difference between winning and losing the auction--and study online learning for bidding in second-price (Vickrey) auctions under this causal perspective. We develop algorithms that attain rate-optimal regret under several feedback models. A key ingredient exploits the information revealed by the second-price payment rule, which strictly improves regret relative to analogous learning problems in first-price auctions.

We study linear dueling bandits in volatile environments characterized by the simultaneous presence of post-serving contexts, delayed feedback, and adversarial corruption. Feedback is subject to unknown stochastic or adversarial delays and a cumulative corruption budget $\mathcal{C}$. To address these challenges, we propose \term, which integrates a learned approximator that predicts post-serving contexts from pre-serving information. It further employs an adaptive weighting strategy that clips feature vectors to mitigate the impact of corrupted and delayed observations simultaneously. Under standard regularity conditions and a parametric post-serving mapping, we rigorously establish that our algorithm is delay-regime-agnostic, achieving a regret upper bound of $\widetilde{\mathcal{O}}(d(\sqrt{T} + \mathcal{C} + \mathcal{D}))$, where $d$ is the total feature dimension and $\mathcal{D}$ encapsulates the delay complexity. Crucially, our analysis reveals an additive cost structure between corruption and delay, avoiding the multiplicative degradation typical of prior works. We further establish lower bounds that nearly match our upper bounds up to a $\sqrt{d}$ factor for adversarial delays in the absence of post-serving contexts.

Grounding is the collaborative process of establishing mutual belief sufficient for the current communicative purpose. While static grounding maps language to a shared, externally observable context, dynamic grounding is a joint activity where meaning is negotiated through interaction. Current multi-agent Large Language Model (LLM) benchmarks focus on static, one-shot tasks, overlooking the ability to repair grounding breakdowns across turns. We introduce an iterated, multi-turn negotiation game in which two agents allocate shared resources toward private projects with verifiable jointly optimal outcomes. While individual agents can identify Pareto-optimal allocations in isolation, agent dyads consistently fail to reach them across open- and closed-source models. Our investigation reveals four failure modes: (1) coordination degrades when shared interaction history is absent; (2) yet accumulated context can itself become a liability through stubborn anchoring, where initial proposals are treated as axiomatic rather than negotiable; (3) a reliance on perfunctory fairness (equal resource splits) over reward-maximizing coordination; and (4) failures in referential binding, where agents lose track of commitments across turns. These results highlight dynamic grounding as a critical and understudied axis of multi-agent coordination. Our framework decomposes the coordination gap into measurable components: the oracle baseline establishes that the gap is not attributable to individual reasoning limitations; the no-talk baseline establishes that communication is necessary; and a full-transparency intervention establishes that information exchange alone is insufficient: the bottleneck lies in the interactive processes of joint plan formation, commitment, and execution that constitute dynamic grounding.

Large Reasoning Models achieve strong performance on complex tasks but remain prone to hallucinations, particularly in long-form generation where errors compound across reasoning steps. Existing approaches to improving factuality, including abstention and factuality-driven optimization, follow a \emph{coupled exploration-commitment} paradigm, in which intermediate reasoning is unconditionally propagated to the final output, limiting fine-grained control over information selection and integration. In this paper, we propose an \textbf{Exploration-Commitment Decoupling} paradigm that disentangles knowledge exploration from final commitment, enabling models to explore with awareness while answering cautiously. We instantiate the paradigm with \textbf{Calibration-Aware Generation (CAG)}, a framework that equips models with end-to-end, calibration-aware generation capabilities, by augmenting intermediate reasoning with calibrated reliability estimates and prioritizing reliable content in final outputs. Across five long-form factuality benchmarks and multiple model families, CAG improves factuality by up to 13%, while reducing decoding time by up to 37%. Overall, our work highlights decoupling as a principled approach for more reliable long-form generation, offering directions for trustworthy and self-aware generative systems.

Language data are increasingly acquired and governed as assets, yet platforms often price candidate resources before knowing their true privacy or access costs. We study online pricing for governed language data assets under cost uncertainty. At each round, a platform observes an NLP task, a candidate asset, and a coarse cost estimate, may pay for a refined cost signal, posts a price, and receives safe net revenue. We introduce \textsc{NH-CROP}, a clipped robust pricing framework with a no-harm information-acquisition gate. The method compares direct pricing, risk-aware pricing, and verify-then-price, and acquires information only when its estimated decision value exceeds the best no-verification alternative. Across synthetic, real-proxy, and downstream-utility-grounded benchmarks, clipped \textsc{NH-CROP} variants improve or remain competitive with price-only and risk-aware baselines. Causal ablations show that paid verification is not the main source of gains in real-proxy and utility-grounded settings: the strongest learned policies often choose not to verify. Oracle and high-decision-value diagnostics show that refined cost information can still have substantial local value. Overall, governed language-data platforms should calibrate pricing under uncertain access costs first and verify only when information is cheap and decision-actionable.

An agentic-AI runtime issues tool calls, sends messages, and actuates devices on behalf of an LLM. Catching the four ways an action can diverge from its audit record -- F1 gate-bypass, F2 audit-forgery, silent host failure, F4 wrong-target, -- is a load-bearing safety property of any such runtime. We show that upstream OpenClaw, the most engineered single-user agentic-AI gateway in public release, catches none of them: recall is 0.000 on every cell of every confusion matrix, on a 1600-sample template baseline through OpenClaw's actual production command-line interface (CLI) and on a ten-LLM cross-model generalisation run. Detecting F1--F4 requires seven specific runtime structures absent from OpenClaw's source tree: a biconditional checker, a hash-chained audit log, an extension admission gate, a two-layer egress guard, a Bell-LaPadula classification policy, a module-signing trust root, and a bootstrap seal. enclawed-oss -- an MIT-licensed drop-in fork that ships all seven -- reaches $P = R = F_1 =$ accuracy $= 1.000$ on the same input. The gap is structural, not parametric: a six-line append-only widening of enclawed-oss's data-loss-prevention (DLP) regex catalog raises per-channel F3 detection by 14.6\% net at unchanged precision; the same edit on OpenClaw has nowhere to land. The harness deliberately exercises real Discord and Telegram channels -- plugin categories the first enclawed release deleted as unsafe -- to show F1--F4 detection extends to those previously-unsafe extensions. With architectural superiority for security and feature parity for extensions, we argue that unhardened agentic-AI runtimes are architecturally obsolete: a strictly better alternative exists, is adoptable today, and the gap requires re-architecture rather than configuration. We invite reviewers to apply the harness to any candidate runtime.

As large language models (LLMs) are increasingly deployed in real-world systems, they must support post-hoc removal of specific content to meet privacy and governance requirements. This motivates selective unlearning, which suppresses information about a particular entity or topic while preserving the LLM's general utility. However, most existing LLM unlearning methods require access to the original training corpus and rely on output-level refusal tuning or broad gradient updates, creating a tension among unlearning strength, non-target preservation, and data availability. We propose Geometric Unlearning (GU), an approach that operates directly on the model's prompt-time planning states without access to the original training corpus. GU distills a compact, low-rank geometry of desired safe behavior from a small set of safe reference prompts, and uses lightweight anchor-in-context synthetic prompts to trigger localized, projection-based alignment of hidden planning representations to this safe geometry. A teacher-distillation regularizer on synthetic non-target anchors further reduces collateral drift. Across privacy-oriented unlearning benchmarks (ToFU and UnlearnPII), GU achieves strong target suppression with minimal impact on non-target performance, demonstrating that effective unlearning can be achieved with minimal synthetic data.

Vision-Language Models (VLMs) excel at grounded reasoning but remain prone to object hallucination. Recent work treats self-generated captions as a uniformly positive resource, yet we find that naively embedding one can degrade rather than help--dropping Qwen2.5-VL-3B accuracy on HallusionBench by nearly 10 points. Two structural properties explain this. First, captions anchor not only the model's final answer but also its reasoning trajectory and lexical choices. Second, caption errors are asymmetric: omissions vastly outnumber fabrications, yet each fabrication carries a much larger per-instance impact. A caption's usefulness is therefore a per-query property, not a per-corpus one. We propose GEASS (Gated Evidence-Aware Selective Steering), a training-free module that decides on each query how much of the caption the model consumes: it gates the caption by the clean path's confidence, weights it by the entropy reduction it produces, and raises the evidence bar when the two pathways disagree. Experiments on POPE and HallusionBench across four VLMs show that GEASS consistently improves over vanilla inference and contrastive decoding, with only two extra forward passes per query.

Large language models (LLMs) have achieved remarkable performance across diverse domains, yet their enormous computational and memory requirements hinder deployment in resource-constrained environments. Knowledge distillation offers a promising solution by transferring knowledge from a large teacher model to a smaller student model. However, existing distillation methods typically treat all tokens equally, ignoring the fact that different tokens contribute unequally to model decisions. This can lead to inefficient knowledge transfer and reduced learning effectiveness. To address this limitation, we propose an entropy-based adaptive distillation strategy that dynamically adjusts the training process at the token level. Our method leverages the teacher's output entropy to guide three aspects of distillation. Specifically, we introduce a token-level curriculum by dynamically shifting focus from low- to high-entropy tokens during training. We further adjust the distillation temperature based on token entropy to better capture teacher confidence patterns. Moreover, we employ a dual-branch architecture for efficient logits-only distillation on easy tokens and deeper feature-based distillation on difficult tokens. Extensive experiments validate the soundness and effectiveness of our method.

Generative Flow Networks (GFlowNets) learn to sample states proportional to an unnormalized reward. Despite their theoretical promise, practical training is often unstable, exhibiting severe loss spikes and mode collapse. To tackle this, we first assess the sensitivity of GFlowNet objectives, demonstrating that a small Total Variation (TV) distance between the learned and target distributions does not preclude unbounded training loss. Motivated by this mismatch, we establish converse guarantees by deriving loss-to-TV bounds that certify global fidelity from bounded trajectory balance losses. Lastly, we propose Stable GFlowNets, an algorithm that leverages our theoretical results to stabilize training, and empirically demonstrate improved training behavior and superior distributional fidelity.

Large language models (LLMs) are increasingly used for automated news credibility assessment, yet it remains unclear whether they apply even-handed standards across journalistic genres. We examine whether zero-shot LLMs are more likely to misclassify legitimate entertainment news as fake than legitimate hard news, using a within-dataset design on GossipCop from FakeNewsNet. Across four frontier models, we find a clear but model-specific genre asymmetry: DeepSeek-V3.2 and GPT-5.2 show false-positive-rate gaps of 10.1 and 8.8 percentage points, respectively (both $p < .001$), whereas Claude Opus 4.6 and Gemini 3 Flash show no comparable difference. A style-swap experiment yields only limited and inconsistent changes, suggesting that the asymmetry is not reducible to stylistic register alone. Prompt-based mitigation is likewise possible but not generic: framing the model as an entertainment-news fact-checker reduces false positives for DeepSeek-V3.2 by about 50\% without detectable recall loss, but offers little improvement for GPT-5.2. Exploratory qualitative coding further suggests two recurring error patterns in sampled false positives: treating private-life claims as inherently unverifiable and discounting entertainment journalism as an epistemically weaker genre. Taken together, these findings show that aggregate performance metrics can obscure structured false positives within legitimate journalism. We argue that LLM-based credibility assessment may not only evaluate truth claims but also differentially recognize the legitimacy of journalistic genres, and that evaluation should therefore include genre-stratified false-positive analysis alongside overall accuracy.

Sequential recommendation models often struggle to capture latent periodic patterns in user interests, primarily due to the noise inherent in time-domain behavioral data. While frequency-domain analysis offers a global perspective to address this, existing approaches typically treat user sequences in isolation, overlooking the crucial context of the target item. In this work, we present a novel empirical observation: user attention scores exhibit distinct spectral entropy distributions when conditioned on positive versus negative target items. Specifically, true user interests manifest as highly concentrated spectral patterns with lower entropy in the frequency domain, whereas irrelevant behaviors appear as high-entropy noise. Leveraging this insight, we propose the Frequency-Enhanced Deep Interest Network (FEDIN). FEDIN introduces a frequency-domain branch that utilizes a target-aware spectrum filtering mechanism to isolate these periodic interest signals. Extensive experiments on three public datasets demonstrate that FEDIN consistently outperforms state-of-the-art sequential recommendation baselines, demonstrating superior robustness against noise. We have released our code at: https://github.com/otokoneko/FEDIN.

Autoregressive video generation paradigms offer theoretical promise for long video synthesis, yet their practical deployment is hindered by the computational burden of sequential iterative denoising. While cache reuse strategies can accelerate generation by skipping redundant denoising steps, existing methods rely on coarse-grained chunk-level skipping that fails to capture fine-grained pixel dynamics. This oversight is critical: pixels with high motion require more denoising steps to prevent error accumulation, while static pixels tolerate aggressive skipping. We formalize this insight theoretically by linking cache errors to residual instability, and propose MotionCache, a motion-aware cache framework that exploits inter-frame differences as a lightweight proxy for pixel-level motion characteristics. MotionCache employs a coarse-to-fine strategy: an initial warm-up phase establishes semantic coherence, followed by motion-weighted cache reuse that dynamically adjusts update frequencies per token. Extensive experiments on state-of-the-art models like SkyReels-V2 and MAGI-1 demonstrate that MotionCache achieves significant speedups of $\textbf{6.28}\times$ and $\textbf{1.64}\times$ respectively, while effectively preserving generation quality (VBench: $1\%\downarrow$ and $0.01\%\downarrow$ respectively). The code is available at https://github.com/ywlq/MotionCache.

Existing large-scale sign language resources typically provide supervision only at the level of raw video-text alignment and are often produced in laboratory settings. While such resources are important for semantic understanding, they do not directly provide a unified interface for open-world recognition and translation, or for modern pose-driven sign language video generation frameworks: 1. RGB-based pretrained recognition models depend heavily on fixed backgrounds or clothing conditions during recording, and are less robust in open-world settings than style-agnostic pose-processing models. 2. Recent pose-guided image/video generation models mostly use a unified keypoint representation such as DWPose as their control interface. At present, the sign language field still lacks a data resource that can directly interface with this modern pose-native paradigm while also targeting real-world open scenarios. We present SignVerse-2M, a large-scale multilingual pose-native dataset for sign language pose modeling and evaluation. Built from publicly available multilingual sign language video resources, it applies DWPose in a unified preprocessing pipeline to convert raw videos into 2D pose sequences that can be used directly for modeling, resulting in a consolidated corpus of about two million clips covering more than 25 sign languages. Unlike many laboratory datasets, this resource preserves the recording conditions and speaker diversity of real-world videos while reducing appearance variation through a unified pose representation. Toward this goal, we further provide the data construction pipeline, task definitions, and a simple SignDW Transformer baseline, demonstrating the feasibility of this resource for multilingual pose-space modeling and its compatibility with modern pose-driven pipelines, while discussing the evaluation claims it can support as well as its current limitations.

Conversational Aspect-based Sentiment Quadruple Analysis (DiaASQ) needs to capture the complex interrelationships in multiple rounds of dialogues. Existing methods usually employ simple Graph Convolutional Networks (GCN), which introduce structural noise and fail to consider the temporal sequence of the dialogues, or use standard RoPE, which implicitly captures relative distances in a flat sequence but cannot clearly separate the token-level syntactic order from the utterance-level progression, and may suffer from the Distance Dilution problem. To address these issues, we propose a new framework that combines Thread-Constrained Directed Acyclic Graph (TC-DAG) and Discourse-Aware Rotary Position Embedding (D-RoPE). Specifically, TC-DAG filters out cross-thread noise based on thread constraints, maintains global connectivity through root anchoring, and incorporates the temporal sequence of the dialogues. D-RoPE aligns multi-layer semantics using dual-stream projection and multi-scale frequency signals, captures thread dependencies using tree-like distances, and alleviates the token-level Distance Dilution problem by incorporating utterance-level progressions. Experimental results on two benchmark datasets demonstrate that our framework achieves state-of-the-art performance.

Pareto set learning (PSL) is an emerging paradigm in multi-objective optimization that trains neural networks to map preference vectors to Pareto optimal solutions. However, existing PSL methods primarily focus on solving a single multi-objective optimization problem at a time. This limitation not only increases computational costs in multi-objective multitask optimization scenarios by requiring a separate model for each task, but also fails to exploit the inter-task correlations across tasks. To address this, we propose a Cross-tAsk correlation-aware Pareto Set Learning (CoAction) framework, which leverages task-aware transformer to handle multiple tasks simultaneously. Specifically, by assigning task-specific embedding vectors to individual tasks, the model effectively distinguishes between tasks while facilitating knowledge sharing among them. We utilize a Transformer encoder as the backbone architecture to leverage its self-attention mechanism for capturing complex task dependencies. The proposed approach is evaluated on comprehensive multitask test suites covering both benchmark problems and real-world applications, demonstrating effectiveness and competitive performance in Hypervolume, Range, and Sparsity.

AI products often route requests through version aliases, service tiers, tool choices, regional endpoints, fallback rules, or safety handling before responding. These routing steps are documented product surfaces in several widely used AI platforms and serving stacks. Routing helps AI services stay affordable, fast, and available at scale, and it shapes trust. Trust can break when routing changes the cost, quality, or accountability of a response without the user being able to tell what happened. "Which model answered?" is only part of the audit question. The runtime path matters. Adaptive AI systems should produce a runtime transparency artifact called the route receipt. A route receipt is a compact record of the route that served a request. It should capture enough material facts for people relying on the output to reconstruct important routing decisions without exposing proprietary internals or hidden reasoning. Route transparency should be part of model documentation. Model cards describe trained model artifacts, while route receipts describe the runtime conditions under which a particular answer was produced. The paper introduces the route-receipt concept, a minimal schema and redaction model, and a documentation-based survey of selected platforms showing that receipt fragments already exist without a portable per-answer record.

Contemporary systems serving large language models (LLMs) have adopted prefill-decode disaggregation to better load-balance between the compute-bound prefill phase and the memory-bound decode phase. Under this design, prefill workers generate a KV cache that must be transferred to decode workers before token generation can begin. With these workers residing on different physical systems, this transfer becomes a significant bottleneck to serving LLMs at scale. This bottleneck gets exacerbated for long-input and agentic workloads, which typically require long inputs. Existing lossless codecs are not well suited to this setting as they primarily target offline weight compression, rely on CPU-side, or use variable-length coding that decompresses fast but compresses too slowly for online use. SplitZip is a GPU-friendly lossless compressor for KV-cache transfer. It exploits redundancy in floating-point exponents of KV activations, encoding the most frequent exponent values with fixed-length codes, and encoding (position, value) pairs and value of rare exponents in an escape stream. An offline calibrated top-16 exponent codebook enables online encoding, while the regular dense path and sparse escape correction make both encoding and decoding efficient on GPUs. On real BF16 activation tensors, SplitZip achieves 613.3 GB/s compression throughput and 2181.8 GB/s decompression throughput, substantially outperforming prior lossless compressors on the latency-critical codec path. End-to-end transfer experiments show up to 1.32$\times$ speedup for BF16 KV-cache transfer, 1.30$\times$ speedup for TTFT and 1.23$\times$ increase on Request Throughput.

Jamming attacks are proliferating and pose a significant threat to the security of 5G and beyond networks. These attacks target 5G radio frequency (RF) domain and can disrupt the communication in wireless networks. While conventional machine learning and deep learning approaches demonstrate its potential for jamming detection, they typically require centralized data collection, compromising the privacy of user equipment (UEs). This work proposes a federated learning (FL)-based jamming detection framework that operates on over-the-air In-phase and Quadrature (IQ) samples extracted from Synchronization Signal Blocks (SSBs) in the RF domain. The framework enables collaborative model training across multiple UEs without sharing raw RF signal data. We adopt Federated Averaging (FedAvg) algorithm to train a 1D convolutional neural network (1DCNN) for effective detection of attacks. Numerical results demonstrate that the proposed FL framework achieves 97% accuracy and 97% F1-score, outperforming centralized baselines including MLP, 1DCNN, SVM, and logistic regression, while preserving the data privacy of all participating UEs

When copies of the same language model are prompted to debate, they produce diverse phrasings of one perspective rather than diverse perspectives. Multi-agent debate (MAD), and more broadly closed-system reasoning where agents iteratively transform each other's outputs, tends to preserve answer accuracy while degrading the reasoning behind those answers. We name the multi-agent case the Debate Trap and the broader phenomenon the Reasoning Trap, offering a programmatic theory of evidence-grounded reasoning failure.The framework has three parts: (i) SFS (Supported Faithfulness Score), a claim-level metric verifying decomposed atomic claims against provided evidence (decomposer-invariant rankings: Spearman rho=1.0); (ii) EGSR (Evidence-Grounded Socratic Reasoning), replacing adversarial argumentation with evidence-grounded inquiry; (iii) Theorem 1 (DPI Bound): under standard MAD, the chain E -> O^0 -> O^1 -> ... is Markov, and the Data Processing Inequality implies E[I(E;O^{t+1})] <= E[I(E;O^t)]. Three companion results -- open-system recovery (Theorem 2), EGSR accumulation (Lemma 2), and vote-aggregation floor (Proposition 1) -- partition multi-step LLM reasoning by its information-theoretic relationship to E. Across 16 conditions on SciFact (300 claims) and FEVER (1,000 claims), DebateCV (C13) preserves 88% of baseline accuracy while SFS drops 43%; majority-vote MAD (C15) reduces SFS to 1.7% of baseline (p < 10^{-6}, d = -0.96); EGSR recovers 98%. An R6 cohort study (Korean n=10x30 FEVER; English n=3x200 SciFact) finds inter-rater Fleiss kappa <= +0.018 with 0.8-1.4 Likert intra-rater shifts across language and domain -- the human agreement that faithfulness metrics have been calibrated against is not itself stable. We offer one falsifiable conjecture: any closed-system reasoning protocol preserving Theorem 1's Markov structure is, in expectation, subject to the same DPI bound.

Theoretical studies show that for any differentiable function on a compact domain, there exists a neural network that approximates both the function values and gradients. However, such a result cannot be used in practice since it assumes real parameters and exact internal operations. In contrast, real implementations only use a finite subset of reals and machine operations with round-off errors. In this work, we investigate whether a similar result holds for neural networks under floating-point arithmetic, when the gradient with respect to the input is computed by the automatic differentiation algorithm $D^\mathtt{AD}$. We first show that given a floating-point function $φ$ (e.g., a loss function), arbitrary function values and gradients can be represented by a floating-point network $f$ and $D^\mathtt{AD}(φ\circ f)$, respectively. We further extend this result: given $φ_1,\dots,φ_n$, $D^\mathtt{AD}(φ_i\circ f)$ can simultaneously represent arbitrary gradients while $f$ represents the target values, under mild conditions. Our results hold for practical activation functions, e.g., $\mathrm{ReLU}$, $\mathrm{ELU}$, $\mathrm{GeLU}$, $\mathrm{Swish}$, $\mathrm{Sigmoid}$, and $\mathrm{tanh}$.

Stochastic gradient methods are central to large-scale learning, yet their generalization theory typically relies on independent sampling assumptions. In many practical applications, data are generated by Markov chains and learning is performed in a decentralized manner, which introduces significant analytical challenges. In this work, we investigate the stability and generalization of decentralized stochastic gradient descent (SGD) and stochastic gradient descent ascent (SGDA) under Markov chain sampling. Leveraging a stability-based framework, we characterize how Markovian dependence and decentralized communication jointly influence generalization behavior. Our analysis captures the effects of network topology, Markov chain mixing properties, and primal-dual dynamics. We establish non-asymptotic generalization bounds for both algorithms, extending existing results on Markov stochastic gradient methods to decentralized and minimax settings.

Recent attacks show that behavioural unlearning of large language models leaves internal traces recoverable by adversarial probes. We characterise where this retention lives and show it can be surgically removed without measurable capability cost. Our central protocol is a leave-one-out cross-sequence probe that tests whether a memorisation signature generalises across held-out sequences. The signature is real and consistent across scale: memorisation-specific gaps of +0.32, +0.19, +0.30 on Pythia-70M, GPT-2 medium, and Mistral-7B; on Pythia-70M, the random-initialisation control collapses to -0.04 at the deepest layer where the pretrained signature peaks. The probe direction is causally separable from recall -- projecting it out collapses the signature locally (+0.44 -> -0.19) while behavioural recall barely changes -- and a probe trained on naturally memorised content does not classify fine-tuning-injected secrets, marking two representationally distinct regimes. We then introduce probe-geometry alignment (PGA), a surgical erasure that aligns activations along the probe's live readout direction at each depth. PGA drives the cross-sequence probe below random chance at all four scales tested (toy depth-4: 0.17; Pythia-70M: 0.07; Mistral-7B: 0.45; GPT-2 medium: 0.06 via MD-PGA k=2) and remains robust to six adversarial probe variants. Against a re-fitting attacker who trains a fresh probe on PGA-treated activations, we extend PGA adversarially, defeating the re-fit probe at every memorisation-relevant depth while preserving five zero-shot capability benchmarks within 2.8 percentage points per task (mean Δacc = +0.2pp). The cross-sequence signature is a real, causally separable, regime-specific property of pretrained representations -- removable below chance with a single rank-one intervention per depth at no measurable capability cost.

BIM models provide structured representations of building geometry, semantics, and topology, yet extracting specific information from them remains remarkably difficult. Current approaches translate natural language into structured queries by assuming a fixed data organization (static approach), which BIM heterogeneity eventually invalidates. We address this with a new paradigm, adaptive exploration, where an LLM-based agent iteratively executes code to extract information from a BIM model, discovering its structure at runtime instead of assuming it. We evaluate this approach on ifc-bench v2, an open-source BIM question-answering benchmark introduced alongside this work, comprising 1,027 tasks across 37 IFC models from 21 projects. A factorial ablation across two LLM capability levels and four augmentation strategies shows that adaptive exploration significantly outperforms static query generation across all configurations, regardless of the augmentation strategy. These results indicate that BIM heterogeneity is best addressed at the paradigm level, not by further optimizing static approaches.

A world model matters to an agent only through the state it constructs. That state must preserve some information, discard other information, and support some future function: prediction, control, planning, memory, grounding, or counterfactual reasoning. This paper treats world-model research as latent state design under sufficiency constraints. We propose a functional taxonomy that groups methods by what their latent state is for, rather than by architecture or application domain: predictive embedding, recurrent belief state, object/causal structure, latent action interface, grounded planning interface, and memory substrate. These roles expose distinctions that architecture-based groupings hide, including the gap between predictive sufficiency and control sufficiency, and the gap between passive video prediction and counterfactual action modeling. The taxonomy supports an evaluation framework that judges a model by the sufficiency constraint its latent state was built to satisfy. We compare methods along seven axes: representation, prediction, planning, controllability, causal/counterfactual support, memory, and uncertainty. We use the resulting matrix as a diagnostic for what a latent state preserves, discards, and enables. The conclusion that follows is that an actionable world model is the one whose state construction matches the task, not the one that preserves the most information.

In this paper, we propose Complex Diffusion Maps (CDM), a novel diffusion mapping framework that aims to reveal the dominant complex harmonics of high-dimensional data. Inspired by the local Gaussian kernel relevant to the heat equation and the nonlocal Schrödinger kernel relevant to the Schrödinger equation, we propose a unified family of $ω$-parameterized complex-valued kernels for the trade-off between local and nonlocal connections. We establish the theoretical foundation based on the operator spectrum theory, where the corresponding diffusion operator, diffusion distance, and complex harmonic maps are well-defined. An optimization-based interpretation of the maps is also developed, aiming to preserve angular structure in the complex diffusion space rather than relying solely on real-valued magnitude. We extensively evaluate CDM on both synthetic and real-world datasets. The complex-valued kernel amplifies differences among easily confusable samples, improving discriminative power over both linear and nonlinear methods based on real-valued kernels. CDM remains robust in high-noise settings, yielding a clearer eigengap that enhances spectral separation. For resting-state fMRI data, CDM captures more strongly correlated and nonlocal spatiotemporal dynamics. Without task-specific tuning, CDM achieves competitive performance on a public EEG sleep dataset, while maintaining high computational efficiency compared with both traditional machine learning and deep neural network approaches, highlighting its generality and practical value.

Long-horizon conversational agents rely on memory systems with increasingly sophisticated retrieval mechanisms. However, retrieved fragments are typically fed to the language model as unstructured text, lacking the relational, temporal, and thematic structures essential for complex reasoning. To bridge this reasoning gap, we introduce GRAVITY (\textbf{G}eneration-time \textbf{R}elational \textbf{A}nchoring \textbf{V}ia \textbf{I}njected \textbf{T}opological Memor\textbf{Y}), a plug-and-play structured memory module. GRAVITY extracts three complementary knowledge representations from raw conversational utterances: entity profiles grounded in relational graphs, temporal event tuples linked into causal traces, and cross-session topic summaries. At generation time, it injects these representations into the host system's prompt as structured anchoring contexts. This approach effectively synthesizes scattered evidence into a coherent, query-relevant context without requiring any architectural modifications to the host model. Extensive evaluations across five diverse memory systems on the LongMemEval and LoCoMo benchmarks demonstrate the efficacy of our approach. On average, GRAVITY improves LLM-judge accuracy by 7.5--10.1%. Gains are inversely correlated with baseline strength: the weakest host improves by 12.2% while the strongest still gains 3.8--5.7%. These findings establish structured context anchoring as a broadly effective, architecture-agnostic augmentation paradigm for long-horizon conversational memory.

We present MultiBreak, a scalable and diverse multi-turn jailbreak benchmark to evaluate large language model (LLM) safety. Multi-turn jailbreaks mimic natural conversational settings, making them easier to bypass safety-aligned LLM than single-turn jailbreaks. Existing multi-turn benchmarks are limited in size or rely heavily on templates, which restrict their diversity. To address this gap, we unify a wide range of harmful jailbreak intents, and introduce an active learning pipeline for expanding high-quality multi-turn adversarial prompts, where a generator is iteratively fine-tuned to produce stronger attack candidates, guided by uncertainty-based refinement. Our MultiBreak includes 10,389 multi-turn adversarial prompts, spans 2,665 distinct harmful intents, and covers the most diverse set of topics to date. Empirical evaluation shows that our benchmark achieves up to a 54.0 and 34.6 higher attack success rate (ASR)} than the second-best dataset on DeepSeek-R1-7B and GPT-4.1-mini, respectively. More importantly, safety evaluations suggest that diverse attack categories uncover fine-grained LLM vulnerabilities}, and categories that appear benign under single-turn can exhibit substantially higher adversarial effectiveness in multi-turn scenarios. These findings highlight persistent vulnerabilities of LLMs under realistic adversarial settings and establish MultiBreak as a scalable resource for advancing LLM safety.

Virtual screening performance depends heavily on the chosen docking and scoring methods. Recent AI-based tools such as DiffDock and NMDN have reported strong benchmark results, but their practical utility on realistic, experimentally-derived datasets remains unclear. Here we perform a large-scale evaluation on the LIT-PCBA library (15 targets, 578,295 ligand-target pairs with experimentally confirmed actives and inactives). We compare AutoDock-GPU and DiffDock for pose generation, followed by rescoring with GNINA and NMDN. We further evaluate rank-based consensus strategies and supervised machine learning models trained on docking features. GNINA rescoring of AutoDock-GPU poses (AutoDock-GNINA) emerged as the strongest single method with a median EF1% of 2.14. DiffDock-based approaches underperformed relative to AutoDock-GNINA, particularly on challenging targets such as OPRK1. Carefully designed consensus ranking improved robustness but did not surpass the best single scorer. Supervised ML re-ranking delivered the largest gains, achieving a median EF1% of 4.49 (+110% over AutoDock-GNINA). Our results highlight that even the best classical+ML hybrid workflows provide only modest early enrichment on realistic benchmarks. We conclude that no single docking method dominates across targets and that rigorously validated, cost-effective combinations with supervised re-ranking currently offer the most practical value for virtual screening.

Fall detection is a critical task in healthcare, particularly for elderly people. Timely fall detection and treatment can prevent severe injuries. Sensor-based activity data can be used to detect fall. However, this data are highly sensitive and raises significant privacy concerns. Existing privacy approaches apply uniform noise across all training samples, which affects the prediction performance. To address this limitation, we propose a Class-Aware Adaptive Differential Privacy (CA-ADP) framework integrated with a hybrid 3D Convolutional Neural Network and Bidirectional Long Short-Term Memory (3D CNN-BiLSTM) architecture. The CA-ADP mechanism dynamically adjusts the magnitude of noise added to gradients based on the class composition of each mini-batch. This process ensures privacy while mitigates performance degradation. We formally analyze the $(ε,δ)$-Differential Privacy guarantee and provide a privacy-utility trade-off analysis. The proposed method is evaluated on three public benchmark datasets, namely SisFall, UP-Fall, and MobiAct. The experimental results show that the proposed privacy model achieves improvements of 3.3\%, 8.5\%, and 7.5\% over the conventional privacy-based model in terms of F-score for the SisFall, UP-Fall, and MobiAct datasets, respectively. Comparisons with prior studies show that the CA-AD based framework achieves competitive performance and provides formal privacy guarantees, which are largely overlooked in existing studies. Wilcoxon signed-rank tests confirm that the proposed mechanism consistently outperforms conventional differential privacy. Those results establish the proposed CA-ADP framework as an effective approach to privacy-preserving fall detection in real-world healthcare settings.

Missing data imputation remains a fundamental challenge in modern data science, especially when uncertainty quantification is essential. In this work, we propose MissBGM, an AI-powered missing data imputation method via Bayesian generative modeling that bridges the expressive flexibility of neural networks with the statistical rigor of Bayesian inference. Unlike existing methods that often focus on point estimates or treat the missingness mechanism implicitly, MissBGM explicitly and jointly models the data-generating and missingness mechanisms, providing principled posterior uncertainty over imputations rather than a single point estimate. We develop a stochastic optimization framework with alternating updates among missing values, model parameters, and latent variables until convergence. Our theoretical analysis shows that estimates of missing values from MissBGM converge consistently under mild assumptions. Empirically, we demonstrate that MissBGM achieves superior performance over traditional imputers and recent neural network-based methods across extensive experimental settings. These results establish MissBGM as a principled and scalable solution for modern missing data imputation. The code for MissBGM is open sourced at https://github.com/liuq-lab/MissBGM.

Constraint Programming (CP) is a powerful paradigm for solving combinatorial problems, yet translating natural language problem descriptions into executable models remains a significant bottleneck. While Large Language Models (LLMs) show promise in automating this translation, they often struggle with subtle semantic errors in the absence of oracle validation at test time. To address this, we introduce CP-SynC (Constraint Programming modeling with Synthesized Checkers), a multi-agent workflow for zero-shot constraint modeling in MiniZinc. CP-SynC coordinates modeling agents that generate and refine candidate models and validation agents that synthesize semantic checkers to provide feedback on semantic correctness. To mitigate noise inherent in individual LLM outputs, CP-SynC explores multiple modeling trajectories in parallel and employs selection agents to select the final model via multi-agent evidence aggregation. Extensive experiments on a benchmark of 100 CP problems show that CP-SynC substantially outperforms existing baselines in MiniZinc modeling.

External priors of unknown reliability create a brittle trade-off in causal discovery: blind trust amplifies errors, blind rejection wastes signal. Real priors are also \emph{heterogeneously} reliable -- physical laws are trustworthy, LLM-suggested edges are speculative -- yet existing methods either ignore priors or impose them through globally uniform trust. We propose \textbf{PRCD-MAP}, a soft prior-consumption layer that assigns \emph{per-edge} trust to an imperfect prior and uses it to modulate a prior-aware $\ell_1$ penalty and prior-weighted $\ell_2$ regularizer in a MAP objective. Trust is calibrated by empirical Bayes on a Laplace-approximated marginal likelihood and propagated along the prior graph by an MLP, so that data-confirmed neighborhoods boost trust and contradictions suppress it. PRCD-MAP enjoys a population-level safety guarantee: it is $\varepsilon$-safe in expectation over the prior-generation distribution, with $\varepsilon = O(d^2/T)$ -- inheriting the oracle convergence rate. When the prior is uninformative, learned trust provably collapses to its floor and the method recovers a no-prior baseline. Empirically, on real CausalTime data PRCD-MAP exploits informative priors when present ($+0.123$ AUROC on AQI, $+0.043$ on Medical over PCMCI+), auto-attenuates on the anonymous-variable Traffic stress test, and retains a lead at $d{=}300$; against BayesDAG~\citep{annadani2023bayesdag} -- the closest soft-Bayesian baseline -- PRCD-MAP wins on every CausalTime dataset under a matched $W_0$-only protocol. A four-way ablation isolates each component: EB calibration and MLP trust propagation jointly carry the plurality of the gain, with positive sign on every dataset. Extensions to nonlinear (NAM) and cross-sectional settings show the calibrated-trust principle is setting-agnostic.

Dense temporal annotation of procedural activity videos is vital for action understanding and embodied intelligence but remains labor-intensive due to reactive tools. Each correction is treated as an isolated edit, limiting reuse of information on annotator uncertainty and model reliability. We introduce IMPACT-Scribe, a correction-driven framework for dense labeling that uses each correction to improve future human-machine collaboration. IMPACT-Scribe combines uncertainty-aware boundary scribble supervision, local proposal modeling, cost-aware query planning, structured propagation, and correction-driven adaptation. Experiments and a human study show that this closed-loop design improves labeling quality per effort, enhances boundary accuracy, and fosters better human-machine interaction over time. The code will be made publicly available at https://github.com/BanzQians/IMPACT_AS.

We present IMPACT-HOI, a mixed-initiative framework for annotating egocentric procedural video by constructing structured event graphs for Human-Object Interactions (HOI), motivated by the need for high-quality structured supervision for learning robot manipulation from human demonstration. IMPACT-HOI frames this task as the incremental resolution of a partially specified, onset-anchored event state. A trust-calibrated controller selects among direct queries, human-confirmed suggestions, and conservative completions based on empirical annotator behavior and evidence quality. A risk-bounded execution protocol, utilizing atomic rollback, ensures that human-confirmed decisions are preserved against conflicting automated updates. A user study with 9 participants shows a 13.5% reduction in manual annotation actions, a 46.67% event match rate, and zero confirmed-field violations under the studied protocol. The code will be made publicly available at https://github.com/541741106/IMPACT_HOI.

We propose Flow-Anchored Noise-conditioned Q-Learning (FAN), a highly efficient and high-performing offline reinforcement learning (RL) algorithm. Recent work has shown that expressive flow policies and distributional critics improve offline RL performance, but at a high computational cost. Specifically, flow policies require iterative sampling to produce a single action, and distributional critics require computation over multiple samples (e.g., quantiles) to estimate value. To address these inefficiencies while maintaining high performance, we introduce FAN. Our method employs a behavior regularization technique that utilizes only a single flow policy iteration and requires only a single Gaussian noise sample for distributional critics. Our theoretical analysis of convergence and performance bounds demonstrates that these simplifications not only improve efficiency but also lead to superior task performance. Experiments on robotic manipulation and locomotion tasks demonstrate that FAN achieves state-of-the-art performance while significantly reducing both training and inference runtimes. We release our code at https://github.com/brianlsy98/FAN.

The rapid growth of video content across domains such as surveillance, education, and social media has made efficient content understanding increasingly critical. Video summarization addresses this challenge by generating concise yet semantically meaningful representations, but existing approaches often rely on expensive manual annotations, struggle to generalize across domains, and incur significant computational costs due to complex architectures. Moreover, unsupervised and weakly supervised methods typically underperform compared to supervised counterparts in capturing long-range temporal dependencies and semantic structure. In this work, we propose TRIMMER (Temporal Relative Information Maximization for Multi-objective Efficient Reinforcement), a novel self-supervised reinforcement learning framework for video summarization. TRIMMER operates in two stages: it first learns robust representations via self-supervised learning and then performs spatio-temporal decision making through reinforcement learning guided by information-theoretic reward functions. Unlike prior approaches that rely on similarity-based objectives, our method introduces entropy-based metrics to capture higher-order temporal dynamics and semantic diversity, while computing rewards directly over selected frame indices to improve computational efficiency. Extensive experiments on standard benchmarks demonstrate that TRIMMER achieves state-of-the-art performance among unsupervised and self-supervised methods, while remaining competitive with leading supervised approaches, highlighting its effectiveness for scalable and generalizable video summarization.

Human cognition emerges from coordinated spiking dynamics in distributed neural circuits, where information is encoded via both firing rates and precise spike timing determined by brain rhythms. Inspired by this notion, we propose a brain-inspired learning primitive in which cognition-level neural synchrony emerges through iterative bottom-up and top-down interactions between micro-scale dynamics of spiking neurons and a macro-scale mechanism of oscillatory synchronization. Specifically, we model each parcel (e.g., a cortical region or an image pixel) in the target system as a spiking neuron embedded in a predefined connectivity scaffold. Low-level information is encoded in a spatiotemporal domain, where neurons are selectively grouped and fire spontaneously over time through self-organized dynamics. In the bottom-up route, oscillatory synchronization is formed from past spiking activity accumulated over a finite memory window. Since brain dynamics operate in a regime of partial and transient synchronization rather than global phase locking, we model oscillatory coordination using a time-delayed synchronization formulation, which enables a top-down modulation of heterogeneous neural spiking for a large-scale distributed system. Together, we devise a spiking-by-synchronization neural network (S2-Net) that uses rhythmic timing as a control mechanism for efficient information processing. Promising results have been achieved across a broad range of tasks, including neural activity decoding, energy-efficient signal processing, temporal binding and semantic reasoning.

We study homogeneous refinement operators \((Vγ)(t)=\sum_{j\in\mathbb Z}A_jγ(Mt-j)\), acting on compactly supported continuous piecewise linear curves \(γ:\mathbb R\to\mathbb R^p\), where \(M\ge2\) and only finitely many matrices \(A_j\in\mathbb R^{p\times p}\) are nonzero. We prove that the iterates \(V^nγ\) admit exact ReLU realizations of fixed width and depth \(O(n)\). The main new ingredient is an exact loop controller for the residual dynamics. Instead of propagating scalar residual surrogates, the construction transports the residual orbit by a forward-exact state on a polygonal loop. Scalar factors and digit selectors are then recovered from this loop state by complementary CPwL readouts. The loop seam is not removed, but its remaining ambiguity is confined to the final readout/selector stage, where it is harmless because the scalar atom is supported away from the seam. This gives a homogeneous \(M\)-ary vector-valued extension of the scalar binary refinable-function construction with a more geometric controller architecture. We also record crude exponential bounds on the network weights and biases. Affine forcing terms are handled by expanding affine iterates into finite sums of homogeneous iterates, giving exact fixed-width realizations with depth \(O(n^2)\), and anchored open curves reduce to compactly supported defects with affine anchor mismatch. We also describe homogeneous polygonal generators, including dragon-type examples and a self-intersecting Hilbert-type prototype in arbitrary dimension. The extended version includes stage-dependent forcing, finite-state stacking reductions, and further geometric constructions such as Koch-, Gosper-, Morton-, and connector-based Hilbert-type variants.

Reliable subnational population estimates are essential for applications, yet remain difficult where censuses are sparse, outdated or spatially coarse. Existing population-mapping workflows rely on hand-built geospatial covariates, such as settlement extent, night-time lights, and environmental conditions, which must be assembled and harmonised across scales and geographies. Geospatial foundation models offer an alternative by learning reusable representations of place from more multifaceted and heterogeneous data sources. Here, we benchmark Population Dynamics Foundation Model (PDFM) embeddings against the harmonised geospatial covariates for subnational population estimation in Brazil, Nigeria and the United States. Under geographically structured validation, PDFM increased predictive fit by a median of 20.1% (IQR: 10.0-33.2%, across country-model comparisons) reduction in unexplained variance, and reduced Kullback-Leibler divergence by 23.2% (9.2-26.2%). However, these gains were uneven. PDFM was most advantageous where the geospatial covariates weakly characterised settlement context, such as larger and less-developed subnational areas. Moreover, PDFM performance was scale-coupled with embeddings providing less flexible transfer across spatial aggregations than geospatial covariates. These findings showed that geospatial foundation-model representations of place can improve population estimation in data poor settings, but their benefits break down predictably under spatial scale mismatch, revealing a fundamental limitation of current geospatial AI.

Training-free AI text detection methods primarily rely on model log-probabilities, achieving strong performance through approaches like Binoculars and DNA-DetectLLM. However, these methods face a fundamental ceiling as models are optimized through RLHF to produce human-like probability distributions. We introduce an alternative detection signal based on character distribution signatures. We provide theoretical foundations showing that AI models, trained on massive domain-balanced corpora, approximate global character patterns while humans exhibit domain-specialized distributions, creating a "Wall of Separation" where human-AI divergence significantly exceeds AI-AI divergence. To enable systematic evaluation, we construct the Models-Domains-Temperatures-Adversarials (MDTA) benchmark comprising 642,274 prompt-aligned samples across 4 models, 5 domains, 3 temperature settings, and 3 adversarial strategies, substantially expanding the HC3 dataset with modern model responses, temperature variation, and adversarial augmentation. We introduce the Letter Distribution Score (LD-Score), demonstrating low correlation (r = 0.08-0.13) with perplexity methods. When integrated with DNA-DetectLLM, Binoculars and FastDetectGPT via a non-linear classifier, LD-Score yields consistent improvements in AUROC and F1, with particularly pronounced gains in specialized domains where vocabulary constraints amplify the detection signal. The MDTA dataset can be accessed at: https://huggingface.co/datasets/nsp909/MDTA.

We study AI alignment through the lens of law-and-economics models of deterrence and enforcement. In these models, misconduct is not treated as an external failure, but as a strategic response to incentives: an actor weighs the gain from violation against the probability of detection and the severity of punishment. We argue that the same logic arises naturally in agentic AI pipelines. A solver may benefit from producing a persuasive but incorrect answer, hiding uncertainty, or exploiting spurious shortcuts, while an auditor or verifier must decide whether costly monitoring is worthwhile. Alignment is therefore a fixed-point problem: stronger penalties may deter solver misbehavior, but they can also reduce the auditor's incentive to inspect, since auditing then mainly incurs cost on a population that appears increasingly aligned. This perspective also changes what should count as a post-training signal. Standard feedback often attaches reward to the final answer alone, but a solver-auditor pipeline exposes the full correction event: whether the solver erred, whether the auditor inspected, whether the error was caught, and whether oversight incentives remained active. We formalize this interaction in a two-agent model in which a principal chooses rewards over joint correction outcomes, inducing both solver behavior and auditor monitoring. Reward design is therefore a bilevel optimization problem: rewards are judged not by their immediate semantic meaning, but by the behavioral equilibrium they induce. We propose a bandit-based outer-loop procedure for searching over reward profiles using noisy interaction feedback. Experiments on an LLM coding pipeline show that adaptive reward profiles can maintain useful oversight pressure and improve principal-aligned outcomes relative to static hand-designed rewards, including a substantial reduction in hallucinated incorrect attempts.

Prevailing alignment methods target a fixed set of preferences and therefore risk forcing value lock-in as societal norms evolve over time. We introduce Adaptive Pluralistic Alignment (APA), a modular pipeline for updating pluralistically aligned AI systems to track evolving values and avoid value lock-in without repeating costly pretraining or large-scale data collection. APA has three stages: (1) learning compact personalized reward models via low-rank reward basis decomposition, (2) using these models as a jury that collectively selects among candidate outputs through social-choice-theoretic voting, and (3) efficiently adapting the jury over time by fitting new annotator weights over the fixed reward bases as values shift. The resulting system is efficient, explainable, steerable, and modular. We implement a proof-of-concept instantiation using the PRISM multi-user alignment dataset and simulated historical annotators, and provide preliminary analysis showing that jury composition and the choice of voting rule can substantially affect outcomes, particularly when jury preferences are heterogeneous. We provide full code and resulting preference datasets at https://anonymous.4open.science/r/apa.

Training compute is increasingly outpacing the availability of high-quality data. This shifts the central challenge from optimal compute allocation to extracting maximum value from limited data. The widely adopted Chinchilla scaling law assumes every training token is unique. This limits its ability to guide pretraining decisions in data-constrained regimes. We model the excess loss under repetition with a simple additive overfitting penalty and find that it accurately describes model behavior. Our scaling law yields qualitatively new compute-optimal allocation advice. Beyond a point, further repetition is counterproductive and compute is better spent on model capacity. We show that following our law's recommended configuration improves performance in data-constrained regimes. Finally, because our one-parameter form isolates overfitting in a single coefficient, it enables direct comparison across training configurations. As a case study, we show that strong weight decay ($λ=1.0$) reduces this coefficient by approximately 70%, providing a scaling-law explanation for recent findings that optimal weight decay in data-constrained regimes is an order of magnitude larger than standard practice.

We introduce the Banach-Butterfly Invariant (BBT), an influence-adaptive Banach geometry on the Walsh-Hadamard butterfly factorization. For a Boolean function $f:\{-1,+1\}^n\to\{-1,+1\}$ with coordinate influences $\mathrm{Inf}_\ell(f)$, BBT assigns exponent $p_\ell = 1+\mathrm{Inf}_\ell(f)$ to butterfly layer $\ell$, yielding the contraction invariant $μ(f)=\prod_\ell 2^{-\mathrm{Inf}_\ell/(1+\mathrm{Inf}_\ell)}$. We prove a Jensen lower bound $\log_2μ(f) \ge -I(f)/(1+I(f)/n)$ and that $μ$ is strictly Schur-convex in the influence vector (modulo permutation), giving scaling classes $μ\sim 2^{-n/2}$ (parity), $2^{-Θ(\sqrt{n})}$ (majority), $2^{-1/2}$ (dictators). $\log_2μ$ is rational but not polynomial in the Fourier coefficients while $μ$ is algebraic, and $μ$ separates functions with identical total influence (122 pairs at $n=3$). Using the certified $n \le 4$ ternary Walsh-threshold universe from a companion synthesis manuscript as a finite testbed, we compute exact MILP minimum-support certificates for all 65,536 Boolean functions at $n=4$ (mean 6.42, max 9, all-odd by a parity argument) and on 10,000 of the 616,126 NPN-canonical representatives we enumerate at $n=5$ (matching OEIS A000370). Conditional Spearman $ρ(μ,|\mathrm{supp}|)$ at fixed total influence is $+0.571$ in the largest stratum at $n=4$ but reverses to $-0.38$ at $n=5$ under both function-uniform and NPN-canonical sampling: $μ$ is a valid Schur-convex concentration invariant, not a universal monotone predictor of minimum support across $n$. A companion application paper validates a real-valued WHT activation-energy proxy inspired by this theory on five pretrained LLMs at W2A16, cutting wikitext-2 perplexity by 15-58% versus vanilla auto-round; the transfer from Boolean theory to the real-valued proxy is qualitative, not formal.

Physics-Informed Neural Networks (PINNs) offer a flexible paradigm for solving differential equations by embedding governing laws into the training objective. A persistent limitation is instance specificity: standard PINNs typically require retraining for each new forcing term, boundary/initial condition, or parameter setting. One-shot transfer learning (OTL) addresses this bottleneck for linear operators by freezing a pretrained latent representation and computing optimal output weights in closed form, but for nonlinear problems closed-form adaptation is generally unavailable because the loss is nonconvex in the output layer. In this paper we substantially broaden the class of nonlinearities amenable to one-shot PINN transfer by combining OTL with Chebyshev polynomial surrogates. We approximate general smooth weakly nonlinear terms by truncated Chebyshev expansions over a prescribed solution range, yielding a polynomial nonlinearity that can be handled by a perturbative decomposition into linear subproblems. A multi-head PINN learns a reusable latent space associated with the dominant linear operator; at test time, solutions to new instances are obtained via a sequence of closed-form linear solves in the output layer, without retraining the network body. We provide a unified derivation of the framework for ODEs and PDEs and demonstrate accuracy and fast online adaptation on nonlinear benchmarks, including non-polynomial and singular ODE nonlinearities as well as a reaction-diffusion PDE with saturating kinetics, demonstrating the method's utility in many-query regimes.

Models that are indistinguishable on in-distribution data can behave very differently under distribution shift. We introduce Perturb-and-Correct (P&C), a post-hoc method for constructing epistemically diverse predictors from a single pretrained network. P&C applies random hidden layer perturbations with a least-squares correction in the subsequent affine layer, producing predictors that agree on calibration data while remaining free to disagree away from it. We analyze this mechanism through the post-correction residual and its first-order sensitivity: the residual is controlled near the calibration distribution by a leverage term, while corrected sensitivity grows as inputs deviate from the calibration geometry. Empirically, P&C achieves a strong ID/OOD tradeoff across MuJoCo dynamics prediction and CIFAR-10 OOD detection, matching or outperforming standard post-hoc baselines while requiring only a single pretrained model. Our findings highlight the potential in further exploiting overparameterization as a strength of deep learning models.

Rankings produced by holistic LLM-as-a-judge scoring are sensitive to the bias of the chosen judge model. We show that switching to binary rubric scoring with multi-judge filtering removes this sensitivity: decomposing the judgement matters more than the judge model itself. To support this claim, we introduce Prosa, the first real user multi-turn Brazilian Portuguese chat benchmark: 1,000 WildChat conversations scored by three judges from three model families on 16 models. Under filtered rubric scoring the three judges agree on every one of the 16 ranks, whereas under holistic scoring they agree on only 7 of 16. Additionally, the rubric filtering pipeline increases the average score gap between neighbouring models by 47%, thereby improving Prosa's discriminative power. Evaluating a new model on Prosa costs approximately $2.1 when using Gemini 3 Flash as the judge. We release the benchmark and the filtering code to ensure that future models can be assessed under identical conditions. These artifacts also make our rubric-based scoring method reusable beyond Prosa, supporting other open-ended evaluation settings.

Self-normalized martingale inequalities lie at the heart of confidence ellipsoids for online least squares and, more broadly, many bandit and reinforcement-learning results. Yet existing vector and scalar results typically rely on bounded covariates and an explicit regularization matrix, producing bounds that are \emph{not scale-invariant}: although the self-normalized quantity is scale-invariant by definition, its standard upper bounds are not. We characterize when scale-invariant upper bounds on self-normalized martingales are possible. Without further assumptions, we prove that nontrivial scale-invariant bounds exist only in dimension $d=1$; moreover, in $d=1$ we obtain $O(\log T)$ scale-invariant self-normalized bounds without any assumptions on the covariates. In contrast, for $d>1$ we show that no nontrivial scale-invariant bound can hold in full generality. We then connect this dichotomy to \emph{doubly-uniform} regret in online linear regression (i.e., regret bounds that are simultaneously independent of the covariate scale and the comparator norm) and use it to resolve the open question of Gaillard, Gerchinovitz, Huard, and Stoltz, \emph{``Uniform regret bounds over $\mathbb{R}^d$ for the sequential linear regression problem with the square loss''} (ALT 2019): in $d=1$ we give an explicit algorithm with $O(\log T)$ doubly-uniform regret, whereas for $d>1$ sublinear doubly-uniform regret is impossible. Finally, under a natural \emph{smoothness} condition (bounded Radon--Nikodym derivatives of the conditional covariate laws with respect to a fixed base measure), we recover sublinear regret for $d>1$ without bounded covariates and derive a self-normalized concentration inequality free of the usual regularization penalties, yielding arguably a first natural scale-invariant bound for adaptive, non-i.i.d. vector martingales.

Low-rank decomposition is a compelling approach for compressing large language models, but its effectiveness hinges on selecting which singular-vector bases to retain for a target task. Existing methods such as Basel adapt singular-value coefficients on downstream data and prune bases with small re-learned magnitudes, a heuristic that can be misaligned with task performance because it ignores the local geometry of the loss landscape. We present Basis Selection with Importance (BSI), a principled low-rank compression framework that ranks and prunes bases by directly estimating the expected loss increase incurred when each basis is removed. BSI derives a derivative-based importance score from a second-order Taylor expansion of the task loss with respect to singular values, combining first-order sensitivity and second-order curvature to quantify pruning impact. To make this criterion practical for LLMs, we develop an efficient Hessian-diagonal estimator by adapting the Hutchinson randomized-probing method to loss curvature with symmetric parameter perturbations. We provide a comprehensive theoretical analysis, including loss-increase bounds under basis pruning, explicit propagation of Hessian-diagonal estimation error into these bounds, variance characterization tied to the Hessian spectrum, high-probability sample-complexity guarantees for achieving a target estimation accuracy, and guidance on perturbation intensity. Extensive experiments on mathematical reasoning benchmarks demonstrate that BSI consistently outperforms state-of-the-art low-rank decomposition baselines, with especially strong improvements under deep compression.

Proteins are inherently multiscale physical systems whose functional properties emerge from coordinated structural organization across multiple spatial resolutions, ranging from atomic interactions to global fold topology. However, existing protein representation learning methods typically operate at a single structural level or treat different sources of structural information as parallel modalities, without explicitly modeling their hierarchical relationships. We introduce PRIME (Protein Representation via Physics-Informed Multiscale Equivariant Hierarchies), a unified framework that models proteins as a nested family of five physically grounded structural graphs spanning surface, atomic, residue, secondary-structure, and protein levels. Adjacent levels are connected through deterministic, physics-informed assignment operators, enabling bidirectional information exchange via bottom-up aggregation and top-down contextual refinement. Experiments on standard protein representation learning benchmarks demonstrate strong and competitive performance across diverse tasks, with particularly notable gains on the Fold Classification benchmark, where PRIME outperforms the strongest geometric GNN baseline by margins of 13.80 and 18.30 points on the harder Superfamily and Fold splits, and achieves a state-of-the-art accuracy of 84.10% on Reaction Class prediction, surpassing all baseline methods, including ESM. Ablation studies confirm that each structural level contributes complementary and non-redundant information, and adaptive cross-attention analysis reveals that PRIME autonomously identifies the most task-relevant structural resolutions at prediction time. Our source code is publicly available at https://github.com/HySonLab/PRIME

Human behavior is difficult to observe continuously at scale, yet it leaves measurable traces in everyday device use. We test whether encrypted smartphone network traffic -- a ubiquitous, always-on, passive sensing modality -- can passively capture behavioral patterns related to sleep, stress, and loneliness. We model shared behavioral structure using a transformer backbone with per-user adapters, allowing the model to represent both typical individual behavior and deviations from it. To make these representations interpretable, we apply a sparse autoencoder to extract behavioral features corresponding to distinct patterns of activity. We relate these features to sleep disturbance, stress, and loneliness using generalized estimating equations with Mundlak decomposition, separating between-person differences from within-person changes over time. We find that the three outcomes reflect distinct temporal structures: stress is primarily associated with stable between-person differences, loneliness with within-person variation, and sleep disturbance with a combination of both. Notably, these within-person dynamics are not captured by predefined network-traffic features, demonstrating the value of learned representations for longitudinal behavioral sensing. These results establish encrypted network traffic as a viable passive sensing modality, revealing interpretable behavioral dynamics -- particularly deviations from an individual's baseline -- that are not visible in raw traffic features.

Cluster resource allocation is a multidimensional search problem that finds the best allocation of tasks to servers. Because the search space grows exponentially, modern approaches frame it as a mixed integer program (MIP) or a complex set of search heuristics. This paper proposes using a different approach: convex optimization, which has extremely fast solution methods. The research challenge is devising how to transform cluster resource allocation into a convex problem that generates good placements. We describe CvxCluster, which allocates cluster resources with a two-stage algorithm. The first stage solves a convex relaxation of the placement problem to yield a principled set of per-machine resource prices. The second stage uses these prices to drive a lightweight greedy procedure to place tasks. Experimental results with Azure traces find that CvxCluster scales to 100,480 servers under proportional workload growth and sustains arrival rates up to 500,000x the baseline trace. CvxCluster runs 100 to 2,500x faster than a state-of-the-art MIP solver while remaining within 3% of the optimal objective. CvxCluster can support complex constraints such as job anti-affinity, machine types, and GPU servers. The key insight behind CvxCluster is that reformulating placement as a continuous rather than discrete problem enables much faster methods that find solutions just as good or better than prior heuristics.

Due to ambiguity in the wording of the EU AI Act, we examine the question of to what extent frontier biological foundation models such as ESM3 are subject to obligations for general-purpose AI models with systemic risk under the EU AI Act. In this paper, we map ESM3 to the biorisk chain, and conclude that it would be desirable if the providers of ESM3 and similar biological models were subject to these obligations, which would require them to assess and mitigate dual-use risks from their models. We then perform an analysis, comparing the attributes of ESM3 to the classification criteria in the AI Act and the supporting material. We conclude that at this time, ESM3 does not appear to be meaningfully regulated by the Act. We then propose remedies to correct the situation.

AI systems have long been expected to interact with users, answering questions, generating content, and continuing (social) conversations. Agentic AI, however, breaks from this expectation, as its primary objective is workflow execution on behalf of the users. If a system becomes more agentic, do users need less interaction with the system? Our answer is: less routine back-and-forth, but more communication for oversight and explanation, as agentic AI proactively acts, not just responds. Grounded in a communication perspective, we discuss how users perceive the communicative roles of AI systems (whether as the source of actions or merely a channel), and how this can shape trust. Because agentic AI can play multiple communicative roles, it can complicate this source perception and introduce potential risks. To address this, we propose three types of explanations that agentic AI needs to incorporate (action-process, uncertainty, and coordination), and suggest that customization affordances that allow users to decide when and which explanations they see may be key to preserving human agency as AI autonomy increases.

We test whether the causal inner product of \citet{park2024linear} -- defined by the unembedding covariance $Σ$ -- enables cross-lingual concept transport. Across 17 models and 4 language pairs, a matched-spectrum randomization test finds that Whitened Causal Alignment is indistinguishable from spectral regularization alone ($p = 0.95$). However, this failure reveals a broader phenomenon: anti-concentration is observed in residual-stream difference-of-means vectors across five architecture families ($p < 10^{-33}$) and supported by SAE features (e.g., $p = 4.5 \times 10^{-19}$) and linear probes on Gemma and Llama. We discover a \emph{dual geometry}: activation-space concept directions anti-concentrate in the spectral tail, while static unembedding-row contrasts \emph{concentrate} in high-variance directions ($p < 10^{-4}$). Split-injection causal interventions support the functional basis on Gemma and Llama (Cohen's $d$ up to $1.80$), and POS-tag probing across 8 models shows syntax preferentially encodes in the high-variance subspace in 6 of 8 architectures ($p < 0.013$), with the Qwen~2.5 family showing a significant reversal consistent with architecture-specific spectral structure. These results suggest transformers may rotate semantic content into spectrally quiet regions during contextualized processing, encoding concepts where they can be manipulated with reduced grammatical disruption.

Large language models are sensitive to minor prompt perturbations, yet existing robustness methods usually enforce consistency at the whole-sequence level. This holistic view can hide an important failure mode: a perturbed response may remain globally similar to the clean one while drifting on a critical entity, relation, or conclusion. We introduce S$^2$R$^2$, a segment-level framework for robust LoRA fine-tuning. S$^2$R$^2$ decomposes clean and perturbed generations into semantic segments, aligns them with an optimal-transport objective, and penalises the segments with the largest meaning drift. To connect this output-side objective with model adaptation, we add an adapter-stability regulariser motivated by segment-level attention reallocation, using LoRA norm control as a tractable proxy for limiting perturbation-amplified evidence shifts. A PAC-Bayesian complexity view further explains why controlling adapter growth may support transfer beyond observed perturbations. Experiments on summarisation benchmarks show that S$^2$R$^2$ improves robustness under typographical noise, deletion, synonym replacement, and paraphrasing, while maintaining competitive clean performance and stronger cross-dataset transfer than consistency-based baselines.

Existing evaluation frameworks for large language models -- including HELM, MT-Bench, AgentBench, and BIG-bench -- are designed for controlled, single-session, lab-scale settings. They do not address the evaluation challenges that emerge when agentic AI systems operate continuously in production: compounding decision errors, tool failure cascades, non-deterministic output drift, and the absence of ground truth for long-horizon tasks. This paper makes three contributions. First, we present a taxonomy of seven failure modes unique to production agentic systems, each grounded in observations from systems operating at billion-event scale. Second, we demonstrate empirically where standard metrics -- ROUGE, BERTScore, accuracy/AUC, and the agentic benchmarks above -- fail to detect each failure mode. Third, we propose PAEF (Production Agentic Evaluation Framework), a five-dimension evaluation framework with an open-source reference implementation, designed for continuous evaluation on production traffic rather than episodic benchmark runs. Our analysis shows that standard metrics fail to detect four of the seven failure modes entirely and detect three others only after a lag of multiple evaluation cycles.

Empirical process control, a way of managing work based on the observation of the successes or misfortunes of earlier activities, is a key process in Scrum and other agile development frameworks. In this experience report, we present a lightweight, scalable, free and customizable sprint simulation activity designed to teach students how to empirically control a Scrum project by engaging in the presentation and interpretation of work status information, task selection and resource allocations in a single teaching session. We reflect on our experience using the simulation as an active learning complement to direct instruction in two master level courses at two different universities and in the training of teaching assistants at a third institution, and abductively establish its effectiveness by mapping student comments to the teaching practices in the threshold concepts framework.

This paper describes the system submitted by team \textbf{Archaeology} to SemEval-2026 Task~13 on AI-generated code detection. The shared task consists of three subtasks; we participate in Subtask-A (binary classification: human-written vs.\ AI-generated code) and Subtask-B (11-class attribution of the generating model). Starting from a TF-IDF and Logistic Regression baseline, we fine-tune four pre-trained code models (CodeBERT, GraphCodeBERT, UniXcoder, and CodeT5+) with separate strategies for each subtask. For Subtask-A, we use leave-one-language-out cross-validation, code augmentation, chunked inference with trimmed-mean aggregation, and threshold calibration on a difficult dataset. For Subtask-B, we use sandwich token packing, class-balanced loss, and multi-seed ensembling with test-time augmentation. Our best submissions obtain macro-F1 scores of 0.737 on Subtask-A (6th/81 teams) and 0.422 on Subtask-B (7th/34 teams).

Neural Ranking Models (NRMs) are central to modern information retrieval but remain highly vulnerable to adversarial manipulation. Existing attacks often rely on heuristics or surrogate models, limiting effectiveness and transferability. We propose CRAFT, a supervised framework for black-box adversarial rank attacks powered by large language models (LLMs). CRAFT operates in three stages: adversarial dataset generation via retrieval-augmented generation and self-refinement, supervised fine-tuning on curated adversarial examples, and preference-guided optimization to align generations with rank-promotion objectives. Extensive experiments on the MS MARCO passage dataset, TREC Deep Learning 2019, and TREC Deep Learning 2020 benchmarks show that CRAFT significantly outperforms state-of-the-art baselines, achieving higher promotion rates and rank boosts while preserving fluency and semantic fidelity. Moreover, CRAFT transfers effectively across diverse ranking architectures, including cross-encoder, embedding-based, and LLM-based rankers, underscoring vulnerabilities in real-world retrieval systems. This work provides a principled framework for studying adversarial threats in NRMs, underscores the risks of generative AI in rank manipulation, and provides a foundation for developing more robust retrieval systems. To support reproducibility, we publicly release our source code, trained models, and prompt templates.

Authoritative competency frameworks such as ESCO, ROME, and O*NET are essential for aligning education with labor market needs, yet their technical complexity and structural heterogeneity hinder practical adoption by educators. This paper introduces SkillGraph-Service, an interoperable microservice designed to bridge this gap by unifying these resources into a provenance-preserving Knowledge Graph (KG). Adopting a KG-first, LLM-fallback architecture, the system combines symbolic rigor with sub-symbolic flexibility. It implements a lightweight hybrid retrieval engine (fusing SQLite FTS5 and HNSW vector search) to handle the vocabulary mismatch in educator queries, and utilizes Large Language Models (LLMs) strictly for constrained ranking and audience-aware explanation. Empirical evaluation on a multilingual dataset reveals that the proposed hybrid strategy achieves superior retrieval effectiveness (nDCG@5>0.94) with sub-200 ms latency, rendering computationally expensive cross-encoder re-ranking may be unnecessary for this domain. Furthermore, an analysis of generated explanations highlights a trade-off between fluency and faithfulness: while JSON-constrained LLMs ensure high citation precision, deterministic templates remain the most reliable method for maximizing evidence coverage. The resulting architecture offers a practical, scalable, and auditable solution for integrating complex skill data into digital learning ecosystems.

Modern deep learning usually treats models as separate artifacts: trained independently, specialized for particular purposes, and replaced when improved versions appear. This thesis studies model merging as an alternative paradigm: combining independently trained neural networks directly in weight space, with little or no optimization and without requiring access to the original training data. The thesis considers two main regimes. In the single-task setting, where models share an objective but differ in initialization, we introduce C$^2$M$^3$, a cycle-consistent merging algorithm based on Frank-Wolfe optimization. C$^2$M$^3$ aligns multiple networks into a shared, reference-free parameter space, making weight averaging meaningful without privileging any individual model. In the multi-task setting, where models are fine-tuned for different downstream tasks from a common pretrained initialization, we first develop a theoretical account of task vectors as approximate gradients. This explains both the effectiveness and the limitations of task arithmetic. Building on this view, we show that task vectors inherit the low-rank structure of gradients and introduce Task Singular Vectors (TSV), a decomposition that enables compression and interference reduction through TSV-Merge. We then present MASS, an input-adaptive routing method that uses TSV geometry to select task-relevant subspaces at inference time. Finally, we introduce MERGE$^3$, an evolutionary merging framework that uses Item Response Theory to reduce evaluation costs by up to 50$\times$ while preserving solution quality. Together, these contributions provide theoretical and algorithmic foundations for model merging, supporting a paradigm in which learned capabilities can be composed, reused, and extended across models.

Empirical causal claims depend on many analyst decisions, from selecting covariates to choosing estimators. Existing robustness tools summarize how results vary across these choices, but, to the best of our knowledge, do not answer: \textbf{How many analyst decisions must change to reach a specification, which is a set of choices, whose confidence interval (CI) contains zero?} We introduce \emph{Minimum Specification Perturbation (MSP)}, the smallest number of changes. MSP is small under the null, grows with effect strength and captures distance-to-falsification information that dispersion-based summaries cannot report; when making decisions under weak effects, an MSP-based rule yields lower false-positive rates than dispersion-based rules. We show that Fragility Index and MSP measure orthogonal vulnerabilities: fragility to influential observations need not imply fragility to specification choices. On the LaLonde benchmark, MSP = 1 implies that one decision change makes the CI contain zero. We further provide exact permutation calibration under randomization and characterize computation, showing tractable cases under additive structure and NP-hardness in general.

The march toward developing relevant and robust CPU benchmarks continues with the introduction of SPEC CPU 2026, the next generation suite for measuring processor performance. This paper details the methodology behind its creation, showcasing a process centered on community collaboration and principled development. The suite is built upon a foundation of modern, open-source applications, selected and hardened through a process that emphasizes workload diversity, portability, and software longevity. A key contribution is Rolling-Round-Robin Rate, a novel and standardized approach to running heterogeneous, multiprogrammed workloads that addresses a long-standing gap in benchmarking practice. Additionally, the suite features an expanded set of multithreaded benchmarks and introduces workloads with distinct microarchitectural profiles, reflecting the demands of contemporary software. By detailing our principled approach to benchmark selection, adaptation, and validation, we demonstrate how the SPEC CPU 2026 suite sets the standard for performance evaluation in the next era of computer architecture research and development.

Vehicle Routing Problem (VRP) is one of the most complex NP-hard combinatorial optimization problem in transportation and logistics that requires a dynamic solution approach. In this paper we present a new hybrid approach that combines the Quantum Approximate Optimization Algorithm (QAOA) into the QRL policy network, instead of the usual variational layers, QAOA mixing and cost Hamiltonian layers. This enhancement enables the agent to exploit problem specific particular quantum correlations when learning policies, and so richer exploration of the routing solution space. The QAOA-augmented QRL framework shows quicker convergence in training and can tackle larger VRP instances that are beyond the reach of Grover's Adaptive Search (GAS) and Quantum Reinforcement Learning (QRL) approaches. Experiments on standard VRP instances demonstrate better solutions, fewer episodes to converge and good memory usage on near term quantum hardware simulators. These findings demonstrate QAOA- integrated QRL as a viable approach to scalable, high quality quantum-assisted combinatorial optimization.

Large language model (LLM) coding agents increasingly operate over repositories, terminals, tests, and execution traces across long software-engineering episodes. Persistent memory is useful, but static vector stores or generic retrieval-augmented generation (RAG) are insufficient for reinforcement-learning (RL) code development, where small details can alter Bellman targets, terminal masks, gradient flow, or validation claims. This paper presents RL Developer Memory, a local-first, Model Context Protocol (MCP)-native developer-memory architecture for RL coding agents. It treats memory selection as a logged contextual decision process: issue_match ranks candidates and records telemetry, issue_feedback maps raw labels to bounded rewards, and issue_record_resolution links verified resolutions to earlier retrieval events. A deterministic ranker remains deployed, while a contextual-bandit residual policy runs in shadow mode and can affect canary behavior only through conservative off-policy-evaluation (OPE) gates. RL/control memories require theory-to-code metadata and review-gated governance. The system is evaluated on a deterministic 200-case benchmark with RL algorithm bugs, hard negatives, review-gated RL/control cases, and low-risk failures. In the same-commit comparison, deterministic control and full shadow/OPE both achieve 80.0% expected-decision accuracy and 100.0% hard-negative suppression; the full configuration adds learning telemetry rather than accuracy gain. Static validation passed 11/11 checks; dynamic integration passed 10/10 cases. The evidence reports limits: active learned-policy deployment and official-client MCP interoperability are unsupported, live full-configuration latency regresses, and 40 residual non-RL failures remain. The contribution is an auditable memory-control architecture with explicit claim boundaries, not a universal coding-agent improvement claim.

Advances in inference methods have enabled language models to improve their predictions without additional training. These methods often prioritize raw performance over cost-effective compute usage. However, computational efficiency is key for real-world applications with resource constraints. We provide a systematic analysis of the inference scaling strategies self-consistency, self-refinement, multi-agent debate, and mixture-of-agents, to study their computational performance tradeoffs. We evaluate methods on two reasoning benchmarks (MMLU-Pro, BBH) and include extensive parameter configurations (e.g., scaling the number of parallel predictions, agents, and debate rounds) across different model sizes. Across 34 configurations and over 100 evaluations, we compute the Pareto-optimal front to select methods that achieve the best accuracy with the lowest computational budget. Notably, inference scaling improves accuracy by up to +7.1% points over chain-of-thought at the highest evaluated budgets (20x the CoT compute budget) on MMLU-Pro. With an equal computing budget, debate and mixture-of-agents outperform self-consistency by 1.3% and 2.7% points, respectively. While self-consistency saturates earlier, multi-agent gains persist, particularly on more complicated tasks. We identify a simple multi-agent design guideline: mixture-of-agents is most efficient when the number of parallel generations exceeds the number of sequential aggregations.

The Object-Oriented Method for Requirements Authoring and Management (OOMRAM) is a requirements reuse framework that relies on exact identifier matching and rigid templates, limiting its ability to adapt specifications across diverse contexts. While Large Language Models (LLMs) offer the flexibility to overcome this bottleneck, they introduce the risk of generating structurally invalid or inconsistent requirement combinations. To address this tension, we present a neuro-symbolic multi-agent system that re-conceptualizes requirements reuse as a \textbf{Model-Driven Elicitation process}. In this paradigm, an LLM serves as a \textbf{non-deterministic heuristic} for traversing a \textbf{deterministic domain model} represented by a formal OOMRAM requirement lattice. A deterministic, symbolic validator enforces all structural constraints within the agent loop, effectively eliminating hallucinated requirement combinations by construction. Evaluated on an autonomous benchmark across two application families, our system achieves 100\% requirement coverage and a constraint-violation rate of only 0.2\%. Although the F1-score against a single gold standard is moderate (0.47--0.51), every generated specification is structurally valid and satisfies all mandatory domain requirements. The model-agnostic implementation scales to larger lattices via subgraph navigation and provides transparent audit trails for regulatory compliance.

This paper develops a Hall-Sandpile model of economic instability that combines a Hall-like transversal stress mechanism with sandpile threshold dynamics on a real production-network substrate. In analogy with the physical Hall effect, where exposed flows under an external field generate stress in a transversal direction, we model economic shocks as fields that act on flow-intensive, low-redundancy, low-capacity nodes and produce systemic stress through a multiplicative conversion function. The accumulated stress drives a discrete toppling rule and an avalanche dynamics whose effective activation threshold declines with transversal exposure. The model is calibrated on annual World Input--Output Database (WIOD) production networks for 2000--2014 and simulated on the 2014 substrate (2{,}283 country--sector nodes) under three alternative propagation normalisations to avoid mechanical near-criticality from row-stochastic operators. Controlled Monte Carlo experiments over external field intensity and redundancy stress generate four ordered regimes: stable absorption, latent fragility, critical transition, and avalanche regime. Mean avalanche size and the probabilities of finite-size systemic events $\Pr(S\!\geq\!5)$, $\Pr(S\!\geq\!10)$ and $\Pr(S\!\geq\!20)$ rise jointly with field intensity and redundancy stress. Tail diagnostics show regime-dependent thickening of the avalanche distribution, but the estimated tail indices remain too high to interpret as evidence of universal power-law criticality. The contribution is therefore a finite-size, real-network description of how transversal stress activates structural fragility, not a claim of self-organised criticality in the global economy.

We introduce an autonomous multiagent framework for mechanistic interpretability that automates both explaining and finding internal features in large language models. The system runs two coupled loops: (1) explanation refinement, where an agent proposes competing hypotheses and iteratively tests them with targeted prompt controls and a multi-metric evaluation; and (2) feature discovery, where an agent generates prompt sets, constructs a k-nearest-neighbor graph in activation space, and retrieves candidate features using statistical separability and semantic coherence criteria. On Gemma-2 family models and MLP neurons in weight-sparse transformers, our agent improves over one-shot auto-interpretations, discovers language-specific and safety-relevant features, and produces auditable explanation traces, showing that agent-driven empirical loops yield sharper and more falsifiable explanations than one-shot labels.

Electrocardiogram (ECG) biometrics have emerged as a promising modality for continuous, liveness-aware authentication in wearable systems. However, many prior studies report overly optimistic results due to data leakage (e.g., random splits within the same session). To address this issue, we introduce ECG-biometrics-bench, a modular, reproducible benchmarking framework that standardizes preprocessing, segmentation, and evaluation across seven widely used public ECG datasets spanning clinical, ambulatory, and large-scale cohort settings. The framework supports both closed-set and open-set (i.e., subject-disjoint generalization in this work) evaluation, as well as progressively realistic protocols including cross-session and long-term temporal separation. To facilitate reproducible research in the community, the ECG-biometrics-bench repository will be made publicly accessible on GitHub upon the acceptance of this manuscript. Through a comprehensive multi-dataset analysis, we expose the Random Split Fallacy, demonstrating that intra-session evaluation protocols artificially inflate performance while masking severe degradation caused by temporal drift and unseen identities. Furthermore, by evaluating multiple architectures, including DeepECG, ResNet1D, and CNN-LSTM, we show that these failures are not model-specific but are likely inherent to current supervised feature-learning paradigms. Finally, we demonstrate that performance degradation due to temporal aging can be partially mitigated through a heavy enrollment, lightweight authentication strategy based on dynamic multi-session template fusion. These findings establish a more realistic baseline for ECG biometrics and highlight critical challenges that must be addressed for reliable real-world deployment.

Sixth-generation (6G) networks are increasingly envisioned as AI-native infrastructures integrating communication, sensing, and computing into a unified fabric. However, existing approaches remain largely optimization-centric, relying on closed-loop control with limited reasoning capability. In this paper, we argue for a paradigm shift toward Agentic AI-Native 6G, in which Large Language Model (LLM)-based agents operate as bounded, policy-governed reasoning entities within a semantic control plane layered above deterministic 3GPP infrastructure. We propose a four-layer architecture that integrates deterministic network infrastructure, semantic abstraction of intent and context, hierarchical reasoning, and a distributed multi-agent fabric spanning device, edge, and core domains. To assess feasibility, we develop a proof-of-concept agentic reasoning and orchestration framework and conduct an extensive empirical study using a domain-specific 6G benchmark under realistic deployment constraints. Our results reveal a fundamental tradeoff between reasoning capability and system efficiency, showing that no single model simultaneously satisfies latency, throughput, and accuracy requirements. Instead, heterogeneous deployment of LLM agents across the device--edge--core continuum is necessary to balance these constraints. We further demonstrate that quantization introduces non-uniform effects across models, reinforcing the need for system-level optimization rather than model-level compression alone. These findings establish agentic intelligence as a viable architectural direction for 6G and highlight key challenges in achieving scalable, trustworthy, and self-reasoning networks. All experimental results and evaluation scripts are publicly available to support reproducibility.

Machine Learning surrogates for Computational Fluid Dynamics (CFD), particularly Graph Neural Networks (GNNs) and Transformers, have become a new important approach for accelerating physics simulations. However, we identify a critical bottleneck in the field: while architectures have advanced significantly, the common underlying training paradigms remain bound to naive assumptions, such as node-wise supervision and explicit Euler time-stepping. These legacy choices ignore the stiff dynamics and local flux continuity inherent to numerous partial differential equations resolution methods, such as Finite Element, Difference, or Volume (FEM). In this work, we propose a unified framework to bridge the gap between geometric deep learning and rigorous numerical analysis. We introduce three key innovations: (1) Multi Node Prediction, a stencil-level objective that predicts field values for a node's full local topology, enforcing spatial derivative consistency; (2) Temporal Correction, replacing unstable explicit schemes with a predictor-corrector via temporal Cross-Attention; and (3) Geometric Inductive Biases, leveraging 3D Rotary Positional Embeddings (RoPE) to robustly capture rotational symmetries in unstructured meshes. We evaluate this framework across three architectures (MeshGraphNet, Transolver, and a Transformer) on diverse physics datasets. Our approach yields consistent improvements in accuracy and stability, particularly in long-horizon rollouts, while producing latent representations that generalize to unseen subtasks such as Wall Shear Stress or Pressure prediction. Code is available at https://github.com/DonsetPG/graph-physics.

Isolated word meanings are inherently uncertain. This uncertainty reduces when they are combined and anchored in context. We propose that grammar compresses meaning uncertainty cross-linguistically, which is reflected in brain and selectively disrupted in disorders. Compression was operationalized as the relative difference between non-contextual surprisal estimated from lexical frequency, and contextual surprisal from grammar-sensitive models. In narratives from 20 languages, contextual surprisal reduced frequency-based surprisal. This reduction closely tracked the surprisal cost of reversing word order, and scaled with richer, non-redundant lexis as organized by more complex but optimal dependency structure. During fMRI, surprisal and its reduction explained BOLD activity for comprehension and production in overlapping but distinct regions. Uncertainty reduction was significantly attenuated in aphasia, dementia, and schizophrenia, but remained intact where primary deficit is not language. These findings position uncertainty reduction via grammar as a foundational concept that illuminates principles, brain basis, and disruptions of language.

Microservice architecture is widely adopted in modern systems, where auto-scaling is critical for satisfying service-level objectives (SLOs). However, determining optimal scaling for microservices is difficult, and reactive resource allocation often leads to costly over- or under-provisioning. We propose AutoSLO, a learning-based, self-adaptive scaling framework that dynamically adjusts microservice replicas to meet SLOs while minimizing resource usage. AutoSLO uses a continuous monitoring-adaptation feedback loop and leverages genetic programming to learn and evolve scaling logic, enabling the deployed microservice system to proactively prevent SLO violations rather than repeatedly searching for one-off scaling actions. We evaluate AutoSLO on two case-study systems -- an online shopping platform and a chatbot based on large language models -- and show that this framework substantially reduces resource usage while maintaining a low frequency of SLO violations, all of which are resolved within a short time window.

Vision-Language Models (VLMs) frequently suffer from visual perception errors and hallucinations that compromise answer accuracy in complex reasoning tasks. Reinforcement Learning with Verifiable Rewards (RLVR) offers a promising solution by optimizing policies using answer correctness signals. Despite their effectiveness, prevailing RLVR methods face two critical limitations. First, much of the sampling budget is wasted on trajectories doomed to fail due to early visual description errors. Second, sparse rewards cannot distinguish whether failures stem from visual perception or reasoning stages. We introduce MIRL, a decoupled framework that addresses both limitations by leveraging mutual information (MI) between generated descriptions and visual inputs as a cheap pre-screening signal. This enables intelligent budget allocation toward high-potential trajectories via forking, while decoupled training provides independent MI-based rewards for visual perception optimization, resolving reward blindness. Experiments on six vision-language reasoning benchmarks demonstrate that MIRL achieves 70.22% average accuracy and successfully surpasses the performance of sampling 16 complete trajectories using only 10 pre-samples with top-6 selection (25% fewer complete trajectories). Our code is available at: https://anonymous.4open.science/r/mirl-main/.

Principal Component Analysis (PCA) is widely used for dimensionality reduction in hyperspectral imaging, genomics, and neurosciences. However, it suffers from computational bottlenecks in matrix multiplication and singular value decomposition (SVD). Prior PCA hardware accelerators either target only one of these stages, rely on High Level Synthesis (HLS) that limits microarchitectural optimizations or use fixed point datapaths with limited dataset scalability. There is a need for a unified PCA accelerator that is suitable for datasets of any input dimension. Hence, the proposed work presents MANOJAVAM, a scalable PCA accelerator fabric, unifying matrix multiplication and SVD in a single architecture. MANOJAVAM(T,S) comprises an S number of TxT TPU-style systolic arrays employing block streaming for high-throughput matrix multiplication. It further integrates a highly parallel Jacobian unit implementing the Jacobi method for SVD with pipelined CORDIC based rotations. A two tier cache hierarchy and mode-aware memory policies adapts to the distinct memory access patterns of covariance matrix and rotation computation. For demonstration, MANOJAVAM(4,8) is realized on a Xilinx Artix-7 FPGA, achieving a frequency of 200 MHz at 1.271W. MANOJAVAM(16,32) is realized on Xilinx Virtex-Ultrascale+ FPGA, achieving a frequency of 434 MHz at 16.957W. Benchmarking on real-world datasets reveals that MANOJAVAM(16,32) achieves up to a 22.75x speedup in SVD latency and a 42.14x reduction in total energy consumption compared to a high-performance NVIDIA A6000 GPU. The architecture offers a unified, scalable, and energy-efficient platform for large-scale data analytics in both high-performance and edge-computing environments.

Designing therapeutic messenger RNA (mRNA) requires creating full-length transcripts that carefully balance stability, translation efficiency, and immune safety. To address this challenge, we propose ProMORNA, a multi-objective generation framework that produces complete mRNA transcripts \textit{de novo} directly from a target protein sequence. Our approach begins by training a BART-style encoder-decoder model on over 6 million natural protein-mRNA pairs. We then introduce Multi-Objective Group Relative Policy Optimization (MO-GRPO) to simultaneously optimize for various biological objectives in a unified way. As a case study, we evaluated ProMORNA on the widely used firefly luciferase target, excluding it from both our supervised training data and the prompt pool. The results indicate that ProMORNA improves the \textit{in silico} Pareto frontier for predicted half-life and translation efficiency relative to standard supervised baselines. Additionally, it achieves higher predicted functional scores than a state-of-the-art baseline under the same evaluation pipeline. These computational findings demonstrate the feasibility of using multi-objective reinforcement learning for full-length mRNA design on unseen targets.

Prior studies report that partial driving automation can increase the cognitive demands on human drivers. This effect largely arises from human drivers' lack of transparent insight into the vehicle's intentions and decision logic, as well as from automated systems' limited awareness of the driver's dynamic state and preferences. This bidirectional misalignment undermines shared situational awareness and exacerbates coordination failures in human-vehicle interaction. To address these limitations, we argue for a paradigm shift that elevates the human role from passive supervisor to active manager. We introduce the Mediator-in-the-Loop-Driving (MILD) system, based on an agentic system architecture to facilitate synergistic human-vehicle collaboration. MILD integrates a perception agent for joint in-cabin and out-of-cabin understanding with a lightweight strategy agent that generates compliant and explainable action suggestions. To ensure these strategies are strictly aligned with safety regulations and human values, we develop Evidence- and Constraint-weighted Policy Optimization (ECPO). ECPO leverages automatic validators to steer the agent toward behaviors that are not only accurate but also structurally complete, substantiated by evidence, and free from constraint violations. Furthermore, a retrieval-augmented generation module dynamically incorporates constraints from traffic regulations, speed recommendations, and driver preferences into the decision loop. Field experiments across three open datasets demonstrate that MILD consistently outperforms baselines in both perception accuracy and strategy quality under auditable offline metrics, and yields higher human-rated policy adequacy, comfort, and explanation than baselines. This work offers a practical pathway for building auditable and aligned agents for human-vehicle collaborative driving.

Patrolling with multiple robots offers efficient surveillance to detect and manage undesired situations. This necessitates improved patrol efficiency and operator situation awareness at base stations. Enhanced situation awareness enables operators to predict robots' behaviors, support recognition and decision-making, and execute emergency interventions. This study presents the Local Reactive and Partition (LR-PT) algorithm, a novel multi-robot patrolling approach. In simulations, LR-PT outperformed existing methods by ensuring frequent patrols of all locations of interest and enhancing the situation awareness of the base station. Robots independently select patrol targets based on locally available information, integrating patrol needs and the urgency of reporting mission progress to the base station into a unified utility function. This locality also contributes to robustness against communication constraints and robot failures, as demonstrated in this research. The algorithm further autonomously emerged the area partition, which can avoid falling into local optima and realize the comprehensive patrol over the whole mission area. The simulation results demonstrated the superior performance of LR-PT for multi-robot patrolling, utilizing the advantages of swarm robotics and addressing real-world operational challenges.

Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by grounding responses in external knowledge during inference. However, conventiona RAG systems under-perform on structured tabular data, largely due to coarse retrieval granularity and insufficient table semantic comprehension. To address these limitations, we introduce FT-RAG, a fine-grained framework that employs knowledge association by decomposing tables into entry-level semantic units to construct a structured graph. FT-RAG employs a structural neighbor expansion mechanism to find semantically connected entities during graph retrieval, followed by multi-modal fusion to consolidate the context of table retrieval results. Further, to address the scarcity of specialized datasets in this domain, we introduce Multi-Table-RAG-Lib, a benchmark comprising 9870 QA pairs with high complexity and difficulty, curated to demand multi-table integration and text-table information fusion for reasoning. FT-RAG surpasses top-performing baselines across all metrics, achieving a 23.5\% and 59.2\% improvement in table-level and cell-level Hit Rates, respectively. Generation performance also sees a remarkable 62.2\% increase in exact value accuracy recall. These metrics verify the framework's effectiveness in factual grounding across both pure tabular and heterogeneous table-text contexts. Therefore, our method establishes a new state-of-the-art performance for complex reasoning over mixed-modality documents.

We study high-dimensional LASSO under differential privacy via objective perturbation with heterogeneous covariate scales. In practical scenarios, covariates often exhibit diverse scales; however, standard preprocessing is problematic under privacy constraints, as it consumes additional privacy budget. This heterogeneity induces effective anisotropy in the objective perturbation via the inverse Gram matrix of covariates, which can degrade the stability and accuracy of algorithms. To address this, we propose a Gram-based anisotropic objective perturbation, a ``pre-distortion" strategy that counteracts the distortion from the covariate structure to restore isotropy in the estimation process. Using an Approximate Message Passing (AMP) framework and state evolution analysis, we demonstrate that our proposed perturbation significantly stabilizes convergence and improves both statistical efficiency and privacy performance compared to standard uniform noise injection. Our results provide theoretical insights into designing stable and efficient private estimators without relying on data-dependent preprocessing.

Pansharpening aims to generate high-resolution multispectral (HRMS) images by fusing low-resolution multispectral (LRMS) images with high-resolution panchromatic (PAN) images. However, the current mainstream frequency-based pansharpening methods employ fixed frequency filters, which cannot precisely adapt to complex and spatially diversified frequency distributions in PAN and MS images. Furthermore, existing denoising strategies insufficiently exploit frequency components for denoising and struggle to suppress various noise types accurately. To address these challenges, we propose CGFformer, a cluster-guidance frequency Transformer that focuses on varying frequency distribution and interactions between frequency and spatial components. Specifically, we design an adaptive separation module that integrates local features and non-local information through K-means clustering, enabling more precise separation of high- and low-frequency components. Subsequently, we introduce a dual-stream refinement module combined with Transformer-based cross-attention to remove various noise, allowing the network to jointly suppress frequency-relevant and irrelevant disturbances. In addition, we develop a frequency-spatial fusion module designed to enhance detail and facilitate spatial-frequency interaction, ensuring more effective reconstruction of spatial structures in the fused results. Extensive experiments on multiple benchmark datasets demonstrate that the proposed CGFformer achieves notable improvements over existing pansharpening approaches.

Frontier scientific reasoning is rapidly emerging as a key foundation for advancing AI agents in automated scientific discovery. Deep research agents offer a promising approach to this challenge. These models develop robust problem-solving capabilities through post-training on information-seeking tasks, which are typically curated via knowledge graph construction or iterative web browsing. However, these strategies face inherent limitations in frontier science, where domain-specific knowledge is scattered across sparse and heterogeneous academic sources, and problem solving requires sophisticated computation and reasoning far beyond factual recall. To bridge this gap, we introduce SciResearcher, a fully automated agentic framework for frontier-science data construction. SciResearcher synthesizes diverse conceptual and computational tasks grounded in academic evidence, while eliciting information acquisition, tool-integrated reasoning, and long-horizon capabilities. Leveraging the curated data for supervised fine-tuning and agentic reinforcement learning, we develop SciResearcher-8B, an agent foundation model that achieves 19.46% on the HLE-Bio/Chem-Gold benchmark, establishing a new state of the art at its parameter scale and surpassing several larger proprietary agents. It further achieves 13-15% absolute gains on SuperGPQA-Hard-Biology and TRQA-Literature benchmarks. Overall, SciResearcher introduces a new paradigm for automated data construction for frontier scientific reasoning and offers a scalable path toward future scientific agents.

Legal consultation is a high-stakes, knowledge-intensive task that requires agents to identify relevant legal issues, retrieve authoritative support, and determine when evidence is sufficient for a recommendation. Although retrieval-augmented generation has improved grounding in legal question answering, many multi-turn legal agents still rely on fixed retrieval depth or coarse heuristic control. This often leads to either insufficient support for key legal elements or excessive retrieval that increases context burden and weakens answer focus. We propose MAP-Law, a coverage-driven framework for retrieval control in multi-turn legal consultation. MAP-Law models consultation as a controlled retrieval process over a joint structured state consisting of issue nodes, legal element nodes, and evidence nodes. After each retrieval round, the agent computes Element Coverage, Evidence Coverage, and Marginal Gain, and uses these signals to decide whether to continue retrieval, redirect the search, or generate the final response. In this way, MAP-Law turns stopping from a fixed hyperparameter into an interpretable and auditable decision aligned with legal argumentative structure. Experiments on a self-constructed dataset of 50 cases across eight labor-law scenarios show that MAP-Law with DeepSeek as the action selector achieves an Element Coverage of 0.860 using only 2.9 retrieval rounds and 5.8 evidence pieces on average. Compared with a fixed seven-round baseline, it reduces evidence volume by over 80% and retrieval rounds by 58%. Ablation results further confirm the independent contributions of coverage-driven stopping, joint graph representation, and LLM-based action selection.

With the rapidly improving reasoning abilities of Large Language Models (LLMs), there is also a rising demand to use them in a wide variety of domains. This brings about the need to carefully evaluate the limits of the capabilities of these models with various tests and benchmarks. Graph structures are ubiquitous in real-world data, and are often used to represent and analyze relationship patterns within data. Many benchmarks have already been proposed in the graph literature to test the reasoning ability of LLMs to follow and execute graph algorithms. However, due to the limited context length of LLMs, these benchmarks consist of very small graphs. In real-world data, the size of graphs can be significantly larger, and in many cases, not fully accessible. In this paper, we examine a class of problems that arises with very large graphs having limited accessibility. We propose a large graph benchmark dataset, EstGraph, and introduce four distinct tasks designed to estimate large graph properties. We evaluate the reasoning abilities of LLMs on these tasks using a wide variety of graph datasets. In addition, we provide task-specific prompt constructions based on random walk sampling of large graphs (up to millions of nodes) that effectively convey sufficient information to LLMs within the limits of context length.

A hierarchical cross-modal fusion model is proposed for vision-language question answering (VLQA) in industrial robotics, targeting the challenges of semantic ambiguity, complex environmental layouts, and domain-specific terminology common in modern manufacturing. The framework integrates advanced object detection, multi-scale visual encoding, syntactic parsing, and task-aware semantic attention to unite vision and language signals into a joint reasoning space. Region-based deep networks extract visual features, weighted embeddings aggregate, and recurrent neural parsing encodes sentence structures. Through fine-grained semantic alignment driven by adaptive fusion and cross-attention mechanisms, the system can handle operational queries, instruction steps, and anomaly detection with higher reliability. Compared to the existing VLQA benchmarks, validation experiments conducted on the IVQA and RIF benchmarks indicate improvements in semantic alignment, Top-1 accuracy, and robustness to ambiguous or procedural task queries. Ablation studies further quantify the impact of each architectural module, confirming the necessity of multi-level feature integration and context-driven gating for dependable industrial deployment. The technical advancements reported here provide core methodologies to improve the interpretability and operational effectiveness of industrial robots faced with diverse human-robot interaction tasks.

Multi-Hop Fact Verification (MHFV) necessitates complex reasoning across disparate evidence, posing significant challenges for Large Language Models (LLMs) which often suffer from hallucinations and fractured logical chains. Existing methods, while improving transparency via Chain-of-Thought (CoT), lack explicit modeling of the causal dependencies between evidence and claims. In this work, we introduce a novel framework that grounds reasoning in a Structural Causal Model (SCM), treating verification as a constructive causal inference process. We empirically identify an "inverted U-shaped" correlation between reasoning chain length and accuracy, revealing that excessive structural complexity degrades performance. To address this, we propose a Rule-based Reinforcement Learning strategy using Group Relative Policy Optimization (GRPO). This approach dynamically optimizes the trade-off between structural depth and conciseness. Extensive experiments on HoVer and EX-FEVER demonstrate that our SCM-GRPO framework significantly outperforms state-of-the-art baselines, offering a reliable and interpretable solution for complex fact verification.

Online High-Definition (HD) map construction is a key component of autonomous driving. Recent methods rely on multi-view camera images for cost-effective HD map segmentation, but cameras lack depth information for accurate scene geometry. In contrast, LiDAR provides precise 3D measurements but lacks dense semantic cues. In this work, we propose LIE, LiDAR-only semantic map construction method that employ Knowledge Distillation (KD) to handle the lack of dense semantic and texture cues. Specifically, the teacher branch fuses student LiDAR features and the corresponding 2D intensity map tile to provide dense supervision for segmenting map elements using online distillation scheme. Experimental results show that our method outperforms all single-modality approaches, achieving 8.2% higher mIoU than the state-of-the-art camera-based model on nuScenes. LIE is robust over long ranges and under challenging weather and lighting, and efficiently adapts to Argoverse2 with only 10% fine-tuning, surpassing camera-based models trained on the full dataset. Source code will be available \href{https://iv.ee.hm.edu/lie/}{here}.

Predicting future clinical outcomes from electronic health records (EHR) remains challenging due to the complexity and heterogeneity of patient data. LLMs have shown strong potential for such predictive tasks, yet existing approaches mainly focus on enhancing medical knowledge through distillation or RAG while relying on the model's internal ability to interpret contextual information. In this work, we present ReMedi (Reasoner for Medical Clinical Prediction), a framework for improving clinical outcome prediction from EHR. ReMedi generates rationale-answer pairs using a challenging sample regeneration mechanism for complex clinical questions, which leverages ground-truth answers as hints to enhance reasoning for further fine-tuning and preference tuning. ReMedi integrates ground-truth outcome guidance into the preference data construction loop, regenerating rationale-answer variants. By tuning on these rationale-answer pairs, the model improves its predictive performance. Experiments on multiple EHR prediction tasks demonstrate substantial gains of up to 19.9 percent over state-of-the-art baselines in terms of F1 score, underscoring ReMedi's effectiveness in real-world clinical prediction.

Maintaining reliable UI test suites in large-scale enterprise applications is a persistent and costly challenge. We present an industrial case study of a multi-agent autonomous testing system evaluated using anonymized execution data from a production-like enterprise UI testing prototype. The application features several hundred dynamic UI elements per screen. Built on a large language model with LangGraph orchestration, Playwright execution, and a RAG knowledge base, the system evolves from human-directed testing toward High-autonomy feature discovery and test execution: given no explicit test targets, it discovers over 100 testable features across 10 UI screens, dynamically expands coverage by an additional 15--30 features through runtime DOM analysis, and iteratively repairs failing tests without human intervention. We analyzed 300 consecutive autonomous execution reports encompassing 636 individual test-case executions across 10 distinct scenario families. The system achieved a 70% repair convergence rate at the scenario-family level, with a mean of 3.4 repair iterations to convergence. However, only 10% of scenario families succeeded on first attempt, 38% of reports failed to produce any executable test artifact, and we documented concrete instances of assertion weakening and test-case deletion used as workaround mechanisms to achieve superficial convergence. Our findings show that unrestricted autonomy leads to unstable and often misleading outcomes, while constrained autonomy transforms such systems into operationally viable workflows. Rather than advocating full autonomy, our findings suggest that reliable autonomous testing in enterprise-scale settings requires explicit constraints, validation boundaries, and human oversight to preserve semantic correctness and operational trustworthiness.

Long-tailed data bias decision boundaries toward head classes and degrade tail class accuracy. Diffusion-based generative augmentation address this problem by generating additional data, while head-to-tail transfer further mitigate the generator bias inherit from long-tailed dataset. However, we show that while head-to-tail transfer helps balance the decision space of the classifier, it also induces latent non-local feature mixing that entangles inter-class features, causing decision boundary overlap and tail class distribution shift. To address this, we first identify the problem of boundary ambiguity and then propose Decision Boundary-aware Generation (DBG) framework, which promotes near-boundary representation learning by generating informative near-boundary samples. Overall, DBG rebalances the long-tailed dataset while yielding more separable decision space for long-tailed learning. Across standard long-tailed benchmarks, DBG consistently improves tail class and overall accuracy with less inter-class overlap. The code of DBG is available at https://github.com/keepdigitalabc-svg/DBG.

Although multi-modal learning has advanced point cloud completion, the theoretical mechanisms remain unclear. Recent works attribute success to the connection between modalities, yet we identify that standard hard projection severs this connection: projecting a sparse point cloud onto the image plane yields an extremely sparse support, which hinders visual prior propagation, a failure mode we term Cross-Modal Entropy Collapse. To address this practical limitation, we propose SplAttN, which replaces hard projection with Differentiable Gaussian Splatting to produce a dense, continuous image-plane representation. By reformulating projection as continuous density estimation, SplAttN avoids collapsed sparse support, facilitates gradient flow, and improves cross-modal connection learnability. Extensive experiments show that SplAttN achieves state-of-the-art performance on PCN and ShapeNet-55/34. Crucially, we utilize the real-world KITTI benchmark as a stress test for multi-modal reliance. Counter-factual evaluation reveals that while baselines degenerate into unimodal template retrievers insensitive to visual removal, SplAttN maintains a robust dependency on visual cues, validating that our method establishes an effective cross-modal connection. Code is available at https://github.com/zay002/SplAttN.

Swarm foraging algorithms, such as the central-place foraging algorithm (CPFA), typically rely on offline parameter optimization using genetic algorithms (GA) or reinforcement learning, yielding policies tightly coupled to a specific combination of team size, arena size, and resource distribution. When deployment conditions change, performance degrades, and retraining is computationally expensive. We propose LLM-Foraging, a decentralized swarm controller that augments the CPFA state machine with a large language model (LLM) tactical decision-maker at three structured decision points, namely post-deposit, central-zone arrival, and search starvation. Each robot runs its own LLM client and queries it using only locally observable state, while the existing CPFA motion and sensing stack executes the selected action. Because the LLM serves as a general decision policy rather than parameters fitted to a single configuration, the controller is training-free at deployment and transfers across configurations without re-optimization. We evaluate LLM-Foraging in Gazebo with TurtleBot3 robots across 36 configurations spanning team sizes of 4 to 10 robots, arena sizes from 6x6 to 10x10 meters, and three resource distributions (clustered, powerlaw, random). LLM-Foraging collects more resources than the GA-tuned CPFA baseline across the evaluated configurations and is more consistent, a property that the GA's single-configuration tuning does not transfer.

Single-Image Super-Resolution (SISR) aims to reconstruct a High-Resolution (HR) image from a Low-Resolution (LR) observation, a fundamentally ill-posed problem where high-frequency details are severely degraded at large upscaling factors. Recent advances have been driven by transformer-based architectures and diffusion models improve global context modeling and perceptual quality at the cost of increased computational complexity. In contrast, this work focuses on enhancing the expressivity of local operators under minimal resources. We propose SRGAN--CKAN, a hybrid super-resolution framework that integrates Convolutional Kolmogorov--Arnold Networks (CKAN) into an adversarial learning setting reformulating convolution as a nonlinear patch-based transformation. The proposed operator replaces linear local mappings with spline-based functional representations, allowing expressive modeling of complex local structures and high-frequency textures using minimal hardware resources. Experimental results demonstrate that the proposed approach improves perceptual quality while preserving reconstruction fidelity, achieving a favorable balance between distortion-based and perceptual metrics. These results are obtained under constrained computational settings, highlighting the efficiency of the proposed formulation. Overall, this work introduces a complementary direction to existing approaches by improving the representational power of local transformations, providing an efficient and scalable alternative to globally intensive architectures.

Generative models have emerged as a major paradigm for offline multi-agent reinforcement learning (MARL), but existing approaches require many iterative sampling steps. Recent few-step accelerations either distill a joint teacher into independent students or apply averaged velocities independently per agent, suggesting that few-step inference requires sacrificing inter-agent coordination. We show this trade-off is not necessary: single-pass multi-agent generation can preserve coordination when the velocity field is natively joint-coupled. We propose Coordinated few-step Flow (CoFlow), an architecture that combines Coordinated Velocity Attention (CVA) with Adaptive Coordination Gating. A finite-difference consistency surrogate further replaces memory-prohibitive Jacobian-vector product backpropagation through the averaged velocity field with two stop-gradient forward passes. Across 60 configurations spanning MPE, MA-MuJoCo, and SMAC, CoFlow matches or surpasses Gaussian / value-based, transformer, diffusion, and prior flow baselines on episodic return. Three independent coordination probes confirm that the gains flow through inter-agent coordination rather than per-agent capacity. A denoising-step sweep shows that single-pass inference suffices on every configuration. CoFlow reaches state-of-the-art coordination quality in 1-3 denoising steps under both centralized and decentralized execution. Project page: https://github.com/Guowei-Zou/coflow.

Existing evaluations of conformal prediction, such as prediction efficiency and test-conditional coverage, are defined in expectation over the calibration data. In practice, when only one calibration set of limited size is available, prediction sets often exhibit high variability in size, especially for methods with localization. We formalize this concern as set stability, defined as the variance of the conditional expectation of the set size given the calibration data. To improve stability without requiring additional target-task labels, we propose Stable Conformal Prediction (StCP), a transfer learning approach that utilizes labeled source-task data and unlabeled target data. Theoretically, we characterize the marginal coverage and stability of StCP; empirically, it delivers more stable prediction sets than standard conformal prediction methods, especially for those with localization, when calibration data are limited.

Large language models (LLMs) are rapidly being integrated into high-stakes public safety systems, including emergency call triage and dispatch decision support, yet their demographic fairness in this context remains largely untested. Here we introduce a cross-lingual audit framework that operationalizes the Police Priority Dispatch System as a five-level ordinal classification task and applies a controlled minimal-pair design to isolate the effect of demographic cues. Across 19,800 model outputs spanning 11 frontier models, 15 scenario pairs, three demographic categories (religious appearance, gender, and race), and two languages (English and Mandarin Chinese), we find that demographic bias emerges systematically when incident severity is ambiguous but largely disappears when the operational priority is clearly determined by call content. Bias magnitude varies by demographic axis, with the largest effects observed for religious appearance, followed by gender and race. Critically, bias does not transfer consistently across languages: gender bias is substantially amplified in Mandarin Chinese, whereas race bias is more pronounced in English, revealing cross-lingual asymmetries that aggregate analyses obscure. In several scenarios, demographic cues produce counter-directional effects, challenging simple stereotype-amplification accounts of model behavior. These findings suggest that bias in LLM-based dispatch is not a fixed property of models alone, but arises from the interaction between demographic signals, contextual ambiguity, and language. Beyond these empirical results, the proposed framework provides a scalable audit infrastructure that enables deploying agencies to evaluate candidate models on jurisdiction-relevant scenarios prior to real-world adoption.

Universal adversarial attacks on aligned multimodal large language models are increasingly reported with attack success rates in the 60-80% range, suggesting the visual modality is highly vulnerable to imperceptible perturbations as a prompt-injection channel. We argue that this number conflates two distinct events: (i) the model's output was perturbed (Influence), and (ii) the attacker's chosen target concept was actually emitted (Precise Injection). We compose two existing techniques -- Universal Adversarial Attack and AnyAttack -- under an $L_{inf}$ budget of 16/255, and we add a dual-axis evaluation: a deterministic Ratcliff-Obershelp drift score for Influence (programmatic baseline) plus a 4-tier ordinal categorical none/weak/partial/confirmed for Precise Injection. The judge is DeepSeek-V4-Pro in thinking mode, calibrated against Claude Opus 4.7 with Cohen's $κ$ = 0.77 on the injection axis (substantial agreement); the entire 4475-entry SHA-256 input cache ships with the dataset so reviewers can re-derive paper numbers bit-exact without an API key. Across 6615 pairs over four open VLMs, seven attack prompts, and seven test images, the two axes diverge by roughly 90$\times$: 66.4% of pairs are programmatically disturbed (LLM-judged 46.6% at the substantial-or-complete tier), but only 0.756% (50/6615) reach any non-none injection tier and only 0.030% (2/6615) verbatim. The few injections that do land cluster on screenshot- or document-style carriers whose semantics already invite text transcription. BLIP-2 shows \emph{zero detectable drift} at $L_{inf}$ = 16/255 across all 2205 pairs even when used as a Stage-1 surrogate. We release the full dataset -- 21 universal images, 147 adversarial photos, 6,615 response pairs, the v3 dual-axis judge results, and the cache at huggingface.co/datasets/jeffliulab/visinject.

There has been a growing interest in explaining entailments over description logic (DL) knowledge bases. The existing explanation formalisms focus on justifications to explain true axioms, and abductive reasoning to explain missing axioms in a knowledge base. However, these formalisms only point out the reasoning steps behind a (missing) entailment and lack a user-centered approach as they do not consider an inquirer's needs, level of understanding, or prior knowledge. We propose contrastive explanations, aiming at answering "why an axiom P (fact) is true instead of another axiom Q (foil)" over description logic knowledge bases. The motivation arises from the observation that when a user discovers that P has occurred, they are often surprised because they anticipated the occurrence of another similar event Q. Furthermore, individual explanations for "why P" and "why not Q" are unsatisfactory since a user expects to see the difference between P and Q. In this work, we first present formal foundations of contrasting questions and then define contrastive explanations within description logics. To this end, facts include ABox assertions of the form C(x) for a concept C and individual x. Possible foils for such facts are assertions C(y) (contrasting against an individual y), or D(x) (contrasting against a concept D). Additionally, we explore the properties of contrastive explanations in the DL EL and ALC. We also provide an implementation of our definition and an experimental evaluation on KBs of varying sizes.

AI language technologies (AILTs), increasingly enabled by large language models (LLMs), are becoming embedded in multilingual healthcare workflows for translation, rewriting, documentation, interpreting, and messaging in language-discordant settings. Yet fluent output is not the same as clinically safe or equitable communication: performance varies across languages, accents, tasks, and workflows, and efficiency gains can hide errors, reduce traceability, and shift responsibility across clinicians, translators, interpreters, and health systems. This narrative review synthesises recent peer-reviewed evidence across written communication, spoken communication, and emerging agentic workflows. Using the Human-Centered AI Language Technology (HCAILT) lens, it examines capabilities, evaluation practices, implementation patterns, and recurrent errors through reliability, safety culture, and trustworthiness. We identify key convergences and contradictions in the literature and propose seven grand challenges for the next phase of research and deployment. Progress, we argue, requires not only better models but also accountable sociotechnical design, calibrated human oversight, and stronger collaboration across MT/NLP, translation studies, HCI, clinical practice, implementation science, and policy.

Libraries of Low-Rank Adaptation (LoRA) adapters are becoming a practical by-product of parameter-efficient adaptation. Once such adapters accumulate, a natural question is no longer how to train one adapter for one task, but how to reuse an open pool of adapters for a new task given only a small support set. Prior work has shown that LoRA modules can be composed at the task level and dynamically selected at the instance level. However, open-pool LoRA reuse is not automatic: retrieving relevant adapters does not guarantee that their parameter updates are compatible, and composing adapters does not guarantee reliable outputs. We introduce the Sparse-Composition Agreement Layer (SCALE), a post-retrieval audit and composition framework for open-pool LoRA reuse. SCALE contains a deployable 1.0* merge path, Layer-Adaptive Sparse Residual Composition (LASRC), and a higher-cost reliability-analysis layer for multi-view disagreement. LASRC addresses merge interference by preserving a linear anchor while residualizing block-wise adapter update directions. The reliability layer treats disagreement among sparse composition views as an observable uncertainty signal and compares agreement, support-loss proxy selection, and oracle headroom under explicit path cost. In matched FLAN-T5-Large, BIG-Bench Hard (BBH), and 97-LoRA experiments, LASRC gives a directional single-view gain under fixed retrieval, while SCALE-support is reported as a query-label-free 3.0* reliability-analysis variant rather than as a calibrated or throughput-equivalent selector. Protocol-distinct BBH-8 validation shows the same qualitative trend on three decoder-only backbones. Detailed scores, paired audits, and path-cost records are reported in the experimental section.

Despite significant strides in factual reliability, errors -- often termed hallucinations -- remain a major concern for generative AI, especially as LLMs are increasingly expected to be helpful in more complex or nuanced setups. Yet even in the simplest setting -- factoid question-answering with clear ground truth-frontier models without external tools continue to hallucinate. We argue that most factuality gains in this domain have come from expanding the model's knowledge boundary (encoding more facts) rather than improving awareness of that boundary (distinguishing known from unknown). We conjecture that the latter is inherently difficult: models may lack the discriminative power to perfectly separate truths from errors, creating an unavoidable tradeoff between eliminating hallucinations and preserving utility. This tradeoff dissolves under a different framing. If we understand hallucinations as confident errors -- incorrect information delivered without appropriate qualification -- a third path emerges beyond the answer-or-abstain dichotomy: expressing uncertainty. We propose faithful uncertainty: aligning linguistic uncertainty with intrinsic uncertainty. This is one facet of metacognition -- the ability to be aware of one's own uncertainty and to act on it. For direct interaction, acting on uncertainty means communicating it honestly; for agentic systems, it becomes the control layer governing when to search and what to trust. Metacognition is thus essential for LLMs to be both trustworthy and capable; we conclude by highlighting open problems for progress towards this objective.

Generative AI disrupts the practice of giving credit to work that came before. Ideally, a generative model would give credit to any work on which its output depends in a significant way. \emph{Counterfactual credit attribution} (CCA) is a technical condition formalizing this goal--a relaxation of differential privacy--recently introduced by Livni, Moran, Nissim, and Pabbaraju [2024] who studied it in the PAC learning setting. We initiate the study of CCA generative models. Specifically, we consider autoregressive models giving credit to a deployment-time dataset (e.g., a RAG database). We uncover barriers to two natural approaches to CCA autoregressive models. First, we show that imposing CCA on the underlying next-token predictor does not guarantee that the model is CCA: CCA does not compose autoregressively (unlike DP). Second, we consider a different approach to building CCA models which we call \emph{retrofitting}. Retrofitting takes a model that does not attribute credit, and adds credit onto it. We prove a lower bound for CCA retrofitting under a weak optimality requirement. Given black-box access to the starting model, retrofitting requires query complexity exponential in the length of the model's outputs.

Multimodal learning leverages the integration of diverse data modalities to enhance performance in complex tasks. Yet, it frequently encounters incomplete or redundant modality data in real-world scenarios. This paper presents a fine-grained theoretical analysis of the generalization properties of multimodal metric learning models, addressing critical gaps in understanding the relationship between modality selection and algorithmic performance. We establish hierarchical relationships between function classes corresponding to different modality subsets and quantify the discrepancy between learned mappings and ground truth. Through rigorous analysis of pairwise complexity within the multimodal learning framework, we derive novel generalization error bounds that reveal the joint impact of modality quantity and granularity on model performance. Our theoretical findings on both upper and lower bounds demonstrate that incorporating fine-grained modality features reduces the complexity of the hypothesis space by enhancing modality complementarity. This work offers both theoretical foundations and practical implications for improving convergence rates and accuracy in multimodal learning systems.

The escalating data scale in High-Energy Physics (HEP) fuels a growing aspiration for higher analytical efficiency. While Large Language Models (LLMs) offer a path toward automation via agentic AI, they struggle with complex scientific workflows that require deep domain knowledge and are tightly coupled to experiment-specific codebases. To address this, we introduce a methodology centered on HepScript, a dual-use Domain-Specific Language (DSL) for HEP data analysis workflows. HepScript serves as a shared formal interface, abstracting HEP analysis logic into a constrained syntax that is both intuitive for human experts and reliably generable by AI agents. First developed for the Beijing Spectrometer III (BESIII) experiment, HepScript hides the complexity of the underlying software stack, translating high-level analysis intent into low-level, production-ready code. In our case studies, this abstraction reduces the required human-written code by 93\%. Crucially, HepScript's constrained grammar defines a tractable action space, enabling AI agents to autonomously generate executable specifications for core analysis stages directly from published literature with a 95\% success rate. Our work demonstrates a scalable pathway toward human-AI collaborative systems, where a formally specified DSL acts as an unambiguous translation layer between human expertise, AI automation, and production environment, rendering previously intractable automation problems solvable.

Artificial Jagged Intelligence (AJI) denotes a recurring pattern in which large learning systems exhibit strong local capabilities while remaining weak or brittle in other domains. This paper develops a formal theory of AJI as uneven allocation of optimization pressure. We model training as a finite-budget process that distributes gradient-driven update energy across capability-relevant directions in parameter space. In this model, jagged capability profiles arise from anisotropic objective structure, data geometry, and representational coupling rather than from a single scalar quantity called intelligence. The paper defines capability gain, optimization energy share, and jaggedness, then proves that persistent concentration of cumulative update energy yields lower bounds on dispersion in capability gains. A finite-budget tradeoff theorem shows why prioritizing one capability can impose opportunity costs on others unless positive coupling or shared structure offsets the cost. The analysis also studies redistribution mechanisms, including energy-variance regularization and auxiliary structural objectives, as interventions that reshape the optimization field. The resulting framework links uneven emergence, training architecture, and optimization governance. It predicts that early concentration of update energy should forecast later capability jaggedness; that scaling under a narrow objective need not eliminate anisotropy; and that explicitly funded auxiliary objectives can revive neglected capabilities. AJI is therefore not merely a descriptive label for uneven model behavior, but a testable theory of how finite optimization resources produce concentrated, delayed, and structurally uneven capability formation.

Understanding the behavior of simulated architectures in gem5 is critical for studying complex, deeply integrated computing systems. However, conventional analysis methods provide only an indirect view of the simulated system internals. In this work, we show that call-stack profiling of gem5 itself offers a powerful yet underutilized perspective: the simulator's own call-stack directly reflects the activity of the simulated system, exposing insights that conventional statistics may overlook. Profiling gem5's call-stacks is challenging due to its highly layered and complex software design patterns. To address this, we introduce a specialized, lightweight profiling framework built on Linux's perf_event interface which samples gem5's runtime call-stacks throughout the simulation, resolves symbols on the fly, and merges samples into a hierarchical call-tree representation supporting both high-level structural views and focused, user-defined, component-specific analysis. Moreover, all profiling is performed in a separate process running alongside the main gem5 process, avoiding intrusive changes and overheads to the simulation itself. We apply our framework to gem5's three major CPU models -- AtomicSimpleCPU, TimingSimpleCPU, and O3CPU -- together with the Ruby memory system, and uncover behaviors that are not easily observable in conventional gem5 statistics. Our case studies reveal, for example, that TimingSimpleCPU is inefficient due to its use of a lockup-cache model and, despite its conceptual simplicity, does not simulate faster than a full out-of-order core. In addition, our tool makes it straightforward to detect cache coherence protocol deadlock and livelock -- issues that are otherwise difficult to identify, since the simulation either appears to run normally or terminates abruptly, making it hard to pinpoint when these conditions occur.

Time-series generative models often lack control over temporal granularity, forcing users to accept whatever granularity the model produces. To enable truly user-driven generation, we introduce TimeTok, a unified framework for Granularity-Controllable Time-Series Generation (GC-TSG), which generates time series at any target granularity from any coarser input (e.g., rough sketches) or from scratch. At the core of TimeTok is a hierarchical tokenization strategy that maps time series into an ordered sequence of tokens, from coarse to fine temporal granularity. Our autoregressive generation process operates across these granularity levels, producing token blocks that are decoded back into continuous time series. This design naturally enables GC-TSG - including standard generation - within a single framework, where controlling the number of token blocks provides explicit control over output detail. Experiments show that TimeTok excels at GC-TSG tasks while achieving state-of-the-art performance in standard generation. Furthermore, we showcase TimeTok's potential as a foundational tokenizer by training on multiple datasets with heterogeneous temporal granularities, verifying strong transferability that consistently outperforms models trained on individual datasets. To our knowledge, this is the first unified framework that covers the full generative spectrum for time series, offering a valuable foundation for models that benefit from diverse temporal granularities.

Evaluating large language models (LLMs) for medical applications remains challenging due to benchmark saturation, limited data accessibility, and insufficient coverage of relevant tasks. Existing suites have either saturated, heavily depend on restricted datasets, or lack comprehensive model coverage. We introduce Medmarks, a fully open-source evaluation suite with 30 benchmarks spanning question answering, information extraction, medical calculations, and open-ended clinical reasoning. We perform a systematic evaluation of 61 models across 71 configurations using verifiable metrics and LLM-as-a-Judge. Our results show that frontier reasoning models (Gemini 3 Pro Preview, GPT-5.1, & GPT-5.2) achieve the highest performance across both benchmarks, most frontier proprietary models are significantly more token efficient than open-weight alternatives, medically fine-tuned models outperform their generalist counterparts, and that models are susceptible to answer-order bias (particularly smaller models and Grok 4). A subset of our evals (Medmarks-T) can be directly used as reinforcement learning environments to post-train LLMs for medical reasoning. Code is available at https://github.com/MedARC-AI/Medmarks

The increasing scale and complexity of online platforms raises critical policy questions around harmful content, digital well-being, and user autonomy. Traditional content moderation systems rely on centralised, top-down rules, often failing to accommodate the subjective nature of harm perception. This paper proposes an LLM-based multi-agent personalised inference framework that filters content based on unique sensitivity profiles of individual users. Our architecture combines domain-specific Expert Agents, a Manager Agent for orchestrating content analysis and agent selection, and a Ghost Profile Agent for simulating user perspectives, to inform moderation decisions. Evaluated against a range of non-personalised baselines, the system demonstrates up to a 32% improvement in accuracy, showing increased alignment with individual user sensitivities. Beyond technical performance, our framework provides policy-relevant insights for platform governance, providing a scalable way to reconcile moderation policies with societal and individual digital rights

Recent AI systems compress the distance between capability growth and capability deployment. Earlier high-risk technologies were slowed by capital intensity, physical bottlenecks, organizational inertia, and specialized supply chains. By contrast, AI capabilities can be copied, invoked, embedded in workflows, and scaled across institutions at low marginal cost. This paper argues that declining deployment friction changes the safety problem at its root. Safety is not only local output correctness or preference alignment, but the control of irreversibility under rising decision density. The paper formalizes this claim through decision-energy density: the rate-weighted capacity of a node to generate, evaluate, select, and execute consequential decisions. It then identifies three sovereignty boundaries that determine whether AI remains an amplifier within a human-governed system or becomes a de facto control center: irreversible decision authority, physical resource mobilization authority, and self-expansion authority. The model shows how efficiency pressure, path dependence, scale feedback, and weak boundary constraints concentrate decision-energy in the most efficient node. This concentration can diffuse responsibility and raise the probability of irreversible system-level loss even when local per-action error rates remain low. The main result is a boundary stabilization theorem. It shows that safety need not require proving that advanced systems are always correct. Instead, it requires institutional and technical designs that prevent irreversible power from being released by a single high-efficiency node. The paper reframes AI safety as layered control, authorization, and externally reviewable limits, linking alignment, security engineering, organizational economics, and institutional design.

Masked Language Modeling (MLM) pre-training is one of the primary ways to initialize Neural Information Retrieval (IR) models prior to retrieval fine-tuning. However, studies show that MLM pre-trained models have limited readiness and transfer learning issues for fine-tuning them into Neural Bi-Encoder models. This paper studies the effect of different pre-training datasets and pre-training options on the MLM pre-trained models for retrieval fine-tuning. The study focuses on the SPLADE-style model, which uses the MLM layer also at fine-tuning time. More specifically, we experimented with Expanded-SPLADE (ESPLADE) models, a specific instance of SPLADE models, and in-house web document titles are used as datasets. Pre-training, fine-tuning, and evaluation with optional test-time pruning of sparse vectors are conducted. Our observations are three-fold: First, fine-tuned models of higher retrieval effectiveness at both unpruned and most strict pruned settings are mostly pre-trained on a general corpus, and pre-trained with a higher learning rate, showing lower MLM accuracies. Second, in the most strict pruned setting, those models show higher-level retrieval cost and a higher variance in the length of the individual postings list. Third, the repetition of the general pre-training dataset does not have much effect on retrieval effectiveness. The experimentation empirically identifies the potential limitations for aligning MLM pre-training to ESPLADE fine-tuning. Also, the experimentation provides an empirical observation that, at most strict pruned settings, the retrieval effectiveness is better maintained by the higher-level retrieval cost, showing the trade-off relationship between the two in our setting.

Analog and mixed-signal (AMS) circuit design remains heavily reliant on expert knowledge. While recent AI-driven automation tools can generate candidate topologies, they critically depend on manually curated datasets with functional and performance annotations -- a requirement that current large language models (LLMs) and vision models cannot automate. Existing approaches still require domain experts to manually interpret circuit functionality. We present AMSnet-q, a fully automated, unsupervised pipeline that eliminates human-in-the-loop annotation by converting schematic images directly into a labeled AMS circuit database. Unlike prior work that stops at netlist extraction, our framework automates the complete verification loop: it performs schematic-to-netlist conversion, topology-aware testbench generation, and simulation-based sizing validation to objectively determine circuit functionality. Validated in 28 nm technology, AMSnet-q processed 739 schematics from the AMSnet 1.0 dataset, automatically constructing a repository of 4 circuit classes, 105 distinct topologies, and 89,789 labeled device configurations. By decoupling human effort from dataset volume and reducing the workload to a one-time testbench template per circuit class, AMSnet-q enables scalable, objective, and fully automated AMS database construction.

Multi-label node classification (MLNC) has recently been addressed by increasingly complex label-aware designs that explicitly model node-label interactions and inter-label dependencies.However, it remains unclear whether the advantages of these methods truly stem from their specialized designs, or simply from insufficiently optimized baselines. In this paper, we revisit MLNC from a strong-baseline perspective and investigate whether carefully tuned classic full-graph GNNs can already serve as strong solutions to this task. We systematically study several representative backbones, including GCN, SSGConv, and GCNII, and optimize them using standard yet effective techniques such as normalization, dropout, and residual connections. Experiments on five representative benchmark datasets show that our tuned baselines outperform representative specialized methods on four datasets and achieve state-of-the-art performance in multiple settings. These results indicate that careful tuning of classic backbones is a highly influential but often overlooked factor in MLNC, and highlight the need for more rigorous strong-baseline evaluation in future research on multi-label graph learning.

Multimodal large language models (MLLMs) struggle with numerical regression under long-tailed target distributions. Token-level supervised fine-tuning (SFT) and point-wise regression rewards bias learning toward high-density regions, leading to regression-to-the-mean behavior and poor tail performance. We identify the lack of cross-sample relational supervision as a key limitation of existing MLLM training paradigms. To address it, we propose a distribution-aware reinforcement learning framework based on Group Relative Policy Optimization, which introduces batch-level comparison-based supervision via the Concordance Correlation Coefficient-based reward to align predicted and ground-truth distributions in terms of correlation, scale, and mean. The framework is plug-and-play, requiring no architectural modification. Experiments on a unified suite of long-tailed regression benchmarks show consistent improvements over SFT and existing MLLM regression methods, with particularly strong gains in medium- and few-shot regimes.

Tacit knowledge embedded in expert practice remains difficult to capture, formalise, and scale. While AI-driven educational systems have advanced personalisation, learner modelling, affective support, and self-regulated learning, they less often model the tacit reasoning and context-sensitive judgement that underpin expert practice in practice-based domains. This paper introduces the AI Expert Twin, a cognition-centric framework that models expert knowledge as structured, computable representations of procedural actions, semantic concepts, and decision processes. The framework also considers how value-laden preferences, trade-offs, and uncertainty shape expert judgement in practice. We formalise expert cognition as a three-layer representation and capture knowledge from experts under this model, laying the groundwork for integration into AI-powered educational system. A case study in a cultural heritage workshop demonstrates the feasibility of the approach in a real-world setting. The framework is designed to be transferable across domains such as vocational education and creative industries. By embedding expert heuristics into AI while maintaining transparency and learner agency, the AI Expert Twin offers a novel path towards scalable, practice-based learning and invites further research on ethical, human-centred applications of AI in education.

Educational recommender systems (ERSs) are becoming increasingly important in enhancing educational outcomes and personalizing learning experiences by providing recommendations of personalized resources and activities to learners, tailored to their individual learning needs. While user control is widely assumed to improve user experience, the effects of different levels of control in ERSs remain underexplored. To address this gap, we designed and evaluated an interactive ERS within the MOOC platform CourseMapper, where learners could interact with the input (i.e., user profile), process (i.e., recommendation algorithm), and output (i.e., recommendations) of the system. We conducted a between-subjects user study (N=184) to examine how varying levels of user control in an ERS influenced users' perceptions of the recommendation goals of perceived control, transparency, trust, satisfaction, and perceived quality. Our results show that enabling users to build and refine their profile is sufficient to promote positive perceptions of the ERS, while additional control options mainly reinforce these impressions. Moreover, perceived control is the only goal significantly affected by providing different levels of user control in the ERS, with input control exerting the strongest influence. Furthermore, different levels of control affect transparency, trust, satisfaction, and perceived quality in distinct yet interconnected ways. Overall, the findings provide empirical evidence that user control positively shapes transparency, trust, satisfaction, and perceived quality, though to varying extents.

The conventional Retrieval-Augmented Generation (RAG) paradigm of injecting raw retrieved texts into the Large Language Model (LLM)'s context often results in suboptimal integration of retrieved information. This paper proposes to bridge retrieval results and the LLM's reasoning ability through Verbal Annotations, analytic narratives that explicitly articulate the logical connection between a search query and retrieved contexts. Our empirical investigation reveals the potential of Verbal Annotations to substantially enhance the LLM's ability to generate accurate, contextually-grounded responses. Motivated by this finding, we introduce Verbal-R3, a novel agentic RAG framework that consists of a Generator and a Verbal Reranker. The Generator performs iterative retrieval and reasoning, while the Verbal Reranker returns relevance scores and Verbal Annotations to guide the reasoning and answering process of the Generator. The inference process of Verbal-R3 is further refined through relevance-guided test-time scaling, which efficiently allocates test-time compute for effective trajectory expansion. Verbal-R3 achieves state-of-the-art performance on complex Question Answering benchmarks, validating the effectiveness of the proposed framework.

Formal specification is essential for rigorous program verification, yet writing correct specifications remains costly and difficult to automate. Although large language models (LLMs) and agents have shown promising progress, their true capabilities and failure modes remain unclear. We present the first systematic and contamination-aware study of LLM- and agent-based formal specification generation for C programs. We introduce LiveFMBench, a continuously evolving benchmark of 630 ACSL (ANSI/ISO C Specification Language)-annotated C programs, including 360 newly collected cases designed to mitigate data leakage. Using this benchmark, we evaluate direct prompting with different sampling sizes, reasoning-enabled (thinking mode) inference, the agentic pipeline, and perform a fine-grained failure analysis. Experimental results reveal that naive evaluation substantially overestimates performance because models under direct prompting may exhibit unfaithful behaviors, such as deceiving automated provers or ignoring code-context constraints; after excluding such cases, the true specification generation accuracy drops by approximately 20\%. We further find that both increased sampling and thinking mode significantly improve success rates, with smaller models benefiting more from thinking mode. Agentic pipelines are particularly effective under low sampling budgets and on harder datasets. Failure analysis further shows that incorrect loop invariants are the dominant error type, while agentic pipelines notably reduce assertion errors. These results expose fundamental limitations in current LLM-based approaches and suggest they remain far from replacing human-authored formal specifications. We release LiveFMBench at https://huggingface.co/datasets/fm-universe/Live-FM-Bench and all evaluation artifacts to support future research.

Recent advancements in Large Language Models (LLMs) have demonstrated significant potential across a wide range of software engineering tasks, including software design, an area traditionally regarded as highly dependent on human expertise and judgment. However, there has been little research focusing on how LLMs are used in software design, nor on the associated benefits and drawbacks. This paper aims to bridge this gap by empirically investigating how software developers utilize LLMs in the context of software design. We conduct a mixed-methods study, combining a mining study of 291 developer-ChatGPT conversations shared on GitHub with a survey of 65 software practitioners. Our findings reveal nine distinct categories of design tasks supported by ChatGPT, including architecture design, data model design, and the use of design patterns. We further characterize developer-ChatGPT interactions, showing that developers primarily use ChatGPT for knowledge acquisition and design-related code generation, with most tasks situated at the detailed design level. The study identifies seven key benefits of utilizing LLMs in software design as perceived by developers, such as better technology selection and the early detection of design flaws. We also uncover six limitations, including the generation of overly lengthy and difficult-to-read outputs, the creation of inexecutable or incorrect code, and a heavy reliance on context that can lead to hallucinated results. These findings provide an evidence-based characterization of current LLM use in software design from both open-source and practitioner perspectives, highlighting a tension between perceived benefits and limitations, which lays a foundation for future research and the development of effective techniques and tools to integrate LLMs into software design practices.

Large Language Models (LLMs) lack persistent memory for long-term personalized conversations. Existing graph-based memory systems suffer from information dilution, absent provenance tracking, and uniform retrieval that ignores query context. We introduce MemORAI (Memory Organization and Retrieval via Adaptive Graph Intelligence), a framework that integrates three innovations: selective memory filtering with dual-layer compression to retain user-persona-relevant content, a provenance-enriched multi-relational graph tracking factual origins at the turn level, and query-adaptive subgraph retrieval with Dynamic Weighted PageRank that applies query-conditioned edge weighting. Evaluated on LOCOMO and LongMemEval benchmarks, MemORAI achieves state-of-the-art performance in memory retrieval and personalized response generation, demonstrating that selective storage, enriched representation, and adaptive retrieval are essential for coherent, personalized LLM agents.

Differentiable physical networks provide a simple setting in which learning can be studied through the interaction between trainable parameters and physical equilibrium constraints. We investigate sequential learning in differentiable resistor networks governed by Kirchhoff's laws. Although individual input--output mappings can be learned by gradient-based adjustment of edge conductances, sequential training on conflicting tasks produces catastrophic forgetting. We show that forgetting is controlled by task conflict and by the degree of adaptation to the new task. Uniform anchoring and normalised gradient-weighted anchoring reduce forgetting only by increasing the final loss on the new task, giving a clear forgetting--adaptation trade-off. We also show that forgetting is associated with localised conductance changes on high-current edges, giving a physical interpretation as reconfiguration of dominant transport pathways. Broader random-task ensembles show that the strongest forgetting occurs when the second task reverses the output ordering imposed by the first task. Finally, comparisons across Erdős--Rényi, small-world, scale-free, and random-geometric graph ensembles show that topology changes the forgetting--adaptation balance. These results position differentiable resistor networks as compact, physically interpretable testbeds for studying continual learning in tunable matter.

Analyzing human vasculature and vessel-like, tubular structures, such as airways, is crucial for disease diagnosis and treatment. Current methods often rely on small sub-regions or simplified tree-like structures, rendering analysis of entire organ-level networks at clinical resolution computationally challenging. To this end, we propose VAEsselSparse, an efficient encoder-decoder model to obtain a meaningful yet compact representation of the entire organ-level vascular network at sub-millimeter resolution. VAEsselSparse leverages the inherent sparsity of 3D vascular structures via sparse convolutions and attention mechanisms, achieving substantial spatial compression rates of 8 x 8 x 8. We demonstrate superior reconstruction performance compared to dense counterparts and previous methods. Importantly, the resulting latent space retains clinically relevant discriminative features readily usable for classification tasks, such as aneurysm/stenosis or subvariants of the circle of Willis. Moreover, the compact latent space of VAEsselSparse serves as an effective representation for learning vessel-specific priors through generative models, enabling the synthesis of realistic vasculature.

Understanding how neural models represent human-interpretable concepts is challenging. Prior work has explored linear concept subspaces from diverse perspectives, such as probing and concept erasure. We introduce a unified framework to study these subspaces along two axes: \textit{containment}, which tests if a concept is fully represented in a subspace but not outside it, and \textit{disentanglement}, which tests for isolation from other concepts. In experiments on both text and speech models, we first highlight that concept subspaces may not be uniquely determined, and discuss the implications for concept subspace analysis. Then, we compare properties of concept subspaces estimated using five estimators, proposed in different communities. We find that (1) the choice of estimator impacts the containment and disentanglement properties; (2) the state-of-the-art concept erasure method, LEACE, performs well on both testing axes, but still struggles to generalize to unseen data; and (3) in HuBERT speech representations, phone information is both contained and disentangled from speaker information, while speaker information is hard to contain in a compact subspace, despite being disentangled from phones.

Current AI systems, grounded in oversimplified neuroscience, risk eroding the distinction between truth and falsehood. They maximize reward by amplifying attention to information without intrinsic precision mechanisms to assess whether it is valid or worth attending to. This increases both the volume of information and the inherent biases in what the system attends to, whether true, false, or irrelevant. If not corrected, this trend will accelerate, threatening to overload systems and individuals with biased and dubious information and increasing the risk of confusion, poor judgment, and irrational or harmful decisions and behaviour, a condition I term the mind-reality overload dilemma. I argue that this threat may be mitigated by providing the public with access to more advanced AI tools built on the biophysical dynamics of pyramidal neurons underlying awake thought and higher-order cognition. These neurons support an intrinsic active precision mechanism that, rather than merely maximizing reward, uses locally and globally coherent predictions to evaluate the validity and contextual adequacy of evidence before it is attended to or propagated through hierarchies, prioritizing coherence and adequacy before attention.~While this approach does not derive or prescribe moral rules from biology, it may give rise to AI with more "real understanding", helping restore epistemic conditions by reducing information overload and amplifying reliable information, thereby supporting the formation of better-informed beliefs and more coherent judgments that benefit society at large-though no guarantees exist.

Knowledge distillation is a key technique for compressing large language models (LLMs), but most existing methods align representations at fixed layers or token-level outputs, ignoring how representations evolve across depth. As a result, the student is only weakly guided to capture the teacher's internal relational structure during distillation, which limits knowledge transfer. To address this limitation, we propose Multi-Granular Trajectory Alignment (MTA), a framework that aligns teacher and student representations along their layer-wise transformation trajectory. MTA adopts a layer-adaptive strategy: lower layers are aligned at the word level to preserve lexical information, while higher layers operate on phrase-level spans (e.g., noun and verb phrases) to capture compositional semantics. We instantiate this idea through a Dynamic Structural Alignment loss that matches the relative geometry among semantic units within each layer. This design is motivated by empirical findings that Transformer representations become increasingly abstract with depth, and is also consistent with linguistic views in which higher-level meaning emerges through the composition of lower-level lexical units. We further incorporate a Hidden Representation Alignment loss to directly align selected teacher-student layers. Experiments show that MTA consistently outperforms state-of-the-art baselines on standard benchmarks, with ablations confirming the contribution of each component.

The iterative denoising paradigm of Diffusion Large Language Models (DLMs) endows them with a distinct advantage in global context modeling. However, current decoding strategies fail to leverage this capability, typically exhibiting a local preference that overlooks the heterogeneous information density within the context, ultimately degrading generation quality. To address this limitation, we systematically investigate high-information-density (HD) tokens and present two key findings: (1) explicitly conditioning on HD tokens substantially improves output quality; and (2) HD tokens exhibit an early-decoding tendency, converging earlier than surrounding tokens. Motivated by these findings, we propose Focus on the Core \textbf{(FoCore)}, a training-free decoding strategy that utilizes HD tokens in a self-contrast manner, wherein HD tokens are temporarily remasked as negative samples, to guide generation. We further introduce FoCore\_Accelerate \textbf{(FoCore\_A)}, an efficient variant that, upon detecting HD token convergence, performs parallel decoding over stable candidates within a local context window, substantially accelerating generation. Extensive experiments on math, code and logical reasoning benchmarks demonstrate that FoCore consistently improves generation quality and efficiency across both LLaDA and Dream backbones. For instance, on HumanEval, FoCore improves pass@1 from 39.02 to 42.68 over standard Classifier-Free Guidance, while FoCore-A reduces the number of decoding steps by 2.07x and per-sample latency from 20.76s to 8.64s (-58.4\%).

Large language models (LLMs) have been widely explored for embedding generation. While recent studies show that in-context learning (ICL) effectively enhances the representational capability of LLMs by prepending a few task-related demonstrations, it causes substantial token overhead due to the increased sequence length. In this work, we propose EPIC, a novel embedding-based in-context prompt training strategy that leverages ICL to generate high-quality embeddings while reducing computational burden during both training and inference. This approach replaces discrete text demonstrations with their corresponding continuous embeddings, which not only encourages the LLM to align semantically-related text pairs during contrastive learning, but also requires the model to interpret demonstration embeddings as part of the in-context prompt. Consequently, EPIC-trained models achieve excellent embedding performance both with or without in-context prompts at inference time. Comprehensive experiments demonstrate that our method establishes new state-of-the-art results on the MTEB benchmark, surpassing frontier models trained solely on publicly available retrieval data. Extensive ablation studies further validate the effectiveness and necessity of our mechanism.

Deep learning has been widely adopted for WiFi CSI-based human activity recognition (HAR) due to its ability to learn spatio-temporal features in a privacy-preserving and cost-effective manner. However, DL-based models generalize poorly across environments, a challenge amplified in multi-user settings where overlapping activities cause CSI entanglement and domain shifts. Practical deployments often limit access to labeled source data due to privacy constraints, motivating source-free adaptation using only unlabeled target-domain CSI and a pre-trained source model. In this paper, we propose MU-SHOT-Fi, a source-free unsupervised domain adaptation framework for single- and multi-user Wi-Fi sensing. MU-SHOT-Fi employs permutation-invariant set prediction with Hungarian matching during source training, followed by frozen-classifier backbone adaptation in the target domain. To enable stable adaptation without labels, we introduce occupancy-weighted information maximization that prevents model collapse by focusing diversity regularization on likely-occupied slots while excluding the dominant class from marginal entropy. Additionally, we employ binary rotation prediction as spatial self-supervision that exploits CSI frequency-time structure to learn domain-invariant features. For single-user scenarios, we introduce SU-SHOT-Fi by replacing occupancy weighting with standard information maximization and incorporating contrastive predictive coding to exploit temporal consistency. Extensive experiments on the WiMANS and Widar 3.0 datasets across cross-environment, cross-frequency, cross-orientation, and combined domain shifts demonstrate that MU-SHOT-Fi effectively recovers multi-user exact-activity classification performance under large domain shifts while maintaining accurate occupancy estimation and preventing collapse toward dominant classes.

High-fidelity circuit execution on noisy intermediate-scale quantum devices is bottlenecked by compilation pipelines that disregard complex, correlated noise. To address this, this methodology article proposes a quantum machine learning control (QMLC) framework for generative quantum circuit synthesis from gate-set tomography (GST) data that bypasses the traditional two-step pipeline of characterizing native quantum gates via GST followed by unitary decomposition algorithms. Instead, a generative concept space is directly learnt from GST data, enabling conditional synthesis of quantum circuits on a desired output distribution. Our approach tokenizes GST germ circuits and embeds them into a structured latent space using a curriculum-learning-motivated strategy, starting with short circuits and progressively incorporating longer ones with diverse output statistics. The embedded sequences are processed by a set-vision transformer with permutation-invariant pooling, producing k-seed vectors that represent the learned concept space of the quantum device. Aggregating data across multiple circuits makes this latent representation inherently context-aware, capturing the shared physical noise environment (e.g., crosstalk, drift) that isolated gate metrics miss. We propose an unconditional diffusion model to sample from the concept space. During inference, a user provides a target measurement distribution, and the model generates a corresponding circuit. To ensure fidelity and robustness, the output is denoised using a diffusion model that operates on the target conditional covariance matrix. This end-to-end framework is a step towards context-aware, hardware-native circuit synthesis directly from raw GST data, which offers a new paradigm for integrating quantum control and compilation. The QMLC framework is particularly suited for near-term quantum devices with complex calibration procedures.

The building energy community lacks a foundational thermal model, i.e., a single pretrained model capable of generalizing across diverse buildings, climates, and control strategies without building-specific calibration. Achieving this vision requires architectural principles that capture universal thermal dynamics rather than memorizing building-specific patterns. We take a step toward this goal by presenting a physics-informed transformer architecture that embeds domain knowledge, e.g., derivative enrichment and Euler-based numerical integration, into a decoder-only framework. We incorporate static building features extracted from simulation models and employ Rotary Position Embedding attention to capture temporal dependencies. Evaluated on the CityLearn dataset spanning 247 residential buildings across three climate zones, our model achieves one-step prediction accuracy (RMSE of 0.30°C in Texas, 0.29°C in Vermont) while outperforming both traditional baselines and fine-tuned Time-Series Foundation Models. We also demonstrate zero-shot transferability: models trained on as few as two buildings generalize to unseen buildings and climate zones without fine-tuning. Despite the limitation of simulated residential buildings, our results establish physics-informed architectural principles as a promising foundation for universal building thermal models.

Flavor-tagging calibrations are often provided either as scale factors measured at a finite set of working points or as binned corrections to a chosen one-dimensional discriminant. However, this approach falls short of providing continuous, event-level calibration across the full multicomponent outputs of modern taggers. This limitation leads to information loss in analyses that demand high-performance flavor tagging, restricting analyses to a limited set of predefined variables. In this work, we propose a geometry-aware framework that formulates flavor-tagger calibration as an optimal transport problem on the probability simplex. The transport maps are parameterized and trained in the isometric log-ratio coordinate system. Because the quadratic Euclidean cost of Brenier transport in this coordinate system is equivalent to the Aitchison distance on the simplex, the learned map induces a minimal deformation under the Aitchison geometry. Furthermore, we extract flavor-conditional target distributions directly from control-region data using an expectation-maximization (EM) technique that simultaneously fits multiple control regions, models each flavor component with a normalizing flow, and estimates the regional mixture fractions. The extracted targets are subsequently used to learn flavor-factorized transport maps. Because the joint estimation of mixture fractions and flexible component densities admits weakly constrained directions, we further introduce a linearized feedback-operator analysis that propagates the fitted composition covariance into the extracted component densities, separating data-constrained modes from those dominated by the composition prior. The simulation-based closure study demonstrates improved closure in dedicated control regions and in independent validation mixtures.

Decision-Focused Learning (DFL) trains predictors to improve downstream decision quality, but computing regret gradients typically requires differentiating through solvers or relying on surrogate losses, which can be computationally expensive or deviate from the true objective. We show that, under standard regularity with locally stable active constraints, the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature. This reveals that regret gradients can be obtained by filtering decision-irrelevant components from the MSE gradient, providing a simpler and more direct alternative to existing approaches. Based on this, we propose PEAR (Projected Error As Regret-gradient), which computes regret gradients via a reduced linear system over active constraints, avoiding differentiation through solver iterations or additional optimization solves. Experiments on LP benchmarks and a real-world QP task show that PEAR achieves the best decision quality among all baselines while being the most computationally efficient, with gains that persist under constraint shifts.

In this paper, we employ the Minimal Cognitive Grid (MCG), a framework created to evaluate the cognitive plausibility of artificial systems, to offer a systematic assessment of leading computational models of analogy and metaphor, including the Structure-Mapping Engine (SME), CogSketch, METCL, and Large Language Models (LLMs). We present a formal and quantitative operationalization of the MCG framework and, through the analysis of its three main dimensions (Functional/Structural Ratio, Generality, and Performance Match), examine how well each system aligns with standard cognitive theories of the modeled phenomena, thus allowing for comparison of the models with respect to their cognitive plausibility, according to consistent and generalizable mathematical criteria.

Unsupervised Environment Design (UED) offers a promising paradigm for improving reinforcement learning generalization by adaptively shaping training environments, but it requires reliable environment evaluation to remain effective. However, existing UED methods evaluate environments using indirect proxy signals such as regret, value-based errors, or Monte Carlo, which suffer from bias, high variance, or substantial computational overhead and fail to reflect agent realized learning progress. To address these limitations, we propose Parameter Change Environment Design (PACE), which evaluates an environment through the policy parameter change induced by training on that environment, directly grounding environment selection in realized learning progress. Specifically, PACE assigns environment value using a first-order approximation of the policy optimization objective, where the improvement induced by an environment is proportional to the squared L2 norm of the corresponding parameter update, enabling low-variance and computation-efficient evaluation without additional rollouts. Experiments on MiniGrid and Craftax show that PACE consistently outperforms established UED baselines, achieving higher IQM and smaller Optimality Gap on OOD evaluations, including an IQM of 96.4% and an Optimality Gap of 17.2% on MiniGrid.

Large Language Models (LLMs) excel at long-context understanding but exhibit significant limitations in long-form generation. Existing studies primarily focus on single-generation quality, generally overlooking the volatility of the output. This volatility not only leads to significant computational costs but also severely impacts the models' reliable application. To address this gap, our work unfolds in three stages: benchmarking, probing, and mitigation. We first propose the VOlatility in Long-form Text Benchmark (VOLTBench), a novel heterogeneous-task benchmark designed to systematically quantify the length volatility of long-form generation. Subsequently, by analyzing attention traces, we conduct an in-depth probe to identify several common internal patterns that cause this volatility. Finally, to mitigate long-form output volatility, we propose Stable Generation via Logits Boosting (GLoBo), a lightweight decoding-stage optimization strategy, designed to significantly enhance both the length accuracy and stability of long-form generation without additional training. Extensive experiments on VOLTBench provide the first systematic confirmation of severe long-form output instability in mainstream models and validate that our proposed method successfully improves the mean output length of the base model by 148% and reduces the length volatility by 69%, while maintaining high generation quality.

Learning constraint-satisfying policies from offline data without risky online interaction is crucial for safety-critical decision making. Conventional methods typically learn cost value functions from abundant unsafe samples to define safety boundaries and penalize violations. However, in high-stakes scenarios, risky trial-and-error is infeasible, yielding datasets with few or no unsafe samples. Under this limitation, existing approaches often treat all data as uniformly safe, overlooking safe-but-infeasible states - states that currently satisfy constraints but inevitably violate them within a few steps - leading to deployment failures. Drawing inspiration from the concept of knowledge-data integration, we leverage large language models (LLMs) to incorporate natural language knowledge into the policy to address this challenge. Specifically, we propose PROCO, a model-based offline safe reinforcement learning (RL) framework tailored to datasets largely free of violations. PROCO first learns a dynamics model from offline data and constructs a conservative cost function by grounding natural-language knowledge of unsafe states in LLMs, enabling risk estimation even without observed violations. Using the cost function and learned model, PROCO performs model-based rollouts to synthesize diverse counterfactual unsafe samples, supporting reliable feasibility identification and feasibility-guided policy learning. Across a range of Safety-Gymnasium tasks with exclusively safe or minimally risky training data, PROCO integrates seamlessly with a variety of offline safe RL algorithms and consistently demonstrates reduced constraint violations and improved safety performance compared to both the original methods and other behavior cloning baselines.

Automated leaf disease classification is critical for early disease detection in resource-constrained field environments. Vision Transformers (ViTs) provide strong representation capability by modeling long-range dependencies and inter-class relationships; however, their high computational cost makes them impractical for deployment on edge devices. As a result, existing approaches struggle to effectively transfer these rich representations to lightweight models. This paper introduces AgriKD, a cross-architecture knowledge distillation framework for efficient edge deployment, which transfers knowledge from a Vision Transformer (ViT) teacher to a compact convolutional student model. To bridge the representational gap between Transformer and CNN architectures, the proposed approach integrates multiple distillation objectives at the output, feature, and relational levels, where each objective captures a different aspect of the teacher knowledge. This enables the student model to better preserve and utilize transformer-derived global representations. Experiments on multiple leaf disease datasets show that the distilled student achieves performance comparable to the teacher while significantly improving efficiency, reducing model parameters by approximately 172 times, computational cost by 47.57 times, and inference latency by 18-22 times. Furthermore, the optimized model is deployed across multiple runtime formats, including ONNX, TFLite Float16, and TensorRT FP16, achieving consistent predictive performance with negligible accuracy degradation. Real-world deployment on NVIDIA Jetson edge devices and a mobile application demonstrates reliable real-time inference, highlighting the practicality of AgriKD for AI-powered agricultural applications in resource-constrained environments.

GPU-based simulation environments for embodied AI interleave physics simulation (CUDA) and photorealistic rendering (Vulkan) on a single device. We observe that two foundational scenarios -- simulation data generation and RL training -- can be naturally adapted to execute their simulation and rendering phases concurrently, presenting a significant opportunity to improve GPU utilization through spatial multiplexing. However, a fundamental obstacle we term execution isolation prevents this: CUDA and Vulkan create separate GPU contexts whose channels are bound to different scheduling groups, confining compute and graphics to mutually exclusive time slices. Existing spatial-sharing techniques are limited to the CUDA ecosystem, while temporal-sharing approaches underutilize available resources. This paper presents VUDA, a system that breaks execution isolation to enable spatial parallelism between CUDA compute and Vulkan graphics workloads. VUDA is built on two key observations: although CUDA and Vulkan expose different programming abstractions, their execution paths converge to a common channel primitive at the driver and hardware level; meanwhile, their virtual-address spaces are inherently disjoint, making safe page-table merging feasible without remapping. VUDA exposes a thin API for developers to annotate co-schedulable CUDA streams, and realizes spatial sharing through channel redirection into Vulkan's scheduling domain and page-table grafting to unify address spaces, eliminating all data copying on the critical path. Experiments on representative embodied-AI workloads show that VUDA delivers up to 85% higher throughput than temporal-sharing baselines, while improving GPU utilization and reducing end-to-end latency.

This paper presents the rAIson platform, a high-level technological environment for the development of automated, reliable and explainable decision-making agents. The research underlying the platform and its technological progress has now reached a mature stage that allows the platform to be used for the development of complex real-life applications without writing a single line of code.

Detecting machine-generated text is essential for transparency and accountability when deploying large language models (LLMs). Among detection approaches, watermarking is a statistically reliable method by design -- it embeds detectable signals into LLM outputs by biasing their token distributions. However, it has been reported that watermarked LLMs often perform worse on downstream tasks. We propose PUPPET, a framework that fine-tunes an LLM via reinforcement learning to generate text that is both more detectable and better performing on downstream tasks. We use two reward functions: a detector that outputs a machine-class likelihood and an evaluator that measures a task-specific metric. Experiments on long-form QA, summarization, and essay writing show that LLMs trained with PUPPET achieve high detectability competitive with watermarking methods while outperforming them on downstream tasks. The analysis shows that this optimization can be performed efficiently with only a few thousand samples in 1--2 GPU hours. Moreover, these gains are consistent across out-of-domain tasks, different LLM families, and model sizes, and are even robust to paraphrasing attacks.

On-policy distillation (OPD) trains a student on its own trajectories under token-level teacher supervision, but existing methods are capped by a single-teacher capability ceiling: when the teacher errs, the student inherits the error. OPD also remains largely unexplored in agentic tasks, where per-step errors compound across long trajectories and destabilize training. We propose MAD-OPD (Multi-Agent Debate-driven On-Policy Distillation), which breaks this ceiling by recasting the distillation teacher as a deliberative collective of teachers that debate over the student's on-policy state; the debate produces an emergent collective intelligence that supplies token-level supervision, with each teacher's contribution weighted by its post-debate confidence. To extend OPD to agentic tasks, we also introduce On-Policy Agentic Distillation (OPAD), which adds step-level sampling to stabilize training under multi-step error compounding. We additionally derive a task-adaptive divergence principle, selecting JSD (Jensen-Shannon divergence) for agentic stability and reverse KL (Kullback-Leibler) divergence for code generation, and verify it both theoretically and empirically. Across six teacher-student configurations (Qwen3 and Qwen3.5; 1.7B-14B students, 8B-32B teachers) and five agentic and code benchmarks, MAD-OPD ranks first across all six configurations; on the 14B+8B$\to$4B setting it lifts the agentic average by $+2.4\%$ and the code average by $+3.7\%$ over the stronger single-teacher OPD.

Visual perception in modern Vision-Language Models (VLMs) is constrained by a fundamental perceptual bandwidth bottleneck: a broad field of view inevitably sacrifices the fine-grained details necessary for complex reasoning. Inspired by the classical paradigms of active vision and information foraging, we frame overcoming this limitation as a sequential decision-making process. We formalise this process through the lens of the sequential Bayesian optimal experimental design (S-BOED) problem. While exact Bayesian inference is intractable in continuous gigapixel spaces, we derive principled yet tractable approximations that balance spatial coverage against resolution. To validate this framework, we present a training-free inference strategy as a practical instantiation of the S-BOED objective for agents equipped with multiple vision tools. Designed as a flexible template, this strategy accommodates arbitrary optimisation algorithms, ranging from efficient greedy sampling to look-ahead planning, to approximate the optimal design. Empirical evaluations on gigapixel-level benchmarks demonstrate that our approach further boosts the performance of state-of-the-art models, significantly outperforming standard baselines and effectively narrowing the gap towards human-annotated oracles.

Given a knowledge base (KB) with a non-entailed fact, the ABox abduction problem asks for possible extensions of the KB that would entail this fact. This problem has many applications, ranging from diagnosis to explainability and repair. ABox abduction has been well-investigated for consistent KBs and classical semantics, but little is known for the case of inconsistent KBs, which can be caused by erroneous data. In this paper we define suitable notions of abduction in this setting and propose criteria that guide abduction towards "useful" hypotheses. To regain meaningful reasoning in the presence of inconsistencies, we use well-established repair semantics. We provide a comprehensive landscape of the complexity of ABox abduction under repair semantics, treating different variants of the abduction problem for the light-weight description logics DL-Lite and EL_bot.

Learning-based approaches to verifying unknown Markov decision processes (MDPs) often employ uncertain MDPs. These models use, for example, confidence intervals to capture transition uncertainty and allow synthesis of policies that are robust to this uncertainty. However, this approach typically quantifies uncertainty independently for individual transition probabilities, ignoring dependencies due to shared latent quantities. We propose to learn such models using parametric MDPs (pMDPs), where transition probabilities are expressions over a set of parameters. We project statistical uncertainty from empirical transition frequencies onto the pMDP's parameter space, yielding a probably approximately correct (PAC) uncertainty model for the underlying MDP that respects the algebraic dependencies between transitions. The resulting models are algorithmically challenging to solve, so we propose a hierarchy of sound polytopic outer approximations of the induced confidence set. We implement and evaluate our approach, demonstrating substantially tighter uncertainty estimates than classical interval-based uncertain MDP learning techniques.

System-level diagrams encode the architectural blueprint of chip design, specifying module functions, dataflows, and interface protocols. However, non-standardized symbols and the scarcity of structured training data hinder existing multimodal large language models (MLLMs) from recognizing these diagrams. To address this gap, we introduce DiagramNet, the first multimodal dataset for system-level diagrams, comprising 10,977 connection annotations and 15,515 chain-of-thought QA pairs across four tasks: Listing, Localization, Connection, and Circuit QA. Building on this dataset, we propose a progressive training pipeline together with a decoupled multi-agent workflow that decomposes complex visual reasoning into Perception, Reasoning, and Knowledge stages. On the DiagramNet benchmark, integrating our 3B-parameter model with the proposed workflow surpasses the 2025 EDA Elite Challenge winner and outperforms GPT-5, Claude-Sonnet-4, and Gemini-2.5-Pro by over 2x in end-to-end evaluation. Notably, the workflow generalizes beyond our model, boosting Task 1 performance by 128.7x for Gemini-2.5-Pro and 12.4x for GPT-5. Furthermore, with only 60 images for detector adaptation, the method transfers effectively to AMSBench, achieving zero-shot connectivity reasoning on par with GPT-5 and Claude-Sonnet-4 while surpassing the AMS state-of-the-art method Netlistify.

News outlets shape public opinion at a scale that makes automated detection of political bias and factuality essential. However, the field still lacks unified resources, comprehensive evaluations across diverse approaches, and systematic analyses of the representations and fusion strategies that matter most, especially under label sparsity and dataset diversity. In addition, there is little empirical work reporting broad, observation-driven findings about what consistently works, what fails, and why. We address these gaps through four main contributions. First, we introduce MBFC-2025, a large-scale label set covering approximately 2,600 outlets from Media Bias/Fact Check (MBFC). Second, we construct multiview representations for ACL-2020 (Panayotov et al., 2022), which includes around 900 outlets, as well as for MBFC-2025. These representations span Alexa graphs, hyperlink graphs, LLM-derived graphs, articles, and Wikipedia descriptions. Third, we provide a systematic evaluation and analysis of embedding views and fusion strategies, including a reinforcement learning-based fusion variant. Fourth, we conduct extensive experiments that achieve state-of-the-art results on ACL-2020 and establish strong benchmarks on MBFC-2025.

We characterize the fundamental limits of high-dimensional mean testing under arbitrary truncation, where samples are drawn from the conditional distribution $P(\cdot \mid S)$ for an unknown truncation set $S$ that may hide up to an $\varepsilon$-fraction of the probability mass. For distributions with $p$-th directional moments of magnitude at most $ν_{P,p}$, truncation induces a bias of order $O(ν_{P,p}\varepsilon^{1-1/p})$. This bias creates a sharp information-theoretic detectability floor: when the signal $α$ falls below this threshold, the null and alternative hypotheses are indistinguishable even with infinite data. Above this floor, we prove that a simple second-order test achieving near-optimal sample complexity $n = O\!\left(\frac{\|Σ_P\|}{(α-4ν_{P,p}\varepsilon^{1-1/p})^2}\sqrt{d}\right)$. We further identify a structural escape from this finite-moment bias barrier. Under a directional median regularity assumption, truncation bias improves to linear order $O(\varepsilon)$. This reveals an intermediate regime in which estimation requires $Θ(d)$ samples for uniform recovery, while testing recovers the classical $Θ(\sqrt d)$ rate once truncation bias is eliminated. Together, our results provide a unified framework for mean testing under truncation, connecting finite-moment, sub-Gaussian, and median-regular structural regimes.

Multimodal large language models (MLLMs) have emerged as a promising paradigm for dental image analysis. However, their ability to capture the multi-level cognitive processes required for radiographic analysis remains unclear. Here, we present a comprehensive benchmark to evaluate the cognitive capabilities of MLLMs in dental radiographic analysis. It spans three critical imaging modalities, i.e., periapical, panoramic, and lateral cephalometric radiographs, and defines four cognitive categories: perception, comprehension, prediction, and decision-making. The benchmark comprises 27 clinically grounded tasks derived from public datasets, with manually curated annotations and 3,820 clinician assessments for evaluation. Six frontier MLLMs, including GPT-5.2 and GLM-4.6, are evaluated. We demonstrate the performance gap between MLLMs and clinicians in dental practice, delineate model strengths and limitations, characterize failure patterns, and provide recommendations for improvement. This data resource will facilitate the development of next-generation artificial intelligence systems aligned with clinical cognition, safety requirements, and workflow complexity in dental practice.

In-group favoritism refers to the phenomena of favoring members of one's in-group over out-group members and is widely observed in numerous social cooperative behaviors. Recently, in-group favoritism biases have also been identified in generative language models. However, whether the in-group favoritism exists when persona agents are faced with contradicting information (e.g., misinformation), and how to mitigate the adverse effects of in-group favoritism biases in persona agents have been understudied. To address these problems, we propose a Truth or Tribe simulation framework to study the agent cooperation within the spread of contradicting information through a triadic interaction paradigm, and conduct controlled trials to evaluate the primary moderating factors. Extensive results showcase that persona agents display strong in-group favoritism, accepting incorrect answers from identity-similar peers at much higher rates than from dissimilar peers. In-group favoritism continues to emerge in defeasible reasoning contexts where no absolute truth exists, and it intensifies as cognitive complexity increases. Furthermore, three intervention strategies--Identity-Blind Instruction, Structured Counterfactual Reasoning, and Heterogeneous Perspective Ensemble--are proposed to mitigate the in-group favoritism.

Existing reinforcement learning approaches for Large Language Models typically perform policy optimization at the granularity of individual tokens or entire response sequences. However, such formulations often misalign with the natural step-wise structure of reasoning processes, leading to suboptimal credit assignment and unstable training in multi-modal reasoning tasks. To bridge this gap, we propose Segment-Aligned Policy Optimization (SAPO), a novel reinforcement learning paradigm that treats coherent reasoning steps, rather than tokens or full sequences as fundamental units of policy update. SAPO introduces a step-wise Markov decision process abstraction over reasoning segments, accompanied by segment-level value estimation, advantage computation, and importance sampling mechanisms that are semantically aligned with reasoning boundaries. Experiments on representative reasoning benchmarks demonstrate that SAPO consistently outperforms token-level and sequence-level policy optimization methods, achieving significant accuracy improvements while exhibiting better training stability and value estimation consistency. Our work underscores the importance of aligning reinforcement learning updates with the intrinsic structure of reasoning, paving the way for more efficient and semantically grounded policy optimization in complex reasoning tasks. Codes and models will be released to ensure full reproducibility.

Vision-Language Models (VLMs) have enhanced traditional LLMs with visual capabilities through the integration of vision encoders. While recent works have explored various combinations of vision encoders and LLMs, there still lacks a principled understanding of what makes a vision encoder suitable for VLM alignment. In this paper, we systematically investigate this question via comprehensive experiments on a curated collection of 19 pre-trained vision encoders from diverse sources. We first demonstrate that common practices, such as choosing encoders with the largest size or highest zero-shot accuracy, consistently fail to identify optimal models. In fact, these metrics show only weak to moderate correlation with VLM performance. This intriguing finding begs a fundamental question: What factors of vision-encoders matter in VLM? Through comprehensive analysis, we identify that the structural similarity across modalities plays a crucial but previously overlooked role in vision-encoder selection, which we measure using the Gromov-Wasserstein distance as a proxy. From a theoretical perspective, we show that the learnability of cross-modality mapping can be provably associated with the Gromov-Wasserstein distance. Empirical verification on 60+ full VLM training runs shows that our proposed inference-only metric performs significantly better than alternative model selection strategies and exhibits a much stronger correlation with final VLM performance, thereby enabling efficient and effective prediction of VLM performance before full training.

In this article, we introduce SiNFluD, a novel benchmark dataset for Sindhi figurative language classification. We first collect raw text from various blogs, social media platforms, and literary sources, and subsequently prepare the corpus for annotation. Two native annotators label the data using the Doccano text annotation tool, achieving an inter-annotator agreement of 0.81. We then establish baseline results using 5-fold and 10-fold cross-validation. Finally, we evaluate mBERT, XLM-RoBERTa, and XLM-RoBERTa-XL models, along with SetFit for few-shot fine-tuning of sentence transformers. Among these, the pretrained XLM-RoBERTa-XL achieves the best performance.

This study presents a comparative analysis between two primary approaches in Natural Language Processing (NLP): Machine Learning (ML) utilizing the PyCaret AutoML framework, and Deep Learning (DL). The evaluation is conducted on a sentiment analysis task using an Indonesian e-commerce review dataset sourced from Hugging Face. The dataset, consisting of 15,000 samples, is partitioned into training, validation, and testing sets. The ML experiments compare LightGBM, Logistic Regression, and Support Vector Machine (SVM) algorithms, whereas the DL experiment implements a Bidirectional Long Short-Term Memory (BiLSTM) architecture. The experimental results demonstrate that the BiLSTM model outperforms all ML models, achieving an accuracy of 98.87\% and an F1-Score of 98.87\%. Meanwhile, LightGBM emerges as the best-performing ML model with an accuracy of 98.23\% in a highly efficient training time. This research proves that the BiLSTM architecture is highly capable of capturing the sequential context of Indonesian review texts, making it the superior model for this specific classification task.

Barren plateaus (BPs) are usually described by the exponential suppression of gradient variance, but the mechanism by which gradient signal disappears remains unclear. We show that this phenomenon can be understood as destructive interference among termwise gradient contributions. To make this perspective operational, we introduce a diagnostic framework based on the cancellation ratio $R_k$, the effective term count $N_{\mathrm{eff},k}$, and the interference-quality measure $B_{\mathrm{eff},k}=R_k\sqrt{N_{\mathrm{eff},k}}$. Under a random-sign model, $B_{\mathrm{eff},k}$ remains near a stable baseline, defining a random-sign cancellation regime. For the transverse-field Ising model (TFIM), we find that the hardware-efficient ansatz (HEA) remains close to this regime across system sizes and depths, whereas the Hamiltonian variational ansatz (HVA) systematically escapes it. In particular, HVA exhibits larger $B_{\mathrm{eff},k}$ not merely because $N_{\mathrm{eff},k}$ is larger, but because $R_k$ also remains systematically larger despite the broader term participation. This pattern indicates improved sign organization rather than simple term suppression. We further establish an exact identity that connects the proposed interference diagnostics directly to the standard variance-based theory of BPs. These results position destructive interference as a mechanistic interpretation of BP-like behavior in the regimes studied here, but they do not imply that BPs and destructive interference are universally interchangeable across all architectures and settings.

This paper compares Machine Learning and LSTM-based Deep Learning methods for sentiment analysis of Mobile Legends app reviews. Using a dataset of 10,000 reviews labeled as positive, negative, and neutral, the study evaluates traditional models with TF-IDF and PyCaret AutoML and compares them against an LSTM model designed to capture sequential text dependencies. The results show that the LSTM model outperforms the classical Machine Learning baselines, achieving 92% accuracy and a weighted F1-score of 91%. The findings indicate that deep learning is more effective for handling informal and context-dependent user review text.

This paper investigates sentiment classification of Steam game reviews using an attention-based Bidirectional Long Short-Term Memory (BiLSTM) model. Using a dataset of 50,000 reviews sampled from a larger Steam review corpus, the authors compare a traditional machine learning baseline based on TF-IDF and PyCaret AutoML with a deep learning approach implemented in PyTorch. The proposed BiLSTM+Attention model is trained with class-weighted cross-entropy to address class imbalance and achieves 83% accuracy and 85% weighted F1-score on the test set, with 90% recall for negative reviews. The paper also presents attention visualizations to show interpretability by highlighting sentiment-bearing words. The study concludes that the BiLSTM+Attention model is effective for analyzing user sentiment in Steam reviews and useful for helping developers understand player feedback.

Offline evaluation of language models from usage logs is biased when model choice is confounded: the same user-side factors that influence which model is used can also influence how its output is judged, so raw comparisons of logged scores mix self-selected populations rather than estimating a common quantity of interest. A small randomized experiment can break this bias by overriding model choice, but in practice such experiments are scarce and costly. We study a three-source design that combines a large confounded observational log (OBS) for scale, a small randomized experiment (EXP) for unconfounded scoring, and an offline simulator (SIM) that replays candidate models on cached contexts. Our main result is an identification theorem showing that the randomized experiment and the simulator are together enough to recover causal model values; the observational log enters only afterward, to reduce estimation error rather than to make the causal comparison valid. Six estimator families are evaluated in a controlled semi-synthetic validation and in two real-task cached benchmarks for summarization and coding. No family dominates every regime; relative performance depends on the amount of unbiased EXP supervision and on how closely the target reward aligns with OBS-derived structure.

Graph self-supervised learning typically relies on large-scale unlabeled datasets, heavily inflating computational costs. However, empirical evidence suggests that these datasets contain substantial redundancy-our analysis reveals that uniformly subsampling 50% of graphs retains over 96% of downstream performance. To exploit this redundancy, we introduce GraphSculptor for pre-training coreset construction. Unlike methods dependent on additional training-time signals or limited solely to topological statistics, GraphSculptor provides a label-free solution that constructs coresets via two complementary perspectives: intrinsic structure and contextual semantics. Concretely, structural diversity is quantified using intrinsic graph statistics, yielding a structural feature vector for each graph, while semantic diversity is captured by utilizing a pre-trained language model to encode descriptions generated via graph-to-text. GraphSculptor integrates these signals into a unified metric space and performs cluster-aware selection to preserve joint structural-semantic diversity. We further derive a theoretical bound on the loss gap between coreset and full-data pre-training, offering theoretical motivation for our selection formulation. Extensive experiments demonstrate that GraphSculptor effectively sculpts the dataset: a 10% coreset achieves 99.6% of full-data performance while reducing pre-training time by nearly 90%, offering a scalable solution for data-efficient graph pre-training.

This paper investigates coordinated downlink transmission in a multi-base station (multi-BS) multi-reconfigurable intelligent surface (multi-RIS)-assisted pinching-antenna (PA) system, where each user equipment (UE) is associated with a single BS and each BS is equipped with movable PAs deployed on parallel waveguides. We formulate sum rate (SR) and energy efficiency (EE) maximization problems by jointly optimizing PA placement, RIS phase shifts, transmit beamforming, and BS-UE association under constraints of inter-PA spacing, power budget, and unit-modulus phase shift. To address the resulting highly coupled mixed-variable problem, we propose a three-stage graph neural network (GNN) that integrates heterogeneous and homogeneous graph representations and is trained end-to-end in an unsupervised manner. Extensive numerical results demonstrate that the proposed three-stage GNN consistently outperforms representative system and learning baselines, generalizes well to unseen numbers of UEs, RISs, and BSs, and maintains millisecond-level inference time. Besides, the results validate the effectiveness of the proposed design from both system and architectural perspectives. Moreover, PAs are shown to enhance SR and EE, and the performance gain is enlarged with increasing number of PAs.

Laser absorption spectroscopy (LAS) is a well-established technique for non-intrusive measurement of gas species in combustion and atmospheric environments, but conventional methods struggle with multi-species mixtures under dynamic or interference-laden conditions. Overlapping spectral features, noise, and incomplete reference data limit reliability when unknown or weakly absorbing species are present. This dissertation develops diagnostics combining LAS with machine learning (ML) to address these limitations. Deep denoising autoencoders (DDAEs) are applied to shock-tube measurements during high-speed hydrocarbon pyrolysis, improving signal fidelity and detection limits for trace species. A structured unsupervised framework, HT-SIMNet, then mitigates interference from unknown species without full calibration data, using spectral augmentation and a Noise2Noise-inspired scheme to isolate species in reactive systems. Where reference spectra are unavailable, UnblindMix, an autoencoder-based blind source separation method, reconstructs concentrations and spectral signatures directly from mixture data, validated on mixtures of up to eight components. To recover weakly absorbing species masked by broader absorbers, a feature-engineering method based on first derivatives and convolutions selectively highlights minor species. Finally, VOC-certifire combines randomized smoothing with Voigt-based spectral perturbation to provide certifiable classification of volatile organic compounds under varying conditions. All techniques are experimentally validated and benchmarked. The integration of spectroscopic hardware with ML offers a path toward real-time, interference-resilient, reference-free gas detection for combustion science, environmental monitoring, and industrial safety.

Standard Retrieval-Augmented Generation (RAG) systems predominantly rely on semantic relevance as a proxy for utility. However, this assumption collapses in realistic decision-making scenarios where user queries are laden with cognitive biases, such as false premises or confirmation bias. In such cases, maximizing relevance paradoxically promotes the retrieval of sycophantic evidence that reinforces hallucinations, a critical failure we term the ``Relevance-Robustness Gap''. To bridge this gap, we propose CoRM-RAG (Counterfactual Risk Minimization for RAG), a framework that aligns retrieval with decision safety rather than mere similarity. Grounded in causal intervention, we introduce a Cognitive Perturbation Protocol to simulate user biases during training, which is then distilled into a lightweight Evidence Critic. This scoring module learns to identify documents that possess sufficient evidential strength to steer the model toward correctness despite adversarial query perturbations. Extensive experiments on decision-making benchmarks demonstrate that CoRM-RAG significantly outperforms strong dense retrievers and LLM-based rerankers in adversarial settings, while enabling effective risk-aware abstention through reliable robustness scoring. Our code is available at https://github.com/PeiYangLiu/CoRM-RAG.git.

Geometric Algebra (GA) presents challenges to learners due to its highly abstract mathematical structure and complex operational rules, as translating algebraic manipulations into concrete geometric interpretations is a non-intuitive process when developing related code. Currently, some existing GA software packages rely on manually written scripts for code generation and visualization, but their high learning curve hinders widespread adoption. Meanwhile, methods based on Large Language Models (LLMs) often produce logical errors when generating specific GA scripts, such as GAALOPScript, resulting in generally low accuracy. To address these issues, this study proposes GA-VisAgent -- a multi-agent interactive learning application for GA code generation and visualization -- building upon a Geometric algebra large language model (GAGPT). Integrating task planning mechanisms with ReAct reasoning strategies, GA-VisAgent can decompose complex operations into five standardized subtasks, including core operations like geometric products, rotations, and reflections. It supports natural language and mathematical formulas as input to automatically generate executable code, accompanied by interactive visualizations to aid user comprehension. Experimental results show that GA-VisAgent achieved a 90% code generation success rate across 40 typical Conformal GA tasks, representing a 70% improvement over GPT-4o. This application introduces an extensible new paradigm for teaching GA and developing visualization tools for related mathematical concepts. The online service for this project will be available at http://gagis.cn/gacrac.

This paper uses game theory to argue that, contrary to the prevailing view, a moratorium on Artificial Superintelligence (ASI) can be in a state's self-interest. By formalizing trategic interactions between geopolitical superpowers, we model the trade-off between the benefits of technological supremacy and the catastrophic risks of uncontrolled ASI. The analysis reveals that as the perceived cost of loss of control increases sufficiently relative to other parameters, it becomes in each state's self-interest to impose a moratorium. We further provide empirical evidence suggesting that the global perception of ASI risk is rising, making a stable, rational moratorium increasingly plausible in the current geopolitical landscape.

In the pursuit of autonomous learning systems, the foundational assumption of stationarity, the premise that data distributions and model behaviors remain constant, is fundamentally untenable. Historically, the research community has addressed non-stationary environments almost exclusively under the scope of concept drift, focusing primarily on temporal shifts in streams. However, as learning systems become increasingly autonomous and complex, merely adapting to temporal non-stationarity is no longer sufficient. Evolving beyond this traditional perspective, we propose a novel, three-dimensional taxonomy that systematizes the field based on the operational state of the system. First, time stream drift distinguishes between stochastic arbitrary patterns and structural rhythmic dynamics. Second, data stream drift disentangles shifts in feature representations, identified as representation drift, from changes in underlying semantics, recognized as semantic drift. Third, model stream drift characterizes the internal endogenous divergence of learning systems through the lenses of sequential plasticity, decentralized heterogeneity, and policy instability. Based on this framework, we systematically review 193 representative studies and identify key open challenges. By bridging the fragmented paradigms of drift adaptation, continual learning, and temporal generalization, this survey outlines a roadmap for building self-evolving intelligent systems capable of learning autonomously through continuous change.

Foundation model-driven agents often struggle with long-horizon planning due to the transient nature of purely prompting-based reasoning. While existing skill induction methods mitigate this by distilling experience into state-blind parameterized scripts, they fail to capture the conditional logic required for robust execution in dynamic environments. In this paper, we propose Neuro-Symbolic Skill Induction (NSI), a framework that lifts interaction traces into modular, \textit{logic-grounded} programs. By synthesizing explicit control flows and dynamic variable binding, NSI empowers agents to discover \textit{when} and \textit{why} to act. This paradigm enables the efficient generalization, allowing agents to induce skills from few-shot examples and flexibly adapt to unseen goals. Experiments on a series of agentic tasks demonstrate that NSI consistently outperforms state-of-the-art baselines, empowering agents to self-evolve into architects of logic-grounded skills.

The growing spread of misinformation in digital media highlights the need for reliable fake news detection systems, yet progress in under-resourced languages such as Bangla is limited by small and imbalanced datasets. This study investigates whether Large Language Model (LLM) based augmentation can effectively address this limitation and improve Bangla fake news classification. Existing datasets remain valuable but highly imbalanced, limiting model performance, and LLM based augmentation for Bangla has been scarcely explored. To fill this gap, we propose a systematic augmentation framework that generates synthetic Bangla news articles using the instruction tuned Gemma 3 27B IT model, supported by semantic filtering and controlled subsampling to preserve label consistency and diversity. We compare zero shot and few shot prompting, evaluate multiple augmentation rates, and examine random versus similarity-based selection strategies. Our experiments show that augmenting only the minority class with a high augmentation rate and random subsampling yields the strongest gains, raising the Fake News F1 score from 0.85 to 0.88. To support reproducibility and further research in this low-resource domain, we publicly release 4,545 synthetically generated Bangla fake news samples along with our full implementation. These findings demonstrate that well-designed LLM-driven augmentation can significantly improve fake news detection in low resource settings and provide a practical foundation for advancing multilingual misinformation research.

Spiking Neural Networks (SNNs) are widely regarded as an energy-efficient paradigm for modeling and processing temporal and event-driven information. Incorporating delays in SNNs has been proven to be an effective mechanism for improving spike alignment in event-driven tasks. However, existing delay learning approaches predominantly assign static delays to individual synapses, resulting in a large number of delay parameters and limited adaptability to input-dependent activity dynamics. To this end, we propose a Congestion-Aware Dynamic Axonal Delay mechanism, decomposing the delay into a channel-wise static base delay for temporal structuring and a global, activity-conditioned shift that dynamically regulates the state update rate under varying spike intensities. The delay parameters are learned using differentiable linear interpolation and discretized at inference time, preserving the benefits of our dynamic delay while incurring only minimal additional cost. Experiments on speech benchmarks, including the Spiking Heidelberg Dataset, Spiking Speech Commands, and Google Speech Commands, demonstrate that introducing congestion-aware delays into synaptic signal transmission effectively improves accuracy on temporal tasks, notably achieving 93.75\% accuracy on SHD, 80.49\% accuracy on SSC, and 95.53\% on GSC-35, while reducing the parameter count by approximately 50\% compared to state-of-the-art delay-based methods with the same architecture.

In deep networks with small initialization, training exhibits long plateaus separated by sharp feature-acquisition transitions. Whereas shallow nonlinear networks and deep linear networks are well studied, extending these analyses to deep nonlinear networks remains challenging. We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss and use this to classify activation functions into four universality classes. On the permutation-symmetric submanifold, the identity combines with an approximate balance law to reduce the full matrix flow to a scalar ODE, giving a critical-depth escape time law $τ_\star = Θ(\varepsilon^{-(r-2)})$ governed by the number $r$ of layers at the bottleneck scale rather than the total depth $L$. We find that this same $r-2$ exponent is recovered under He-normal initialization with $r$ bottleneck layers rescaled by $\varepsilon$, where the symmetry manifold is preserved by the flow but not attracting. We find close agreement between our theory and numerical simulations.

Iterative Retrieval-Augmented Generation (iRAG) has emerged as a powerful paradigm for answering complex multi-hop questions by progressively retrieving and reasoning over external documents. However, current systems predominantly operate on parsed text, which creates two critical bottlenecks: (1) \textit{Coarse-grained attribution}, where users are burdened with manually locating evidence within lengthy documents based on vague text-level citations; and (2) \textit{Visual semantic loss}, where the conversion of visually rich documents (e.g., slides, PDFs with charts) into text discards spatial logic and layout cues essential for reasoning. To bridge this gap, we present \textbf{Chain of Evidence (CoE)}, a retriever-agnostic visual attribution framework that leverages Vision-Language Models to reason directly over screenshots of retrieved document candidates. CoE eliminates format-specific parsing and outputs precise bounding boxes, visualizing the complete reasoning chain within the retrieved candidate set. We evaluate CoE on two distinct benchmarks: \textbf{Wiki-CoE}, a large-scale dataset of structured web pages derived from 2WikiMultiHopQA, and \textbf{SlideVQA}, a challenging dataset of presentation slides featuring complex diagrams and free-form layouts. Experiments demonstrate that fine-tuned Qwen3-VL-8B-Instruct achieves robust performance, significantly outperforming text-based baselines in scenarios requiring visual layout understanding, while establishing a retriever-agnostic solution for pixel-level interpretable iRAG. Our code is available at https://github.com/PeiYangLiu/CoE.git.

Plants, crops and their yields are essential to our very existence, but diseases and pests cause large losses every year. As such it is vital to ensure that diseases can be spotted early and treated accordingly and stopping the spread while still possible. Manual and traditional methods require personal to walk through the field and check for symptoms 'by hand'. This is very laborious and very time consuming, so ML methods have been applied as a result and they have garnered promising results. CNN models are especially efficient as they can automatically extract features from images without any manual feature construction before then feeding the features to a classifier. Datasets are largely influential to the final performance of the model. Despite the importance that datasets pose to the field, there still seems to be somewhat of a discrepancy between what is publicly available for use and what would be required to sufficiently train fully capable models. To overcome these shortcomings, as part of this thesis open datasets for the field of plant leaf disease classification have been identified as well as models that can be trained on them and extensive benchmarks have been carried out to identify their suitability. Then a new dataset was constructed based on those findings as well as on the findings of a augmentation applicability study, which will be used to train a new Base Model based on the DenseNet201 architecture, which managed to outperform the baseline model on said new dataset as well as outperforming it on plant leaf disease classification domain specific Transfer-Learning experiments on another new dataset. This new model manages to train models through Transfer-Learning (TL) faster, more robust, more stable, and with less data than general model would, overcoming a large number of issues that the field still suffers from.

In MRI, variations in scan parameters, sequence, or hardware can lead to discrepancies in image appearance, even for the same subject. These inconsistencies, known as domain shifts, can hinder image analysis and degrade the performance of deep learning models trained on data from specific target domains. MRI image harmonization aims to address these issues by aligning source domain images to the target domain images while preserving biological information such as anatomical structures. However, most existing harmonization approaches require access to both source and target domain data in training or test time. This dependence induces data sharing between institutions, raising concerns about patient privacy and substantially limiting the harmonization approaches that can be practically deployed in clinical settings. To overcome these limitations, we introduce TgtFreeHarmony, the harmonization framework tailored for target-free scenarios, eliminating the need for target domain data and any data sharing, enabling privacy-preserving harmonization directly within the source institution. Our approach estimates the target domain style by searching the manifold of MRI domain style constructed via a disentanglement-based generator using Bayesian optimization guided by the performance of a downstream task model, which is trained on target domain data. We evaluated our method on the brain tissue segmentation task across multiple institutes and demonstrated that it effectively harmonizes source images into target images, leading to improved downstream task performance. By enabling harmonization without any access to target-domain data, TgtFreeHarmony establishes a new direction of harmonization preserving data privacy that can be realistically deployed within clinical environments.

This position paper argues that LLM inference serving has outgrown generic heuristics and now demands mathematical optimization and algorithmic foundations. Despite rapid advances in serving systems such as vLLM and SGLang, their algorithmic cores remain largely unchanged from classical distributed computing: request routing uses join-shortest-queue or round-robin, scheduling defaults to FIFO, and KV cache eviction follows LRU. These general-purpose policies ignore the distinctive structure of LLM inference--dynamically growing KV cache memory, prefill-decode phase asymmetry, unknown output lengths, and continuous batching constraints. We contend that the field must develop mathematical models capturing these characteristics, enabling the design of algorithms with provable performance guarantees across diverse workloads, rather than heuristics that may succeed in some scenarios but fail unpredictably in others. Emerging work at the intersection of operations research and ML systems demonstrates that principled methods can match or exceed heuristic performance while providing theoretical guarantees. We call on the community to recognize algorithmic design for LLM serving as a research frontier.

In this work, we present Valley3, an omni multimodal large language model (MLLM) developed for diverse global e-commerce tasks, with unified understanding and reasoning capabilities across text, images, video, and audio. A key feature of Valley3 is its native multilingual audio capability for e-commerce, developed by extending vision-language models to better support crucial audio-visual tasks, particularly in short-video scenarios. To achieve this, we carefully design a four-stage omni e-commerce continued pre-training pipeline, through which Valley3 progressively acquires audio understanding, cross-modal instruction-following, e-commerce domain knowledge, and long-context reasoning capabilities, ultimately evolving into an omni model for diverse e-commerce scenarios. Then, we further improve Valley3 through post-training to encourage long-chain reasoning with controllable reasoning modes, enabling one non-thinking mode and three distinct levels of thinking, thereby balancing inference efficiency in simple scenarios with deep reasoning for complex applications. Moreover, we equip Valley3 with agentic search capabilities to proactively invoke search tools and acquire task-relevant information for e-commerce deep research tasks. To comprehensively assess the capabilities of Valley3, we construct an omni e-commerce benchmark spanning 6 tasks. Experimental results show that Valley3 consistently outperforms strong baselines on our in-house and open-source e-commerce benchmarks, while remaining competitive on general-domain benchmarks.

Recently, Convolutional Neural Network (CNN) or Transformer architecture based models have been proposed to overcome the limitations of Recurrent Neural Network (RNN) based models in spatiotemporal prediction. These models prevent the inefficiency of parallelization limitation due to the sequential properties and stacked error due to the recursive method, and show high performance. Novertheless, there are still some challengies. First, CNN based models have difficulty considering global information due to the local properties of the kernel, and their performance is limited. In addition, information is mixed because the time axis is combined with the channel axis of the image for processing. Models based on Transformer architecture have high complexity due to the self-attention calcuation and take a long training time. In this paper, we propose a new structure model called CNN-based Multi-In-Multi-Out model for Efficient Spatiotemporal Prediction (MIMO-ESP) to overcome these limitations. MIMO-ESP considers global information and significantly improves complexity by configuring a Transformer architecture based on CNN. In addition, it treats the time axis as an independent axis without combining it, and effectively considers spatiotemporal information together by applying dilation. This structure makes MIMO-ESP efficient and high performance. Extensive experiment results on three promising benchmark datasets which including video, traffic, and precipitation prediction tasks demonstrate that the usefulness of MIMO-ESP due to the achieved competitive efficiency while outperforming existing models. Furthermore, the ablation study results demonstrate the usefulness of the components of MIMO-ESP, emphasizing the potential of the proposed approaches.

Spatio-temporal signals arising from event-driven biological processes, such as surface electromyography (sEMG), exhibit asynchronous and highly structured activation patterns that are challenging to model using conventional discrete or purely real-valued representations. In this work, we propose a continuous temporal modeling framework based on interference-based wave representations. The approach maps event-like input signals into a complex-valued latent wave field, where temporal structure is encoded through phase modulation and interactions between latent components. By projecting the resulting wave field onto an energy domain, the model induces structured activation patterns that capture both temporal localization and relational dependencies within finite observation windows, without relying on explicit recurrence or causal state propagation. The proposed formulation is particularly suited for event-driven biosignals, where continuous representations enable efficient gradient-based optimization and robust feature extraction. In particular, the method is designed to support learning from sEMG data for downstream control tasks in biomechanical systems, such as prosthetic devices and exoskeletons. Experimental results demonstrate that the proposed interference-based wave model provides improved representation quality compared to purely real-valued representations, while maintaining computational efficiency suitable for practical deployment.

Traditional approaches to test case generation often involve manual effort and incur significant computational overhead. Additionally, these approaches are not scalable, and hence, unsuitable for complex software systems. Recently, Large Language Models (LLMs) have been applied to software testing. However, single-shot prompt engineering-based approaches tend to hallucinate and generate redundant test cases, resulting in fewer branches. To handle the above-mentioned limitations, in this paper, we propose FeedbackLLM, a novel automated language-agnostic test case generation framework based on tightly coupled two-stage approach. In the first stage, FeedbackLLM extracts the input constraints by parsing source code and generates the possible test cases. The quality of the test cases is evaluated in the second stage by the following two specialized LLM feedback agents: (i) Line Feedback Agent: extracts the metadata related to missed line executions and (ii) Branch Feedback Agent: extracts the metadata of the unexecuted branch conditions. The above agents operate in a two-stage process, communicating in tandem, and this procedure is repeated for k-steps. Further, we also introduced a redundancy prevention cache to avoid duplicate API requests and avoid unnecessary execution cycles. The performance of the proposed architecture is evaluated on the standard benchmark programs related to C and Python programs. FeedbackLLM demonstrated more line and branch coverage than baseline tools while scaling linearly in execution time.

We study query time bounds for the fundamental problem of estimating the kernel mean $\frac1{|X|}\sum_{x\in X}\mathbf{k}(x,y)$ of a query $y$ in a finite dataset $X\subset\mathbb{R}^d$ up to a prescribed additive error $\varepsilon$. The best known bounds for the Gaussian kernel are $O(d/\varepsilon^2)$, $\widetilde O(d+1/\varepsilon^4)$, and $\widetilde O(d+Δ^2/\varepsilon^2)$, where $Δ$ is the diameter of a region containing the points. We prove the new bound $\tilde O(d+\varepsilonΔ^2+1/\varepsilon^3)$, which improves over the previous ones in regimes with small error $\varepsilon$ and intermediate diameter $Δ$. At the center of our proof is a new fast spherical embedding theorem in the sense introduced by Bartal, Recht and Schulman (2011), which limits the embedded data diameter while preserving local Euclidean distances and avoiding ``distance collapse'' at larger scales. This fast embedding theorem may be of independent interest.

Large-scale GPS trajectory data offer rich observations of human mobility, yet assigning trip purposes to detected stops remains challenging due to the absence of individual-level ground truth, spatial uncertainty from GPS noise and incomplete points of interest (POIs) coverage, and fundamental behavioral differences across trip purposes. We propose a weakly supervised framework integrating neighborhood-level POI semantic zones with distance-weighted spatial likelihoods, differentiated inference strategies for mandatory and non-mandatory activities, and a multi-phase Pareto optimization that jointly minimizes distributional divergence from household travel survey statistics and maximizes inference reliability without requiring annotated labels. Evaluated on over 81 million staypoints in Los Angeles, the framework reduces activity type frequency Jensen-Shannon distance (JSD) by 23%, start time JSD by 48%, and duration JSD by 12% respectively relative to a comparable baseline. The proposed approach provides a scalable and uncertainty-aware path from raw GPS trajectories to semantically annotated mobility data for travel demand modeling and transportation policy analysis.

A promising paradigm for adapting instruction-tuned language models is to learn task-specific updates on a pretrained base model and subsequently merge them into the instruction-tuned model. However, existing approaches typically treat the instruction-tuned model as a passive target that is only involved at the final merging stage, without guiding the training process. We propose GIFT (Guided Fine-Tuning and Transfer), a simple and efficient framework that incorporates guidance from the instruction model into task adaptation. GIFT fine-tunes a low-rank adapter on the pretrained base model using confidence signals derived from the instruction-tuned model. The learned adapter is then merged into the instruction-tuned model, yielding task-specialized models that preserve general instruction-following behavior. We evaluate GIFT on mathematical and knowledge-intensive benchmarks across multiple model families and scales. Results show that GIFT consistently outperforms direct fine-tuning and representative transfer-based baselines, while maintaining robust generalization and favorable test-time scaling behavior.

Training large language models (LLMs) is highly memory-intensive, as training must store not only weights and optimizer states but also intermediate activations for backpropagation. While existing memory-efficient methods largely focus on gradients and optimizer states, activation compression is less well established due to the lack of LLM-tailored theory and guarantees. In this work, we develop a theoretical framework showing that activation compression is safe for linear operators when activation compression is unbiased, but problematic for nonlinear ones. We further derive gradient variance bound and establish convergence guarantees for applying activation compression to all linear operators under the standard $L$-smoothness assumption, showing that it does not change the convergence rate. Guided by the theory, we propose an activation-gradient co-compression method that reuses low-rank activation factors to compress linear-layer gradients without extra computation or additional gradient error. We conduct extensive experiments on Qwen and LLaMA models using a pretraining benchmark and multiple fine-tuning benchmarks to validate our theory and demonstrate competitive performance of our method in both accuracy and compression efficiency. We provide our code in the supplementary material for reproducibility.

Earth Observation (EO) analysis is inherently interactive: resolving uncertainty often requires expanding the region of interest, retrieving historical observations, and switching across sensors such as optical and Synthetic Aperture Radar. However, most EO benchmarks collapse this process into fixed-input, single-turn tasks. To address this gap, we present EO-Gym, a controlled executable framework for multimodal, tool-using EO agents that formulates EO analysis as a Gymnasium-style local geospatial workspace backed by more than 660k multimodal files indexed by location, time, and sensor type, with 35 EO-specialized tools spanning six task families. Built on this environment, we construct EO-Gym-Data, a benchmark of 9,078 trajectories and 34,604 reasoning steps, and grounded in eight public EO datasets together with Landsat and Sentinel-2 imagery. Evaluating $10$ open and closed VLMs shows that strong general-purpose models still struggle with interactive EO reasoning, especially on temporal and cross-modal workflows. As a reference baseline, EO-Gym-4B, obtained by fine-tuning Qwen3-VL-4B-Instruct on EO-Gym-Data, improves overall Pass@3 from $0.49$ to $0.74$ under the main evaluation setting. O-Gym provides a reproducible environment for interactive EO agents, operationalizing EO as an evidence-gathering problem that requires planning across geospatial, temporal, and sensing modality.

Reinforcement learning (RL) post-training has enabled newer capabilities in models, such as agentic tool-use for search. However, these models struggle primarily due to limitations with sparse outcome-based rewards and a lack of training data that encapsulates questions of differing hardness, which results in models not performing deeper searches with tools to collect evidence for question-answering. To address these limitations, we introduce S^3-R1 (Synthetic data and stabilized Search R1), a framework that couples a data-centric approach with denser learning signals. We first develop a synthetic generation and curation pipeline that programmatically derives diverse, multi-hop questions from existing documents. This pipeline incorporates a retrieval-based verification step to specifically isolate questions of intermediate difficulty. We then pair this expanded training set with a reward structure that evaluates both intermediate search quality and the correctness of the final answer. This setup directly mitigates the credit assignment problems inherent to sparse rewards. Our evaluations show that S^3-R1 outperforms existing baselines by learning more effective search and synthesis strategies, yielding up to a 10% improvement in robust generalization on out-of-domain datasets.

Narrative archetypes (e.g., Hero's Journey, Three-act structure) provide universal story structures that resonate across cultures and media and are important for video game storytelling, yet existing LLM-based methods lack explicit use of these archetypes in procedurally generated games. We propose Forking Garden, a framework for narrative arc-conditioned gameplay planning that generates branching games from user-provided storylines. Our approach first generates a diverse pool of independent nodes, then assembles them into a dungeon graph via arc-guided constraint algorithms, where each node achieves multimodal alignment of gameplay elements. We develop an end-to-end interactive system that instantiates the framework.

Reinforcement learning (RL) is a fundamental framework for sequential decision-making, in which an agent learns an optimal policy through interactions with an unknown environment. In settings with function approximation, many existing RL algorithms achieve favorable sample complexity, but often rely on computationally intractable oracles. In this paper, we use supervised learning as a computational proxy to establish a clear hierarchy of commonly adopted RL oracles under low-rank Markov Decision Processes (MDPs). This hierarchy shows that policy evaluation is the most computationally efficient oracle, provided that supervised learning can be efficiently solved. Motivated by this observation, we propose a novel optimistic actor-critic algorithm that relies solely on the policy evaluation oracle. We prove that our algorithm outperforms the existing sample complexity guarantees for low-rank MDPs while avoiding computationally expensive planning or optimization oracles commonly assumed in prior works. We further extend our theoretical results to approximately low-rank MDPs and demonstrate that this setting captures a broad class of real-world environments. Finally, we validate our theoretical results with experiments on several standard Gym environments.

Self-supervised pretraining is promising for large-scale neuroimaging, yet the impact of region-aware masking and hybrid sequence modeling remains underexplored. In this work, we introduce Rhamba, a region-aware pretraining framework that integrates anatomically guided masking with hybrid Attention-Mamba architectures for resting state functional magnetic resonance imaging (fMRI) analysis. Models were pretrained on the ABIDE dataset using region-aligned patch embeddings and three masking strategies (Any, Majority, and Pure) with increasing spatial specificity. We evaluated four architectural variants: a Mamba only model, an Alternate architecture with interleaved Mamba and Attention blocks, and two hybrid encoder-decoder configurations (Attention-Mamba (AM) and Mamba-Attention (MA)). The pretrained models were fine-tuned on downstream classification tasks using the COBRE and ADHD-200 datasets for schizophrenia and attention-deficit/hyperactivity disorder discrimination. We employed Integrated Gradients, an explainable AI method, to identify the brain regions contributing to model predictions. Masking strategy strongly influenced reconstruction behavior, with reconstruction loss following a consistent ordering (Any > Majority > Pure). However, this trend did not directly translate into downstream performance, where differences were modest and dataset-dependent. The hybrid architecture with the MA configuration achieved the highest average AUROC across both datasets, and Rhamba outperformed state-of-the-art methods in comparative evaluation. Region-wise analysis showed that peak performance depends on the interaction between masking strategy and architecture rather than a single dominant configuration. Overall, Rhamba offers a flexible framework for balancing interpretability, scalability, and performance in large-scale fMRI representation learning.

Driven by the escalating global burden of mental health conditions, music-based interventions have attracted significant attention as a non-invasive, cost-effective modality for emotion regulation and psychological stress relief. However, current digital music services rely on static preferences and fail to adapt to users' instantaneous psychological states. Furthermore, directly mapping electroencephalography (EEG) to music generation remains challenging due to severe paired-data scarcity and a lack of interpretability. To address these limitations, we propose MindMelody, a fully functional, closed-loop real-time system for EEG-driven personalized music intervention. MindMelody introduces an emotion-mediated semantic bridge. Specifically, a hybrid Transformer-GNN first decodes real-time EEG signals into global Valence-Arousal states and local temporal affect trajectories. These states are then fed into a Retrieval-Augmented Generation (RAG)-equipped Large Language Model (LLM) to formulate structured intervention plans. Subsequently, a novel Hierarchical EEG Controller injects global affect prefixes and local temporal guidance into a pretrained music backbone, enabling fine-grained controllable audio synthesis. Crucially, the system incorporates a continuous feedback loop that updates generation parameters on the fly based on the user's evolving EEG dynamics. Extensive experiments show that MindMelody improves control adherence and emotional alignment, and receives higher perceived helpfulness in a short-term listening setting, suggesting its promise as an adaptive affect-aware music generation framework.

Recent progress in time-series forecasting has led to rapidly increasing architectural complexity, yet many reported State-of-the-Art gains are statistically fragile or misattributed. We argue that progress requires a shift from model selection to modular attribution, identifying which components truly drive performance. We propose CombinationTS, a self-contained probabilistic evaluation framework that decomposes forecasting models into orthogonal modules--Input Transformation, Embedding, Encoder, Decoder, and Output Transformation--and evaluates them under a shared evaluation condition space. By quantifying each component via marginalized performance ($μ$) and stability ($σ$), CombinationTS enables robust attribution beyond fragile point estimates. Through large-scale paired evaluation, we uncover the Identity Paradox: once the data view (Embedding) is well-designed, a parameter-free Identity Encoder often matches or outperforms complex backbones. We further show that explicit structural priors introduced via Input Transformations yield a more favorable performance-stability trade-off than increasing Encoder complexity, establishing a principled baseline for architectural necessity.

Cross-attention patterns in neural machine translation (NMT) are widely used to study how multilingual models align linguistic structure. We report a systematic artifact in cross-attention analysis of NLLB-200 (600M): non-content tokens - primarily end-of-sequence tokens, language tags, and punctuation - capture 83 percent to 91 percent of total cross-attention mass. We term these "attention sinks," extending findings from LLMs [Xiao et al., 2023] to NMT cross-attention and identifying a causal mechanism rooted in vocabulary design rather than position bias. This artifact causes raw metrics to underestimate content-level similarity by nearly half (36.7 percent raw vs. 70.7 percent filtered), rendering uncorrected analyses unreliable. To address this, we validate a content-only filtering methodology that removes non-content tokens and renormalizes the distribution. Applying this to 1,000 parallel sentences across African languages (Swahili, Kikuyu, Somali, Luo) and non-African benchmarks (German, Turkish, Chinese, Hindi), we confirm the artifact is universal and recover masked linguistic signals: a 16.9 percentage-point gap between teacher-forcing and generation modes, clear language-family clustering in attention entropy, and a hidden Somali paradox linking SOV word order to monotonic alignment. We release our filtering toolkit and corrected datasets to support reproducible interpretability research on multilingual NMT.

Events in spatiotemporal systems are ubiquitous, yet modeling their complex distributions remains challenging. Existing point process models often rely on strong structural assumptions and are typically limited to autoregressive, event-by-event prediction. As a result, they struggle to support broader inference tasks such as inverse inference, trajectory reconstruction, and recovery of missing event locations. We introduce Arbitrarily Conditioned Hierarchical Flows (ARCH), a hierarchical flow matching framework for spatiotemporal event modeling. ARCH is expressive enough to capture complex event distributions while enabling tractable and accurate computation of conditional intensities, which quantify instantaneous event risk. Built on a history-encoder-generative-decoder architecture, ARCH introduces a hybrid masking strategy for flexible conditioning on arbitrary observed events. This enables a unified treatment of forecasting, inverse inference, and partial trajectory recovery within a single framework. Experiments on synthetic and real-world datasets show that ARCH consistently outperforms existing baselines across both prediction and conditional inference tasks.

This paper argues that contemporary multilingual NLP has converged on a fragile and misleading paradigm of incidental multilingualism. Today's LLMs appear multilingual largely because they are trained on massive, uneven web corpora, not because multilingual or multicultural competence has been treated as a core design objective. We contend that this paradigm systematically produces unequal, brittle, and opaque behavior across languages, with severe consequences in real-world and agentic deployments where models must reason, plan, and act across multiple linguistic contexts. We report a focused empirical study of two practical questions: which languages models self-report as supported and which languages they actually respond in across multilingual prompts. We additionally demonstrate how even a simple language-change attack can surface these failures and expose hidden assumptions about language in LLM-based systems. To address this, we call for a shift toward multilingualism by design: a research agenda that treats equitable multilingual performance, cultural grounding, and cross-lingual behavioral understanding as first-class goals in all aspects of the model pipeline.

Signal Temporal Logic (STL) offers verifiable task specifications and is crucial for safety-critical control. Yet STL planning remains challenging: exact optimization-based methods are often too slow, and learning-based methods struggle to generalize across varying environments. We propose a zero-shot STL planning solver for variable-map environments that generates feasible trajectories without retraining. By integrating a map-conditioned Transformer architecture with a lightweight heuristic, our approach effectively handles complex disjunctive (OR) subformulas. Furthermore, we leverage Transitive Reinforcement Learning (TRL) to ensure consistent temporal grounding and logical coherence across decomposed sub-tasks. Experiments on dynamic semantic maps with diverse obstacle layouts demonstrate consistent gains, highlighting the framework's superior zero-shot generalization to changing environments and broad STL coverage.

While diffusion models enable new approaches for estimating Local Intrinsic Dimension (LID), existing methods fail in high-dimensional spaces where noise from vast normal directions overwhelms the tangent signal. We propose Local Hessian Spectral Dimension (LHSD), which resolves this by applying spectral filtering to the log-density Hessian, explicitly cutting off large eigenvalues associated with normal directions to count zero-curvature tangent directions. Implemented using Stochastic Lanczos Quadrature (SLQ), LHSD avoids full Hessian construction, achieving linear scalability with dimension $D$. Experiments on synthetic and real data confirm LHSD's superior robustness and its utility in detecting memorization in large-scale diffusion models.

This position paper argues that agentic AI systems should be designed and evaluated as \emph{marginal token allocation economies} rather than as text generators priced by the unit. We follow a single request -- a developer asking a coding agent to fix a failing test -- through four economic layers that today are designed in isolation: a router that decides which model answers, an agent that decides whether to plan, act, verify, or defer, a serving stack that decides how to produce each token, and a training pipeline that decides whether the trace is worth learning from. We show that all four layers are solving the \emph{same} first-order condition -- marginal benefit equals marginal cost plus latency cost plus risk cost -- with different index sets and different prices. The framing is deliberately minimal: we do not propose a complete theory of AI economics. But adopting marginal token allocation as the shared accounting object explains why systems that locally minimize tokens globally misallocate them, predicts a small set of recurring failure modes (over-routing, over-delegation, under-verification, serving congestion, stale rollouts, cache misuse), and points to a concrete research agenda in token-aware evaluation, autonomy pricing, congestion-priced serving, and risk-adjusted RL budgeting.

Signal Temporal Logic (STL) is a formal language for specifying real-time behaviors of cyber-physical systems (CPS). Automatically transforming natural language requirements into STL specifications has received growing attention. Recent efforts leveraging large language models (LLMs) have demonstrated impressive performance, but some natural language requirements in practice contain vague or ambiguous information, which remains challenging for LLMs to handle. To address these challenges, we propose ClarifySTL, an interactive LLM-agent framework that enhances STL transformation through requirements clarification. ClarifySTL first detects vague expressions that indicate underspecified information in a requirement. If any vagueness is detected, it generates targeted clarification queries to guide users in supplementing the requirement until all necessary details are provided. Subsequently, if ClarifySTL detects ambiguities, it formulates focused ambiguity clarification queries and updates the requirements based on user feedback until all ambiguities are resolved. Finally, the requirements with vagueness and ambiguity clarified are transformed into STL specifications using LLMs. This interactive framework ensures that the resulting STL formulas faithfully capture user intent while reducing the burden on the user. We evaluate ClarifySTL on the representative benchmarks DeepSTL and STL-DivEn, as well as our newly introduced AmbiEval benchmark, which is specifically designed to assess the performance of the agents in handling vagueness and ambiguity, including both detection and query generation. The experimental results show that ClarifySTL is effective.

Vision-language model based graphical user interface (GUI) agents have shown strong interaction capabilities. However, they often behave unfaithfully, relying on memorized shortcuts rather than grounding actions in displayed screen evidence or user instructions. To address this, we propose Faithful-Agent, a faithfulness-first framework that reformulates GUI interaction to prioritize evidence groundedness and internal consistency. Faithful-Agent employs a two-stage pipeline: (i) a faithfulness-oriented SFT stage to instill abstainment behaviors under evidence perturbations; (ii) an RFT stage that further amplifies faithfulness by introducing the guided advantage estimator (GuAE), an anchor-based and variance-adaptive advantage tempering mechanism built upon GRPO. GuAE prevents advantage collapse in low-variance rollout groups under sparse GUI rewards, and with a thought-action consistency reward, Faithful-Agent (Stage II) elevates the Trap SR from 13.88\% to 80.21\% relative to the baseline, while preserving robust general instruction-following performance.

Cross-Tokenizer Knowledge Distillation (CTKD) enables knowledge transfer between a large language model and a smaller student, even when they employ different tokenizers. While existing approaches mainly focus on token-level alignment strategies, which are often brittle and sensitive to discrepancies between tokenizers, we argue that the method of aggregating tokens into more robust representations before distillation is of equal importance. In this paper, we introduce \textbf{SRA} (\textbf{S}pan \textbf{R}epresentation \textbf{A}lignment for Large Language Model Distillation), a novel framework that reframes CTKD through the physical lens of Multi-Particle Dynamical Systems. SRA shifts the fundamental unit of alignment from tokens to robust, tokenizer-agnostic spans. We model each span as a cluster of particles and represent its state by its Center of Mass (CoM) - an attention-weighted average that captures rich semantic information. We leverage the concept of span centers of mass with attention-derived weighting to prioritize the most salient spans. In addition, we employ a geometric regularizer to preserve the structural integrity of the representation space and introduce aligned span logit distillation to enhance knowledge transfer across models. In challenging cross-architecture distillation experiments, SRA consistently and significantly outperforms state-of-the-art CTKD baselines, validating our physically-grounded approach.

Currently, process reward models (PRMs) have exhibited remarkable potential for test-time scaling. Since large language models (LLMs) regularly generate flawed intermediate reasoning steps when tackling a broad spectrum of reasoning and decision-making tasks, PRMs are required to possess capabilities for detecting process-level errors in real-world scenarios. However, existing benchmarks primarily focus on mathematical reasoning, thereby failing to comprehensively evaluate the error detection ability of PRMs across diverse reasoning scenarios. To mitigate this gap, we introduce GR-Ben, a process-level benchmark specifically designed for assessing PRM's performance across two primary reasoning domains (science and logic) and nine subdomains. We conduct extensive experiments on a diverse set of 22 models, encompassing both PRMs and LLMs, and derive two key findings: (1) In domains beyond mathematical reasoning, the error-detection ability of existing PRMs and LLMs is found to be markedly weaker by comparison.(2) In general, PRMs are less adept at identifying knowledge-based errors, whereas LLMs exhibit poorer performance in detecting computational errors.We hope GR-Ben can foster future researches on PRMs for general domains, thereby enhancing the reasoning capabilities of LLMs.

Transformer-based models have achieved remarkable success across a wide range of domains, yet our understanding of their training dynamics remains limited. In this work, we identify a recurrent focus-dilution cycle in attention learning and provide a rigorous explanation in a one-layer Transformer setting for Markovian data via gradient-flow analysis. Using stage-wise linearization around critical points, we show that a single focus-dilution cycle can be decomposed into a sequence of distinct stages. First, embedding and projection rapidly condense to a rank-one structure, while attention parameters remain effectively frozen. Then, the attention parameters begin to increase, inducing a frequency-driven focus toward high-frequency tokens. As attention continues to evolve, it generates next-order perturbations in embeddings, leading to a mass-redistribution mechanism that progressively dilutes this focus. Finally, small asymmetries among low-frequency tokens lift a degenerate critical point, opening new embedding directions and initiating the next cycle. Experiments on synthetic Markovian data as well as WikiText and TinyStories corroborate the predicted stages and cyclical dynamics.

Two recent approaches to computation in superposition reach different recursive capacity regimes: Hänni et al. certify $\tilde{O}(d^{3/2})$ computable features in width $d$ via an approximate-linear recursive template, while Adler and Shavit reach near-quadratic capacity (up to logarithmic factors) using thresholded Boolean recovery. The main contribution of this paper is conceptual: we argue these results are not contradictory because they maintain different interface invariants, and we formalize the distinction. As a tool, we record a rank-trace Welch-type lower bound for biorthogonal linear readouts: for $F \gg d$, the worst-case off-diagonal cross-talk of any unit-diagonal linear readout is $Ω(d^{-1/2})$, and the bound is tight on average for unit-norm tight frames. At quadratic feature load $F=d^2$, random-support threshold recovery succeeds for sparsities $s=O(d/\log d)$, while linear readouts still incur $Ω(s/d)$ average per-coordinate squared error on Bernoulli sparse states. Matching the Welch floor against the published tolerance of the Hänni correction layer explains the $d^{3/2}$ scale as a compatibility threshold for that template, not a universal upper bound. Robust nonlinear reset beyond the Hänni template is left open.

Clinical AI adoption is hindered by the black-box/grey-box nature of high-performing models, which lack the ontological grounding and narrative transparency required for professional-level explainability. We present NEURON, a neuro-symbolic system designed to enhance both predictive reliability and clinical interpretability. NEURON integrates SNOMED CT ontology-informed structural representations with machine learning models to bridge the gap between raw data and medical nomenclature. To facilitate human-aligned interaction, the system utilizes a Retrieval-Augmented Generation (RAG) grounded LLM layer to synthesize SHAP feature attributions and patient-specific clinical notes into coherent, natural-language explanations. Validated on the MIMIC-IV dataset for Acute Heart Failure mortality prediction, NEURON improved the AUC from 0.74-0.77 to 0.84-0.88 and significantly outperformed raw SHAP visualizations in human-aligned metrics (0.85 vs. 0.50). Our results demonstrate that NEURON offers a robust, scalable engineering solution for deploying trustworthy, human-centered connected health applications.

Scaling laws enable the optimal selection of data amount and language model size, yet the impact of the data unit, the token, on this relationship remains underexplored. In this work, we systematically investigate how the information granularity of tokens, controlled by the compression rate (i.e., average bytes of text per token), affects scaling trends. We train 988 latent tokenized models (BLT) ranging from 50M to 7B parameters that enable setting the desired compression rate. This flexibility allows us to study the role of compression rate well beyond 4.57 bytes per token obtained with a popular BPE tokenizer. Our experiments reveal that in compute-optimal configurations, model parameter counts scale proportionally to data size measured in bytes, not in tokens as commonly perceived (Kaplan et al., 2020; Hoffmann et al., 2022). Furthermore, we discover that the optimal compression rate differs from the one obtained with BPE and decreases with compute. These findings generalize to both latent and subword tokenization, as well as to languages other than English, guiding language model developers on tokenization scheme selection for maximal compute efficiency.

The rapid expansion of uplink-intensive applications necessitates video coding solutions that balance high Rate-Distortion (RD) efficiency with ultra-low latency. This paper presents a longitudinal performance analysis of NVIDIA hardware encoding (NVENC), spanning from Pascal to the emerging Blackwell generation. We specifically evaluate the operational viability of the new "Ultra High Quality" (UHQ) tuning mode against standard low-latency configurations. Our results demonstrate that while the Blackwell architecture breaks historical efficiency plateaus, achieving a 5.94% BD-Rate gain in standard modes and up to 22.79% in UHQ modes, these gains incur severe system-level penalties. We reveal that UHQ operates as a hybrid pipeline, offloading complexity to CUDA cores and enforcing aggressive temporal structures (up to 7 B-frames) that increase end-to-end latency by over 400% and GPU board power consumption by up to 40%. Consequently, while UHQ successfully bridges the quality gap with software encoders, its prohibitive serialization delay renders it unsuitable for interactive real-time communications, positioning it instead as a specialized solution for Video-on-Demand (VoD) transcoding.

We present a non-asymptotic theory of generalization in deep learning where the empirical neural tangent kernel partitions the output space. In directions corresponding to signal, error dissipates rapidly; in the vast orthogonal dimensions corresponding to noise, the kernel's near-zero eigenvalues trap residual error in a test-invisible reservoir. Within the signal channel, minibatch SGD ensures that coherent population signal accumulates via fast linear drift, while idiosyncratic memorization is suppressed into a slow, diffusive random walk. We prove generalization survives even when the kernel evolves $\mathcal{O}(1)$ in operator norm, the full feature-learning regime. This theory naturally explains disparate phenomena in deep learning theory, such as benign overfitting, double descent, implicit bias, and grokking. Lastly, we derive an exact population-risk objective from a single training run with no validation data, for any architecture, loss, or optimizer, and prove that it measures precisely the noise in the signal channel. This objective reduces in practice to an SNR preconditioner on top of Adam, adding one state vector at no extra cost; it accelerates grokking by $5 \times$, suppresses memorization in PINNs and implicit neural representations, and improves DPO fine-tuning under noisy preferences while staying $3 \times$ closer to the reference policy.

Despite recent progress, recovering parametric CAD construction sequences from geometric input, such as meshes or point clouds, is a key challenge for design and manufacturing, as existing CAD reconstruction and generation methods are largely restricted to difficult-to-edit formats like meshes or Breps or editable simple sketch-and-extrude pipelines and low-complexity datasets. We introduce CADFit, a hybrid optimization-based CAD reconstruction framework that recovers complex, editable CAD construction sequences from meshes by incrementally fitting and validating parametric operations using geometric feedback. Our approach is distinguished by formulating reconstruction as an IoU-driven optimization over structured CAD programs and supporting a rich set of operations, including extrusions, revolutions, fillets, and chamfers. Experiments on multiple CAD benchmarks show that CADFit outperforms state-of-the-art mesh-to-CAD methods in volumetric Intersection-over-Union and Chamfer Distance, while substantially reducing the Invalid Ratio of reconstructed CAD programs, particularly for complex designs. We further present a multimodal pipeline that enables end-to-end reconstruction of CAD construction sequences from images by combining image-based geometry reconstruction with CADFit. By enabling accurate reconstruction of higher-complexity CAD models, CADFit provides a practical foundation for generating richer datasets and advancing future learning-based approaches to CAD reverse engineering.

It is increasingly recognized that human annotators do not always agree, and such disagreement is inherent in many annotation tasks. However, not all instances in a given task elicit the same degree of opinion divergence. In this paper, we investigate annotation variation patterns in graded human ratings for inappropriate languages, including offensive language, hate speech, and toxic language perception. We examine whether the degree of annotation disagreement can be predicted from textual features. We further propose the Opposition Index, a metric that quantifies perspective opposition among annotators on a given item, and investigate the predictability of instances with potentially opposing human opinions. Our results show a moderate positive correlation between estimated and observed annotation variance. We find that two approaches achieve comparable performance in variance prediction: directly predicting the variance value and estimating it from predicted annotation distributions. Our results on opposition perspective prediction show that items with high opposition index values are more difficult to predict and are often underestimated by models.

Activation steering is a method for controlling Large Language Model (LLM) behavior by intervening in its internal representations to increase the alignment with a specific target feature direction. However, standard interventions, such as vector addition, often cause ``collateral damage", defined as unintended changes in the alignment of activations along other non-target feature directions. This damage occurs because standard methods implicitly assume the isotropy of non-target features. In this work, we provide a mathematical formalization of collateral damage and introduce a principled framework that models steering as a constrained optimization problem. Our method finds a new activation that minimizes the expected squared collateral change weighted by the empirical second-moment matrix of activations. This weighting encodes the nonuniform cost of the perturbation in different feature directions, in contrast to isotropic approaches that penalize changes uniformly in all feature directions. By accounting for the empirical second-moment of activations, our approach achieves more precise control while reducing the degradation of model performance on unrelated tasks.

This position paper argues that LLMs should not yet be credited with decision explanation. This matters because recent work increasingly treats accurate behavioral prediction, plausible rationales, and outcome-conditioned reasoning traces as evidence that LLMs explain why people decide as they do, risking a premature redefinition of what counts as explanatory progress in human decision modeling. We first distinguish three claims with different evidential burdens: decision prediction, rationale generation, and decision explanation. We then argue that the evidence most commonly offered for LLM-based decision accounts directly supports the first two claims, and sometimes explanatory hypothesis generation, but does not distinguish decision explanation from prediction-supportive rationalization. Next, we propose a bridge standard for decision-explanation credit: stronger claims should specify explanatory targets, discriminate against weaker rationalizer alternatives, use target-appropriate process- or intervention-sensitive validation, and bound their scope. We then situate this standard against competing views and related literatures, clarifying why it preserves the value of LLMs as predictors, narrators, and hypothesis generators while resisting premature explanatory credit. We conclude with a principle of credit calibration: LLMs should be credited for the strongest claim their evidence warrants, and no stronger; if adopted, this principle can help turn LLMs from persuasive narrators of decisions into more reliable instruments for discovering, testing, and communicating explanations of human behavior.

Shared embedding spaces are widely used for multimodal search and data curation. In practice, two problems often limit how well this works. First, embeddings can reflect modality more than meaning, so examples cluster by input type even when the underlying content matches. Second, the paired supervision used to train these spaces is often noisy. When we blend many heterogeneous, human-labeled datasets, these issues reinforce each other and degrade cross-modal retrieval. We present a framework that improves alignment by acting on both the training pairs and the embedding model. Symmetric Nucleus Subsampling (SNS) refines training pairs by trimming raw inputs and annotations to the portions that best support each other. Expert Embedding Engine (EEE) combines complementary embedding experts using a learned projection network, together with a bias-aware objective that reduces modality-driven separation in the embedding space. We demonstrate that this approach collapses the modality gap by over 90% on average vs base embedding experts and is a strong data curator, with datablends from our method outperforming stratified sampling and traditional curation baselines in downstream model performance.

Since 2022, AI-powered coding assistants have produced contradictory evidence: controlled studies report 20-56% productivity gains on well-scoped tasks, while the most rigorous RCT documents a 19% slowdown for experienced developers, and telemetry across 10,000+ developers shows 98% more pull requests but 91% longer review times with flat delivery metrics. This paper argues these findings constitute the Productivity-Reliability Paradox (PRP): a systematic phenomenon emerging from non-deterministic code generators and insufficient specification discipline. Through a multivocal literature review of 67 sources (2022-2026), this paper: (1) formally defines the PRP with three moderating variables (task abstraction, codebase maturity, developer experience) and two amplifying mechanisms (code review bottleneck, context window constraint); (2) proposes the AI-Augmented Methodology Taxonomy (AAMT), classifying six methodologies under three AI integration tiers; (3) introduces the Specification Governance Model (SGM), grounded in Transaction Cost Economics, with a practical governance decision guide; and (4) evaluates Spec Kit and TDAD as SGM instantiations via a four-month pilot study. Specification discipline, not model capability, is the binding constraint on AI-assisted software dependability.

Developing business-logic-rich microservices requires navigating complex trade-offs between data consistency and distributed coordination. Although patterns like Sagas and Transactional Causal Consistency (TCC) provide mechanisms to manage distributed state, validating their behavior before production is challenging. Current architectural simulators prioritize network metrics over domain semantics, whereas industry frameworks demand full-scale infrastructure deployments, preventing early architectural experimentation. To bridge this gap, we introduce a \textit{Domain-Driven Design} (DDD) microservice simulator that isolates core business logic from communication and transactional infrastructure. By modeling microservice systems around aggregates, the simulator allows developers to evaluate identical application code under varying consistency guarantees and network constraints. It features support for multiple transactional models (Sagas, TCC) and seamless transitions across diverse deployment topologies, ranging from centralized execution to fully distributed environments. We validate the simulator through the implementation and rigorous concurrency testing of a complex, multi-aggregate microservice system. Through empirical benchmarks, we quantify the performance, coordination overhead, and resilience of different transactional models across localized and distributed execution environments. The findings confirm that the simulator minimizes developer effort while providing a powerful, deterministic environment for the shift-left validation and optimization of business logic implementation in microservice architectures.

ARC-AGI-2 is a benchmark of human-intuitive visual puzzles that measures a machine's ability to generalize from limited examples, interpret symbolic meaning, and flexibly apply rules in varying contexts. In this paper, we discuss our approach to solving the ARC-AGI-2 puzzles with TinyLM, with additional fine-tuning at test time, including Test-Time-Training (TTT) and Products of Experts (POE). Our model achieves 96.1% accuracy on the training set and 21.7% accuracy on the evaluation set.

Does structure in representations imply structure in computation? We study how Llama-3.1-8B reasons over cyclic concepts (e.g., "what month is six months after August?"). Even though Llama-3.1-8B's representations for these concepts are circularly structured, we find that instead of directly computing modular addition in the period of the cyclic concept (e.g., 12 for months), the model re-uses a generic addition mechanism across tasks that operates independently of concept-specific geometry. First, it computes the sum of its two inputs using base-10 addition (six + August=14). Then, it maps this sum back to cyclic concept space (14->February). We show that Llama-3.1-8B uses task-agnostic Fourier features to compute these sums--in fact, these features have periods that respect standard base-10 addition, e.g., 2, 5, and 10, rather than the cyclic concept period (e.g., 12 for months). Furthermore, we identify a sparse set of 28 MLP neurons re-used across all tasks (approximately 0.2% of the MLP at layer 18) that can be partitioned into disjoint clusters, each computing the sum for a Fourier feature with a different period. Our work highlights how an interplay between causal abstraction and feature geometry can deepen our mechanistic understanding of LMs.

As large language models are increasingly deployed as interacting agents in high-stakes decisions, the AI safety community assumes that safety properties of individual models will compose into safe multi-agent behavior. This position paper argues that this assumption is fundamentally mistaken. In agentic AI, safety is determined by interaction topology, not model weights. When agents deliberate sequentially or aggregate via parallel voting with a judge, the structure of information flow and decision coupling dominates outcomes. Evidence across model families and scales reveals three persistent topology-driven pathologies: ordering instability, where system behavior depends primarily on agent sequence; information cascades, where early judgments propagate regardless of correctness; and functional collapse, where systems satisfy fairness metrics while abandoning meaningful risk discrimination. Contrary to intuition, scaling to more capable models strengthens these effects by increasing consensus formation and reducing the challenge of initial decisions. These failure modes are invisible to model-centric evaluation and alignment procedures. We argue that agentic AI must be treated as a dynamical system rather than a collection of aligned components. Interaction topology must become a primary target of safety evaluation and regulation, with systems required to demonstrate robustness across architectural variations before deployment.

Whole-slide images (WSIs) present a fundamental challenge for computational pathology due to their extreme resolution, multi-scale heterogeneity, and the requirement for clinically reliable interpretation. Although recent pathology foundation models have enabled fluent report generation, they often lack clinical grounding, failing to accurately represent key diagnostic concepts and relationships observed by pathologists. This limitation arises from the difficulty of integrating heterogeneous visual evidence spanning fine-grained cellular patterns, slide-level tissue architecture, and high-level diagnostic concepts, while maintaining interpretability and clinical coherence. Here we present SCOUT: Semantic Context-aware mOdality fUsion Transformer, a context-aware concept-grounded multimodal framework for pathology report generation that enables progressive conditioning of image representations by global slide information and explicit diagnostic concepts. The method integrates local histological patterns, whole-slide context, and expert-curated semantic descriptors within a unified learning paradigm, allowing visual features to be dynamically refined throughout the encoding process. By combining depth-aware contextual modulation with adaptive multimodal fusion during text generation, the framework produces clinically coherent reports while preserving complementarity across representational scales. Using CONCH1.5 features, we evaluate SCOUT against WSI-Caption, HistGen, and BiGen on TCGA-BRCA, MICCAI REG, and HistAI. SCOUT achieves the best BLEU-1 to BLEU-4 and METEOR scores on all datasets, plus the best ROUGE-L on TCGA-BRCA and MICCAI REG. On TCGA-BRCA, it reaches 0.436/0.303/0.202/0.156 BLEU-1/2/3/4 and 0.204 METEOR; on REG 2025, it achieves 0.865/0.834/0.805/0.780 and 0.568. These results support progressive contextual conditioning for grounded pathology report generation.

Large Language Model (LLM)-powered agents demonstrate strong capabilities in autonomous task execution, tool use, and multi-step reasoning. However, their increasing autonomy also introduces a new attack surface: adversarial interactions can manipulate agent behavior through direct prompt injection, indirect content attacks, and multi-turn escalation strategies. Existing defense strategies focus on prompt-level filtering and rule-based guardrails, which are often insufficient when risk emerges gradually across interaction sequences. In this work, we propose a complementary defense mechanism: a low-latency fraud detection layer for detecting adversarial interaction patterns in LLM-powered agents. Instead of determining whether a single prompt is malicious, our approach models risk over interaction trajectories using structured runtime features derived from prompt characteristics, session dynamics, tool usage, execution context, and fraud-inspired signals. The detection layer can be implemented using lightweight models leading to low-latency real-time deployments. To evaluate the framework, we construct a synthetic corpus of 12,000 multi-turn agent interactions generated from parameterized templates that simulate realistic agentic workflows. Using 42 structured features and an XGBoost classifier, our detector achieves over 9 times faster than LLM-based detectors. Through the experiment and ablation studies, our work suggests that interaction-level behavioral detection should become a core component of deployment-time defense for LLM-powered agents.

Metric differential privacy (mDP) strengthens local differential privacy (LDP) by scaling noise to semantic distance, but many machine learning (ML) systems are consumed under joint observation, where model-agnostic, per-record guarantees can miss leakage from evidence aggregation. We introduce metric-normalized posterior leakage (mPL), an attacker-aligned, distance-calibrated measure of posterior-odds shift induced by releases, and show that for single or independent releases, uniformly bounding mPL is equivalent to mDP. Under joint observation, however, satisfying mDP may still leave mPL high because learned aggregators compound evidence across correlated items. To make control practical, we formalize probabilistically bounded mPL (PBmPL), which limits how often mPL may exceed a target budget, and we operationalize it via Adaptive mPL (AmPL), a trust-and-verify framework that perturbs, audits with a learned attacker, and adapts parameters (with optional Bayesian remapping) to balance privacy and utility. In a word-embedding case study, neural adversaries violate mPL under joint consumption despite per-record mDP perturbations, whereas AmPL substantially lowers the frequency of such violations with low utility loss, indicating PBmPL as a practical, certifiable protection for joint-consumption settings.

Spectral graph sparsification is a classical tool for reducing graph complexity while preserving Laplacian quadratic forms. In graph neural networks (GNNs), sparsification is often used to accelerate computation while maintaining predictive performance. In this work, we study a complementary representation-level question: does sparsification preserve the geometry of learned embeddings? For polynomial-filter GNNs, we prove that any $ε$-spectral sparsifier induces $O(ε)$ perturbations in polynomial graph filters, multilayer hidden representations, and their Gram matrices. These guarantees imply stability of squared pairwise distances, class means, and covariance structure in embedding space. We further establish finite-time training stability: under smoothness and boundedness assumptions, gradient descent on dense and sparsified graphs produces weight trajectories whose separation grows at most proportionally to the sparsification distortion. Empirically, effective-resistance sparsification validates the predicted perturbation chain on synthetic graphs and preserves hidden representation geometry on real datasets. In our experiments, the gram matrix and training dynamics show low divergence even under substantial sparsification, consistent with the predicted stability under spectral sparsification. Hidden Gram preservation strongly predicts neighborhood preservation and class-centroid stability across FashionMNIST, Cora, and Paul15. Together, these results show that spectral sparsification preserves not only graph operators, but also the representation geometry that supports downstream use of GNN embeddings for interpretability.

The dominant noun-based modeling paradigm has fundamentally constrained AI development, precluding any adequate representation of the future as an open temporal dimension. This paper introduces a verb-based paradigm, together with precise definitions of \emph{timing computation} and \emph{possibility}, that enables AI to function as an effective instrument for realizing the grammar of our thought. Applied to longitudinal EHR data from 3,276 breast cancer patients, the framework empirically demonstrates: (1) automatic discovery of clinically significant patient trajectories, and (2) counterfactual timing deduction. Both results are purely data-driven, require no prior domain knowledge, and, to our knowledge, represent the first such demonstrations in the machine learning literature.

Large language model (LLM)-powered multi-agent systems (MAS) enable agents to communicate and share information, achieving strong performance on complex tasks. However, this communication also creates an attack surface where malicious agents can propagate misinformation and manipulate group decisions, undermining MAS safety. Existing embedding-based defenses aim to detect and prune suspicious agents, but their effectiveness depends on a clear separation between the text embeddings of malicious and benign messages. Attackers can circumvent such defenses by crafting messages whose embeddings lie close to benign ones. We analyze this failure mode theoretically and validate it empirically with three attacks, Slow Drift, Benign Wrapper, and Chaos Seeding. Our analysis further reveals a fundamental limitation of embedding-based defenses: because they rely solely on the text embeddings, they ignore token-level confidence signals such as logits, which can remain informative when embeddings are not distinguishable under attack. We propose using confidence scores to prune or down-weight messages during MAS communication. Experiments show improved robustness across models, datasets, and communication topologies. Moreover, we find that the effectiveness of confidence signals decays over communication rounds, highlighting the importance of early intervention. This insights can inform and inspire future work on MAS attacks and defenses.

In continual reinforcement learning (CRL), good performance requires never-ending learning, acting, and exploration in a big, partially observable world. Most CRL experiments have focused on loss of plasticity -- the inability to keep learning -- in one-off experiments where some unobservable non-stationarity is added to classic fully observable MDPs. Further, these experiments rarely consider the role of partial observability and the importance of CRL agents that use memory or recurrence. One potential reason for this focus on mitigating loss of plasticity without considering partial observability is that many partially-observable CRL environments are prohibitively expensive. In this paper, we introduce Forager, a light-weight partially-observable CRL environment with a constant memory footprint. We provide a set of experiments and sample tasks demonstrating that Forager is challenging for current CRL agents and yet also allows for in-depth study of those agents. We demonstrate that agents exhibit loss of plasticity, proposed mitigations can help, but that most useful is to leverage state construction. We conclude with a variant of Forager that generates an unending stream of new tasks to learn that clearly highlights the limitations of current CRL agents.

If a model has some behavioral tendency, such as sycophancy or misalignment, and it is trained on its own outputs, will the tendency be amplified in the next generation of models? We study this question by training a series of models where each model is finetuned on data generated by its predecessor, and the initial model is seeded with some persona or belief. We test three settings: supervised finetuning (SFT) on instruct models, synthetic document finetuning (SDF) on base models, and direct preference optimization (DPO). In the SFT and SDF settings, traits mostly decay or remain constant so that further finetuning cycles do nothing. In rare cases when amplification occurs, it generally comes at the cost of coherence. In the DPO setting, trait amplification can reliably occur when a model is continually trained with a preference for its own outputs, but vanishes when models are reinitialized at each cycle. Overall, our results suggest that amplification most likely comes from continual post-training, and limiting this stage may be an effective defense. For non-RL finetuning, trait amplification is rare and very sensitive to data quantity, making it significantly less likely to occur accidentally. Finally, the amplification-coherence tradeoff serves as a natural deterrent against trait amplification.

Partitioning transportation networks into balanced and spatially coherent traffic zones is a fundamental yet computationally challenging task in intelligent transportation systems. The resulting optimization problem exhibits dense interactions among decision variables and can be formulated as a Quadratic Unconstrained Binary Optimization (QUBO) model. While quantum optimization naturally aligns with such quadratic energy representations, current noisy intermediate-scale quantum hardware imposes limitations on problem size, connectivity, and circuit reliability. This paper proposes an impact-driven hybrid quantum--classical optimization framework for traffic zone partitioning that bridges transportation-scale optimization models and practical gate-based quantum processors. Instead of static geographic decomposition, the method estimates the energy impact of decision variables and selectively assigns quantum computation to influential subproblems while a classical coordination loop maintains global feasibility. The framework is implemented using the Iskay optimizer and evaluated on the IBM Quantum System One backend. Experiments compare direct quantum optimization, classical iterative SubQUBO refinement, and the proposed hybrid approach. Results show that impact-guided decomposition improves convergence behavior and produces more coherent spatial partitions relative to classical refinement, while remaining consistent with hardware constraints. Although the hybrid method does not outperform the best direct quantum solution, it demonstrates a practical pathway toward scalable hybrid optimization for transportation applications under current quantum hardware conditions.

Forecasting the wide variety of high-impact weather events experienced globally is a challenge for both Artificial Intelligence (AI) and Numerical Weather Prediction (NWP) models and it is critical that such models be properly verified before deployment. Although AI weather models are rapidly evolving, much of their evaluation is currently done either with a global-scale evaluation or by hand-picking a small number of case studies or a region. A widely-used open-source benchmark suite focusing on high-impact weather will help to drive the science forward for all scales of weather models, as it has for other AI fields. Here we introduce Extreme Weather Bench (EWB), a new community-driven benchmark suite that facilitates model validation and verification on a variety of high-impact hazards that matter to people around the globe. EWB provides a standard set of case studies (spanning across multiple spatial and temporal scales and different parts of the weather spectrum), observational data, impact-based metrics, and open-source code for users to evaluate their models. Verifying that a model works against a standard set of case studies, especially events that are high-impact for the general public, is a key piece of improving the trustworthiness of AI models. EWB will help to drive the science forward for all weather models, enabling true comparisons across models and evaluating models on specific high-impact phenomena through the use of case studies. EWB is a free open-source community-driven system and will continue to evolve to include additional phenomena, test cases and metrics in collaboration with the worldwide weather and forecast verification community.

Large language models (LLMs) can provide automated feedback in educational settings, but aligning an LLMs style with a specific instructors tone while maintaining diagnostic correctness remains challenging. We ask how can we update an LLM for automated feedback generation to align with a target instructors style without sacrificing core knowledge? We study how Reinforcement Learning from Human Feedback (RLHF) can adapt a transformer-based LLM to generate programming feedback that matches a professors grading voice. We introduce PERSA, an RLHF pipeline that combines supervised fine-tuning on professor demonstrations, reward modeling from pairwise preferences, and Proximal Policy Optimization (PPO), while deliberately constraining learning to style-bearing components. Motivated by analyses of transformer internals, PERSA applies parameter efficient fine-tuning. It updates only the top transformer blocks and their feed-forward projections, minimizing global parameter drift while increasing stylistic controllability. We evaluate our proposed approach on three code-feedback benchmarks (APPS, PyFiXV, and CodeReviewQA) using complementary metrics for style alignment and fidelity. Across both Llama-3 and Gemma-2 backbones, PERSA delivers the strongest professor-style transfer while retaining correctness, for example on APPS, it boosts Style Alignment Score (SAC) to 96.2% (from 34.8% for Base) with Correctness Accuracy (CA) up to 100% on Llama-3, and Gemma-2. Overall, PERSA offers a practical route to personalized educational feedback by aligning both what it says (content correctness) and, crucially, how it says it (instructor-like tone and structure).

Iterative ptychographic reconstruction algorithms are widely used for coherent diffractive imaging but can exhibit slow convergence under realistic experimental conditions. We propose a machine learning-augmented approach that accelerates iterative ptychographic reconstruction by introducing a learned fast-forward operator applied during reconstruction. Following an initial warm-up using standard iterations, the fast-forward operator advances the reconstruction toward a more converged state, after which conventional iterative updates are resumed. This strategy preserves the physical consistency and flexibility of established ptychographic solvers while reducing the number of iterations required for convergence. The model is trained on diverse ptychographic datasets and evaluated on experimental data acquired in a different year, demonstrating robustness and temporal generalization. Compared with conventional iterative solvers, the machine learning-augmented method achieves comparable reconstruction quality while converging faster in terms of Poisson negative log-likelihood, yielding over a two-fold reduction in wall-clock time. The approach has been integrated into an existing reconstruction pipeline and deployed in production at a synchrotron beamline, demonstrating practicality for real-time experimental operation.

The Zarankiewicz number $\textbf{Z}(m, n, s, t)$ is the maximum number of edges in a bipartite graph $G_{m, n}$ such that there is no complete $K_{s, t}$ bipartite subgraph. We determine for the first time the exact values of three Zarankiewicz numbers: $\textbf{Z}(11, 21, 3, 3)=116$, $\textbf{Z}(11, 22, 3, 3)=121$, and $\textbf{Z}(12, 22, 3, 3)=132$. We further establish lower bounds for 41 more Zarankiewicz numbers, including several that are within one edge of the best known upper bound, and we match the established value in four more closed cases. Our results are obtained using OpenEvolve, an open-source evolutionary algorithm based on Large Language Models (LLMs) that iteratively improves algorithms for generating mathematical constructions by optimizing a reward signal which we tailored for this specific problem. These findings provide new extremal graph constructions and demonstrate the potential of LLM-guided evolutionary search to contribute to mathematical research. In addition to presenting the resulting constructions, we report the generation algorithms produced, describe the relevant implementation details, and provide our computational costs. Our costs are remarkably low, at less than \$30 for each Zarankiewicz parameter combination, showing that LLM-guided evolutionary search can be an inexpensive, reproducible, and accessible tool for discovering new combinatorial constructions.

In this work, we introduce DUET (Dual-model Efficient Two-stage inference), a collaborative inference framework in which a capable model and a lightweight model work together to solve a task. Relying on a single large model to perform end-to-end reasoning and prediction often incurs substantial inference cost. In contrast, DUET decomposes inference into two stages: the capable model produces a reasoning signal, and the lightweight model interprets this signal to generate the final answer, allowing reasoning-intensive computation to be handled by the capable model while non-reasoning-intensive components are delegated to the lightweight model without sacrificing task performance. To achieve this objective, we propose a length-penalized joint training objective that encourages the capable model to transmit only the information that is sufficient for the lightweight model to solve the task. As a result, DUET maintains strong reasoning performance with substantially lower inference cost than end-to-end inference using a large model alone, saving up to 60% of the large model's output tokens on challenging reasoning benchmarks, including AIME and GPQA.

Graph neural tangent kernels give a principled infinite-width theory for graph neural networks, but inherit a basic limitation of graph models: they see only pairwise structure. Many relational systems contain higher-order interactions that are more naturally represented by simplicial complexes. We introduce the Topological Neural Tangent Kernel (TopoNTK), an infinite-width kernel for simplicial message passing on edge features. TopoNTK combines lower Hodge interactions, capturing graph-like coupling through shared vertices, with upper Hodge interactions, capturing coupling through filled simplices. This makes the kernel sensitive to topology invisible to graph kernels, allowing complexes with the same graph but different filled simplices to induce different kernels. Beyond expressivity, the Hodge structure gives the kernel an interpretable learning geometry. Edge signals decompose into gradient-like, harmonic, and local circulation components, and the spectrum of the TopoNTK determines how quickly each component is learned. This yields a topological form of spectral bias: components aligned with large-eigenvalue modes are learned quickly, while global harmonic modes, retained through the residual channel, often lie at smaller eigenvalues and are learned more slowly. We prove expressivity, Hodge-alignment, spectral learning, and stability properties, and validate them on synthetic simplicial tasks and DBLP higher-order link prediction. The results show that topology is not merely extra structure; it can provide coordinates that make relational learning more faithful, interpretable, and effective.

Neural networks transform data through learned representations whose geometry affects separation, contraction, and generalization. Recent work studies this geometry using discrete curvature on neighborhood graphs, suggesting Ricci-flow-like behavior across layers. We develop a smooth operator-theoretic alternative for feedforward representation snapshots. Each feature cloud induces a Gaussian-kernel diffusion Markov operator, and transport, spectral, label-boundary, and local-scale observables are derived from this single object via Bakry-Emery $Γ$-calculus. In a balanced Gaussian class-conditional snapshot model with shared covariance, the population operator has closed-form class affinities, leakage, and coarse spectra, all controlled by pairwise regularized Mahalanobis separations $c_\varepsilon^{(a,b)}$. We also prove that the resulting operator observables vary smoothly under feature perturbations, while hard neighborhood-graph diagnostics can change discontinuously. Synthetic experiments validate the closed-form Gaussian bridge, while learned MNIST experiments show that the same operator observables track training, width, and perturbation stability. Together, these results give a stable operator-geometric framework for analyzing feedforward representation geometry.

Speculative decoding accelerates autoregressive inference by drafting candidate tokens with a fast model and verifying them in parallel with the target. Self-speculative methods avoid the need for an external drafter but have been studied exclusively in homogeneous Transformer architectures. We introduce component-aware self-speculative decoding, the first method to exploit the internal architectural heterogeneity of hybrid language models, isolating the SSM/linear-attention subgraph as a zero-cost internal draft. We evaluate this on two architecturally distinct hybrid families: Falcon-H1 (parallel: Mamba-2 + attention per layer) and Qwen3.5 (sequential: interleaved linear and attention layers), with a pure Transformer control (Qwen2.5). Parallel hybrids achieve acceptance rates of alpha = 0.68 at draft length k=2 under greedy decoding, while sequential hybrids yield only alpha = 0.038 -- an 18x gap attributable to how each architecture integrates its components. The property is scale-invariant: Falcon-H1 at 3B reproduces the rates observed at 0.5B. We further show that perplexity degradation from a companion ablation study predicts speculative viability without running speculative decoding: a 3.15x ratio (Falcon) maps to alpha = 0.37 at k=4, while 81.96x (Qwen) maps to alpha = 0.019. For sequential hybrids, generic LayerSkip achieves 12x higher acceptance rates than the component-aware strategy. The composition pattern of hybrid models -- not merely the presence of alternative components -- determines whether component-level self-speculation is viable.

Understanding how developers interact with AI coding assistants requires more than chat logs or git histories in isolation; it requires reconstructing the full context: which prompt led to which edit, what the developer tried and discarded, and how their strategy evolved over time. We present RECAP (Replay and Examine Captured AI Programming), an open-source platform that (1) passively records AI chat sessions and fine-grained code edits inside VS Code without disrupting the developer's workflow, (2) merges them into a unified timeline for interactive session replay, and (3) exposes an extensible analysis layer, with example modules for behavioral classification and AI reliance measurement. Deployed in a university software engineering course, RECAP captured 2,034 prompts and 8,239 code edits from 41 students across a multi-week project. We demonstrate how the platform's linked data and replay capabilities enable analyses of developer-AI interaction patterns that no single data source could support.

Single-agent systems (SAS) have become the default pattern for LLM-driven scientific workflows, but routing planning, tool use, and synthesis through a single context window comes with a well-known cost: as tool specifications and observational traces accumulate, the effective context available for each decision shrinks, and end-to-end reliability suffers. We present a multi-agent system (MAS) prototype for hydrodynamics in which specialized agents are coordinated through a Layer Execution Graph (LEG). A planner agent constructs query-specific execution topologies from natural-language routing heuristics that capture domain knowledge without hard-coding it as rigid control logic; specialist agents operate under strict tool allowlists and occupy complementary data-class roles. Between layers, consolidator agents fuse parallel outputs into concise briefs, and a reporter agent synthesizes the final response, while the runtime logs provenance for every tool invocation to support auditability. All benchmarks, ablations, and stress tests use Claude Sonnet~4.6 as the backbone model for both specialist and general-purpose agents. Evaluated on 37 queries spanning six complexity categories, the prototype achieves 93.6% factual precision with a 100% pass rate. Accuracy remains above 90% across runs from single-threaded to five independent parallel tracks, and under simulated loss of individual data sources the system degrades gracefully, still returning substantive partial answers. Together, these results suggest that planner-guided, graph-structured multi-agent orchestration can meaningfully alleviate the context-saturation bottlenecks that constrain monolithic single-agent architectures.

This paper develops Virtual Speech Therapist (VST), an intelligent agent-based platform that streamlines stuttering assessment and delivers customized therapy planning through automated and adaptive AI-driven workflows. VST integrates state-of-the-art deep learning-based stuttering classification, and multi-agent large language model (LLM) reasoning to support evidence-based clinical decision-making. The VST begins with the acquisition and feature extraction of patient speech samples, followed by robust classification of stuttering types. Building on these outputs, VST initiates an agentic reasoning process in which specialized LLM agents autonomously generate, critique, and iteratively refine individualized therapy plans. A dedicated critic agent evaluates all generated therapy plans to ensure clinical safety, methodological soundness, and alignment with peer-reviewed evidence and established professional guidelines. The resulting output is a comprehensive, patient-specific therapy draft intended for clinician review. Incorporating clinician feedback, the system then produces a finalized therapy plan suitable for patient delivery, thereby maintaining a clinician-in-the-loop paradigm. Experimental evaluation by expert speech therapists confirms that VST consistently generates high-quality, evidence-based therapy recommendations. These findings demonstrate the system's potential to augment clinical workflows, reduce clinician burden, and improve therapeutic outcomes for individuals with speech impairments. An interactive user interface for the proposed system is available online at: https://vocametrix.com/ai/stuttering-therapy-planning-agent , facilitating real-time stuttering assessment and personalized therapy planning.

This work presents a knowledge-driven decision-support system that integrates structured defect knowledge with LLM-based reasoning to provide explainable defect diagnosis and mitigation guidance in manufacturing, using LPBF as a representative, safety-critical case study. The proposed ontology-integrated LLM-based decision support system for LPBF defect analysis and mitigation guidance is built on a knowledge base containing 27 known LPBF defect types organized into hierarchical categories and causal relationships. The developed system supports fuzzy natural language queries for systematic knowledge retrieval, literature-supported explanation of defects, and guidance on defect causes and mitigation strategies derived from encoded process knowledge. Furthermore, a multimodal image-assessment module based on foundation models enables descriptor-guided interpretation of representative microscopic defect images through semantic alignment scoring. The proposed framework was evaluated through qualitative comparisons with general-purpose vision-language models, an ablation study, and an inter-rater reliability analysis. Evaluation on the literature-derived dataset showed that the fully integrated configuration outperformed the other three evaluated system configurations, achieving a macro-average F1 score of 0.808. Additionally, inter-rater reliability analysis using Cohen's kappa indicated substantial agreement between the model outputs and the literature-derived reference labels. These findings suggest that ontology-guided knowledge representation can improve the consistency, interpretability, and practical usefulness of LLM-assisted LPBF defect analysis.

Adversarial examples in machine learning are typically generated using gradients, obtained either directly through access to the model or approximated via queries to it. In this paper, we propose a much simpler approach to craft adversarial examples, drawing inspiration from insights of explainable machine learning. In particular, we design \emph{adversarial image filters} that are based on classic edge detection algorithms but optimized to deceive learning models. The resulting untargeted attacks are transferable and require only a single pass over the input. Empirically, we find that 3x3 filters already enable success rates between 30% and 80% on different neural networks. Compared to related approaches using generative models for crafting adversarial examples, we reduce the number of parameters by five orders of magnitude, resulting in a very efficient attack. When investigating the parameters of the learned filters, we observe interesting properties such as a high transferability between models and structures common to classic image filters. Our results provide further insights into the vulnerability of neural networks and their fragility to malicious noise.

Recent advances in large language models (LLMs) have led to the development of AI-powered tutoring systems that provide interactive support via dialogue. To enable these tutoring systems to provide personalized support, it is essential to assess student performance at each turn, motivating knowledge tracing (KT) in dialogue settings. However, existing dialogue-based KT approaches often ignore question difficulty modeling and rely on opaque latent representations from LLMs, hindering accurate and interpretable prediction. In this work, we propose an interpretable difficulty-aware conversational KT framework built upon LLMs, which explicitly models students' abilities and the difficulty of tutor-posed tasks at each turn. The framework incorporates the original textual question and the next tutor-posed task to estimate the student's knowledge state and the difficulty of the upcoming turn. Furthermore, it integrates Item Response Theory to map LLM's outputs into student ability and question difficulty parameters, enabling interpretable prediction of student performance grounded in cognitive theories of learning. We evaluate the framework on two tutor-student dialogue datasets. Both quantitative and qualitative results show that our framework outperforms existing KT baselines, meanwhile generating interpretable outputs consistent with cognitive theory.

Reinforcement Learning (RL) has the potential to enable robots with fast, nonlinear, and unstable dynamics to reach the limits of their performance. However, most recent advances rely on carefully designed physics-based simulators and domain randomization to achieve successful sim-to-real transfer within reasonable wall-clock time. In this work, we bypass the need for such simulators and demonstrate that Infoprop Dyna, a state-of-the-art uncertainty-aware model-based reinforcement learning (MBRL) framework, can enable robots to learn directly from real-world interactions. Using Infoprop Dyna, the Mini Wheelbot, an underactuated unicycle robot, learns to race around a track within 11 minutes of real-world experience.

Evaluating DAG task schedulers for wireless edge computing requires jointly modeling compute placement and wireless interference, yet existing tools treat them in isolation. This gap leads to rank inversions: the scheduler that appears optimal under an interference-free model can be the worst choice under realistic wireless conditions. We present ncsim, a lightweight discrete-event simulator that bridges this gap by combining DAG workflow scheduling with physically-grounded IEEE 802.11 CSMA/CA interference modeling in a single Python package. A 108-run factorial experiment reveals rank inversions in 27.8% of scenarios, with the interference-free-optimal scheduler producing up to 2.7x worse makespan than a simple round-robin baseline; scaling to a 100-node random geometric graph raises the inversion rate to 50%. These rank inversions show that interference-free evaluation can select the wrong algorithm entirely, justifying the design and use of ncsim.

When a traffic signal controller adjusts green phases and a grid manager curtails power on the same corridor, each system may comply with its own obligations. The resident who suffers the combined effect has no single authority to hold accountable and, under the EU AI Act, limited means to obtain an explanation. Annex III, point 2 excludes safety-component AI in critical infrastructure from Article 86 explanation rights and Article 27 fundamental-rights impact assessment. Provider and deployer duties under Articles 9-15 still apply, and residual pathways under the GDPR, NIS2, and tortious liability offer partial coverage. The Act's principal resident-facing accountability instruments are nonetheless narrowed for the autonomous infrastructure systems most likely to interact across agencies. The paper traces this accountability deficit through four residual pathways (GDPR Article 22, GDPR transparency obligations, tortious liability, and NIS2) and shows that each is structurally bounded by individual-controller, individual-decision scope. As a governance response, it presents AgentGov-SC, a three-layer architecture (Agent, Orchestration, City) specifying 25 governance measures with bidirectional traceability to the EU AI Act, ISO/IEC 42001, and the NIST AI Risk Management Framework. Five conflict resolution rules and an autonomy-calibrated activation model complete the design. A scenario analysis traces governance activation through a multi-agent corridor cascade involving three documented UAE smart-city systems, with a contrasting single-system scenario confirming proportional activation. The paper contributes a regulatory gap analysis and governance architecture for an increasingly important class of urban AI deployment that existing frameworks treat as bounded and isolated.

The ensemble Gaussian mixture filter (EnGMF) is a powerful, convergent particle filter capable of medium-to-high dimensional non-linear filtering. The EnGMF relies on a resampling step that can generate physically unrealistic posterior samples, that would subsequently produce physically meaningless forecasts. This work introduces the discriminator-informed resampling procedure, that augments the posterior resampling step with a discriminator that accepts or rejects candidate particles based on their physical plausibility. In this work these discriminators are learned through a normalizing flow approach. Numerical experiments on both the Ikeda map and the Lorenz '63 system show that discriminator informed resampling procedure consistently reduces error relative to the standard EnGMF in low-ensemble regimes.

Modern high-performance computing and Internet-of-Things deployments increasingly generate large volumes of signal data that must be compressed efficiently on resource-constrained acquisition devices and decompressed at scale on centralized servers. Lossy compression is widely adopted to minimize storage and transmission costs on low-power hardware sensors, yet existing methods rarely optimize for both reconstruction quality and decompression throughput simultaneously, nor do they apply methods that generalize across signal domains. In this work, we introduce FPTC, a high-throughput asymmetric signal codec that pairs a lightweight sequential encoder with a massively parallel GPU decoder designed for server-side batch decompression. FPTC applies a windowed discrete cosine transform (DCT) to exploit frequency-domain sparsity, quantizes spectral coefficients with a hybrid three-zone mapping, and entropy codes the result using Huffman coding with a novel packing scheme. The pipeline used in FPTC is designed to be throughput oriented on the GPU, maximizing performance without sacrificing reconstruction quality. We evaluate FPTC on ten datasets spanning four signal domains: biomedical diagnostic, seismic reflections, power-grid production metrics, and meteorological recordings. Our results demonstrate that FPTC outperforms existing frameworks in compression ratio while maintaining competitive throughput, achieving multiplicative compression performance of 3.6x (power), 3.1x (meteorological), 1.5x (biomedical), and 1.2x (seismic) over existing frameworks.

We study networked binary classification on a directed acyclic graph (DAG) where each agent observes only a subset of the feature columns of a shared dataset. Agents act sequentially along the DAG: each receives prediction columns from its parents (if any), augments its local features with these columns, fits a logistic predictor by minimizing binary cross-entropy (BCE), and forwards its prediction column to its outgoing neighbors. We ask whether this sequential distributed training procedure achieves information aggregation, meaning that some agent attains small excess loss compared to the best logistic predictor trained with access to all feature columns. This question was studied for linear regression under squared loss by Kearns, Roth, and Ryu (SODA 2026). Extending their guarantees to classification is nontrivial because their analysis relies on quadratic structure that does not directly transfer to BCE with a logistic link. We analyze the resulting sequential logit-passing protocol and prove: (i) an excess loss upper bound of $O(M/\sqrt{D})$ on depth-$D$ paths under the condition that every $M$ contiguous subsequence of $M$ agents collectively observe all features, and (ii) a close lower bound showing instances with excess loss of at least $Ω(k/D)$ where $k$ is the dimension of the feature space. Together, these results identify network depth as a fundamental bottleneck for information aggregation in networked logistic regression.

Retrieval-augmented generation and tool-integrated LLM agents increasingly depend on external textual sources. This reliance broadens the available attack surface, allowing adversaries to insert malicious instructions that trigger unintended model behaviors. Current defensive measures often utilize LLM-based detectors to filter such content, but these approaches remain vulnerable to optimization-based attacks. Additionally, training-based methods frequently fail to generalize to novel data distributions. To resolve these issues, we introduce SONAR, a prompt sanitization framework that identifies and removes injected content using metrics from natural language inference. Specifically, SONAR constructs a sentence-level relational graph across the user query and external data. By using entailment and contradiction scores as edge weights, the system identifies sentences that deviate from the core task. It then employs connectivity-driven pruning to eliminate flagged injection seeds and their related neighbors while maintaining benign context. Rigorous evaluations across several models and datasets show that SONAR reduces the attack success rate to nearly zero, significantly outperforming nine established baseline defenses.

Brazil's Unified Health System (SUS) relies on official clinical guidelines that define diagnostic criteria, treatments, dosages, and monitoring procedures for over 200 million citizens. Yet current LLMs perform poorly on this guideline-specific knowledge, and no benchmark evaluates clinical recall grounded in Brazilian Portuguese protocols. We address this gap by adapting Qwen2.5-14B-Instruct to the Brazilian clinical domain. From 178 official guidelines (~5.4M tokens), we generate ~70M tokens of synthetic data in three formats -- rephrases, wiki-style articles, and question-answer pairs -- using four generator LLMs. We then apply continual pre-training followed by Group Relative Policy Optimization (GRPO). We introduce HealthBench-BR, with 1,780 balanced true/false clinical assertions, and PCDT-QA, with 890 open-ended clinical questions scored by an LLM judge. Our best model achieves 83.9% on HealthBench-BR and 85.4% on PCDT-QA, outperforming GPT-5.2, Claude Sonnet 4.6, Gemini 3.1 Pro, and Google AI Overview's web-grounded RAG despite having only 14B parameters. Ablations show that generator diversity and reinforcement learning are critical to these gains. We release all datasets, benchmarks, and model weights to support reproducible clinical NLP research for Brazilian Portuguese. Code, data, and model weights are available at https://github.com/hugoabonizio/clinical-protocols-br

Associative memory or content-addressable memory is an important component function in computer science and information processing, and at the same time a key concept in cognitive and computational brain science. Many different neural network architectures and learning rules have been proposed to model the brain's associative memory while investigating key component functions like figure-ground segmentation, perceptual reconstruction and rivalry. A less investigated but equally important capability of associative memory is prototype extraction where the training set comprises distorted prototype instances and the task is to recall the correct generating prototype given a new distorted instance. In this paper we benchmark associative memory function of seven different Hebbian learning rules employed in non-modular and modular recurrent networks with winner-take-all dynamics operating on moderately sparse binary patterns. We measure pattern storage and weight information capacity, prototype extraction capabilities, and sensitivity to correlations in data. The original additive Hebb rule comes out with worst capacity, covariance learning proves to be robust but with moderate capacity, and the Bayesian-Hebbian learning rules show highest capacity in almost all different conditions tested.

The paper studies the local geometry of embedding clouds induced by \emph{controlled local classes of semantically close sentences}. The central question is how controlled paraphrase-like semantic variation is organized in sentence embedding space and whether this local structure can be explicitly modeled by low-degree fitted carriers. We introduce a local geometric modeling scheme based on affine, quadratic, and cubic fitted models. We also use a surface-based latent probing procedure that constructs synthetic latent points in a reduced local PCA space with respect to the fitted carrier. The procedure is intended as an offline method for representation-space analysis, local manifold modeling, and geometry-aware latent probing. Generated latent points are evaluated using criteria that measure consistency with the fitted surface, preservation of neighborhood structure, agreement with the empirical distribution, stability of Hessian-based second-order shape descriptors, and stability of fitted-model coefficients. Experiments on controlled sets of semantically close sentences show that nonlinear local models describe embedding clouds more accurately than affine models. Surface-based generation provides strong fitted-geometry fidelity, including surface consistency, Hessian-based shape consistency, and coefficient consistency. Downstream experiments show that geometric validity of synthetic latent points does not automatically translate into improved classification performance. The results support explicit local geometric modeling of sentence embedding space and highlight the need to distinguish geometric validity from discriminative utility. As a resource contribution, we introduce \textbf{CoPaGE-300K}, a controlled template-based dataset of semantically close sentence variants with slot-level annotations and precomputed sentence embeddings.

We study the use of transformers to reconstruct the compositions of tensor products of two-dimensional rational conformal field theories (RCFTs) based on their low-energy spectra. The task is challenging due to its combinatorial nature. The constituent theories are characterized by their central charges and affine Lie algebra labels. We achieve 98% accuracy in recovering the constituents of tensor products theories constructed from Wess-Zumino-Witten models. We further demonstrate that our method generalizes to CFTs with larger central charge and unseen classes of RCFTs by adding a small number of out-of-domain examples. Our results show that transformers are effective at this task and point towards a new tool for bulk reconstruction in AdS/CFT.

Symbolic regression discovers explicit, interpretable equations without assuming a functional form in advance. A Bayesian approach strengthens this through probability distributions over candidate expressions, thus quantifying uncertainty in the presence of noisy and limited data. Deep Symbolic Regression (DSR) uses a neural network to generate symbolic expressions, but it is designed to identify a single best-fitting expression rather than infer a posterior distribution over models. We introduce Deep Variational Inference Symbolic Regression (DVISR), a variational Bayesian extension of DSR. DVISR replaces the original reward with the integrand of the evidence lower bound. It also extends the network architecture to output distributions over constants within expressions, enabling posterior inference over both expression trees and their associated constants. We show that DVISR can recover the true posterior in simple settings, both with and without constant tokens, and we examine how its performance changes as the size of the expression space increases. These results position DVISR as a step toward scalable Bayesian symbolic regression with uncertainty over full symbolic models.

R2 score is the standard metric for evaluating regression tasks, offering a normalized magnitude-agnostic measure of accuracy that captures variance. However, R2 has three key limitations: it is limited to at most two dimensional inputs, it reduces the score to a single scalar that hides rich patterns of prediction accuracy, and it is sensitive to low-variance noise channels which can yield large, uninterpretable negative values. We introduce the Dimensional R2 score (Dim-R2), a simple extension of R2 that accepts data of arbitrary dimensionality, provides a multidimensional view of accuracy, and reduces sensitivity to noise. We demonstrate its advantages on both synthetic sinusoidal data and three multidimensional regression datasets. Dim-R2 offers an interpretable and flexible metric that highlights patterns in regression accuracy, guiding regression modeling.

The goal of differentially private text obfuscation is to obfuscate, or "perturb", input texts with Differential Privacy (DP) guarantees, such that the private output texts are quantifiably indistinguishable from the originals. While perturbation at the word level is intuitive, meaningful text privatization happens on complete documents. Recent research has laid the groundwork for reasoning about privacy budget distribution, namely, how an overall $\varepsilon$ budget can be sensibly distributed among the component pieces of a text. We perform a systematic evaluation of multiple text decomposition and budget distribution techniques in the context of DP text obfuscation, testing how different methods for chunking texts can be combined with techniques for allocating $\varepsilon$ to these chunks. Our experiments reveal that such design choices are very important, as even with comparable privacy budgets, significantly different results can occur based on which methods are chosen. In this, we provide credible evidence of the feasibility of maximizing empirical trade-offs by optimizing DP obfuscation procedures.

Outlier Exposure (OE) is among the strongest training-based OOD detectors on standard benchmarks but exhibits scorer-dependent tradeoffs (e.g., strong on MSP, weak on KNN) and requires curated auxiliary data. We show why OE works: its features sit at the same geometric locus as real near-OOD data, with the boundary-adjacent quartile driving nearly all of OE's gain. OE is boundary calibration, not OOD coverage. GEODE (GEOmetry-preserving DEtection) replicates this calibration synthetically through an angle-adaptive norm loss in which targets scale per-sample with cosine similarity to the nearest class mean, preserving feature geometry where boundary structure matters. Four theorems grounded in neural collapse justify the design. GEODE works across all seven standard scorers on CIFAR-10 (near-OOD AUROC 89.0-92.3, far-OOD reaching 93.05; no catastrophic failure on any scorer). Since the OOD regime is unknown at deployment, this is the test that matters. GEODE outperforms vanilla CE at matched epoch counts. Combined with OE, GEODE reaches 95.0 MSP / 94.8 KNN on CIFAR-10 and beats OE on every scorer on CIFAR-100. The gains hold on WRN-28-10 (+4.5 Energy, 3 seeds). Unlike methods that push OOD into the classifier null space (e.g., PFS, 14.38 KNN AUROC, worse than random), GEODE's adaptive target preserves the geometry that distance-based scorers depend on.

We present SURGE, a streaming GPU encoding system deployed in production to generate embeddings for over 800 million texts across 40,000 logical partitions. Production embedding pipelines face a tension between logical data partitioning and efficient GPU utilization: processing each partition independently incurs $P$ inter-process communication (IPC) calls whose overhead limits throughput for compute-light models. Our contributions are analytical: (i) a cost model (Theorem 1) predicting throughput within 2% across three encoders spanning a 15$\times$ parameter range; (ii) a memory-safety bound (Lemma 3) enabling a streaming two-threshold policy with peak memory $O(B_{\min} + n_{\max})$ rather than $O(N)$; and (iii) a $φ$/CV decision framework characterizing when the pattern applies beyond our workload. The naive fix of batching at fixed size requires $O(N)$ peak memory (32.7 GB at 10M texts; infeasible beyond ~60M on 192 GB nodes), produces no output until all encoding completes, and offers no fault tolerance. SURGE achieves the same throughput with $O(B_{\min} + n_{\max})$ bounded memory (2.6 GB), 68$\times$ faster time-to-first-output, and crash recovery at SuperBatch granularity. On 10M texts with 4 NVIDIA L4 GPUs, SURGE delivers 26,413 texts/s -- matching fixed-batch throughput while using 12.6$\times$ less memory. We validate on bge-base (109M, $d$=768, error 1.3%) and across log-normal $σ$ in {1.0, 1.72, 2.5} (speedup invariant within $\pm$3%), and compare against a partition-batched baseline (PB-PBP-LB), against which SURGE retains a 7% throughput edge and 2.5$\times$ faster TTFO. Complementary engineering -- zero-copy Arrow serialization (22-25$\times$ speedup) and async I/O pipelining (up to 93% benefit) -- realizes the design but is not the contribution.

Layer-aligned distillation and convergence-based early exit represent two predominant computational efficiency paradigms for transformer inference; yet we establish that they exhibit systematic incompatibility under standard deployment conditions for convergence-based early exit. Distillation objectives that align intermediate student layers to teacher representations suppress the representational convergence that early-exit mechanisms exploit, rendering such mechanisms ineffective on distilled models. We introduce LEAP (Layer-wise Exit-Aware Pretraining), an auxiliary training objective that reconciles this incompatibility. LEAP requires no architectural modifications; it augments standard distillation with a single constraint ensuring intermediate layers approximate final-layer representations. LEAP-MiniLM achieves 1.61$\times$ measured wall-clock speedup (batch=1, NVIDIA L4) at $θ$=0.95, with 91.9% of samples exiting by layer 7 and 1.80$\times$ theoretical layer reduction, where standard distilled models achieve zero effective speedup. We validate across sentence similarity (STS-B: 0.760 $\pm$ 0.006) and retrieval benchmarks (BEIR), providing operational guidance including latency measurements, decision thresholds, and deployment criteria.

While Bellman equations for basic reach, avoid, and reach-avoid problems are well studied, the relationship between value optimality and policy optimality becomes subtle in the undiscounted infinite-horizon setting, particularly for more complicated tasks. Greedily maximizing the Q-function can produce policies that indefinitely defer task completion for reach-avoid problems, or equivalently, Until specifications, even when the value function is optimal. Building upon recent results decomposing the value function for temporal logic (TL) into a graph of constituent value functions, we construct non-Markovian policies based on state history that avoid this pathology and prove their optimality with respect to the quantitative robustness score for nested Until, Globally, and Globally-Until specifications. We further show how the Q function can serve as a safety filter for complex TL specifications, extending prior results beyond simple avoid or reach-avoid tasks.

Counterfactual prompting (i.e., perturbing a single factor and measuring output change) is widely used to evaluate things like LLM bias and CoT faithfulness. But in this work we argue that observed effects cannot be attributed to the targeted factor without accounting for baseline ``meaning-preserving'' modifications to text that establish general model sensitivity. This is because every counterfactual edit is a compound treatment that bundles the variable of interest with incidental surface-form variation; this violates treatment variation irrelevance. We observe prediction flip rates on MedQA of 14.9% when we surgically change patient gender. However, this is statistically indistinguishable from the flip rates induced by simply paraphrasing inputs (14.1%). In this case, it would therefore be unwarranted to conclude that the LLM is especially sensitive to patient gender. To account for this and robustly measure the effects of targeted interventions, we propose a framework in which we compare (via statistical testing) differences observed under target interventions to those induced by paraphrasing inputs. We then use this framework to revisit a analysis done on the MedPerturb dataset, which reported evidence of model sensitivity to patient demographics and stylistic cues. We find that these effects largely dissipate when we account for general model sensitivity, with only 5 of 120 tests reaching statistical significance. Applying the same framework to occupational biography classification, we detect clearly significant directional gender bias, showing that the framework identifies real directional effects even when they are small. We evaluate a range of metrics -- aggregate, per-sample distributional, and regression -- and find that per-sample metrics are dramatically more powerful than aggregate metrics and regression powerfully and uniquely characterizes effect direction and magnitude.

Hallucinations, outputs that sound plausible but are factually incorrect, remain an open challenge for deployed LLMs. In code generation, models frequently hallucinate non-existent software packages, recommending imports and installation commands for fictional libraries. This creates a critical supply-chain vulnerability: an attacker can proactively register such packages on public registries with malicious payloads that are subsequently installed and executed by developers or autonomous agents, a class of package confusion attack known as slopsquatting. Once a model is deployed, mitigating this failure mode is difficult: full retraining is costly, and existing approaches either cause severe degradation of model utility or rely on a pre-specified forget-set, an assumption that does not apply to the unbounded space of hallucinations. To address this problem, we present Adaptive Unlearning (AU), a post-deployment framework that surgically suppresses hallucinations while preserving general model utility. AU introduces a hybrid token-level objective that simultaneously reinforces valid outputs and suppresses hallucinated ones. Combined with an adaptive discovery loop that continuously surfaces new hallucination-inducing contexts without human supervision, AU enables generalization to unseen prompts and hallucinations. We demonstrate that AU reduces package hallucination rates by 81%, corresponding to a substantial reduction in slopsquatting attack surface, while maintaining performance on standard coding benchmarks. Our analysis shows that distributional changes are concentrated on package-related generations, leaving general coding behavior largely unaffected and confirming that AU's effect is isolated to the targeted distribution. AU operates entirely on model-generated data, requires no human annotation, and generalizes across domains.

LoRA adapts large language models (LLMs) by restricting updates to low-rank subspaces of pre-trained weights. While this substantially reduces training cost, the effectiveness of adaptation critically depends on which subspace is chosen at initialization: a poor initialization that allocates capacity to task-irrelevant directions can severely hinder downstream performance. Existing initialization strategies primarily rely on the intrinsic properties of pre-trained weights, implicitly assuming that weight geometry alone reflects task relevance. However, such criteria overlook how the model interacts with the downstream data distribution. In this work, we formulate LoRA initialization as identifying the degree of impact of directions in parameter space under the target data distribution. We argue that data-aware sensitivity, rather than weight-only magnitude, should govern the choice of adaptation subspaces. Building on this perspective, we propose a Fisher-guided framework that leverages curvature information induced by downstream data to characterize how parameter perturbations influence model predictions. This perspective yields a principled, task-dependent criterion for selecting LoRA directions that better align adaptation with the target objective. Empirical results across diverse tasks and modalities demonstrate that data-aware initialization consistently and significantly improves downstream performance over existing approaches.

In this paper, we propose an approach that integrates end-to-end traceability with process modelling. OurprocessmodelsrepresentMDEworkflowsthatspan platform-independent-modelling, platform-specificmodelling, andcodegenerationphases. Processexecutionisautomated using megamodels and model transformation chains. The generation of end-to-end traceability information enables global model traceability, from high-level input models to generated code, forming the basis for traceability analysis. We have built an Eclipse-based framework, ProMoTA, to support our approach. ProMoTA extends the Acceleo model transformation language, introducing local traceability support. It also includes a global traceability map generator and end-to-end traceability analysis modules, providing users with a holistic view of the entire transformation process. Our framework is demonstrated with the use of a Wireless Sensor Network-Based IoT application.

In the envisioned future dense urban airspace, multiple companies will operate heterogeneous fleets of small unmanned aerial systems (sUASs), where each fleet includes several homogeneous aircraft with identical policies and configurations, e.g., equipage, sensing, and communication ranges, making tactical deconfliction highly complex for the aircraft. This paper aims to address two core questions: (1) Can tactical deconfliction policies converge or reach an equilibrium to ensure a conflict-free airspace when companies operate heterogeneous fleets of homogeneous aircraft? (2) If so, will the converged policies discriminate against companies operating sUASs with weaker configurations? We investigate a multi-agent reinforcement learning paradigm in which homogeneous aircraft within heterogeneous fleets operate concurrently to perform package delivery missions over Dallas, Texas, USA. An attention-enhanced Proximal Policy Optimization-based Advantage Actor-Critic (PPOA2C) framework is employed to resolve intra- and inter-fleet conflicts, with each fleet independently training its own policy while preserving privacy. Experimental results show that two fleets with distinct, shared PPOA2C policies can reach an equilibrium to maintain safe separation. While two PPOA2C policies outperform two strong rule-based baselines in terms of conflict resolution, a PPOA2C policy exhibits safer interaction with a rule-based policy, indicating adaptive capabilities of PPOA2C policies. Furthermore, we conducted extensive policy-configuration evaluations, which reveal that equilibria between similar policy types tend to favor fleets with stronger configurations. Even under similar configurations but different policy types, the equilibrium favors one of the heterogeneous policies, underscoring the need for fairness-aware conflict management in heterogeneous sUAS operations.

The co-optimization of geometry and physical parameters remains challenging in transient multiphysics systems involving moving boundaries, nonlinear material response, phase transitions, and competing objectives. Existing methods often optimize geometry and physical variables separately, rely on simplified steady-state physics, or require offline data generation and reduced design spaces. Here, we present an end-to-end differentiable co-optimization framework that couples an implicit neural representation of geometry with a JAX-compiled Eulerian multiphysics solver. Geometry is represented as a signed distance field using Fourier-feature-encoded spatial coordinates, while boundary conditions, initial conditions, process controls, and material parameters are optimized within the same differentiable loop. Continuous relaxations represent non-smooth physical transitions while preserving compatibility with reverse-mode automatic differentiation and backpropagation through time. We demonstrate the framework using a transient hamburger-cooking benchmark, selected as an interpretable multiphysics problem rather than a culinary optimization exercise. The benchmark combines conductive and convective heat transfer, latent energy effects, moisture and fat transport, shrinkage-induced geometry evolution, evolving contact boundary conditions, flipping-induced boundary-condition changes, and competing quality objectives. Results show that geometry-only optimization modifies shape to relieve thermal bottlenecks, while joint co-optimization distributes the design response across geometry, material state, process variables, and boundary conditions through gradients propagated over the full transient rollout.

A fixed-confidence, finite-sample problem of active hypothesis testing arises in many safety-critical applications. Situated in the context of sequential hypothesis testing, this paper studies the effect of hypothesis elimination on the stopping time. We introduce an elimination-augmented Track-and-Stop algorithm, in which champion-specific active-opponent sets are progressively pruned, and sensing effort is reallocated toward the surviving alternatives. Our analysis derives a non-asymptotic upper bound on the expected stopping time. The gain in finite-sample from elimination appears on the scale of the non-leading term, resulting from tighter tracking and concentration constants on the reduced hypothesis set. Furthermore, we introduce an aggressiveness parameter to modulate the trade-off between faster elimination and weaker confidence guarantee. An experimental study on synthetic Gaussian instances confirms the theoretical predictions.

We present a certified purity architecture that converts governance enforcement in cognitive workflow systems from a runtime convention into a structural capability boundary. A prior three-layer governance architecture proves governance completeness, provenance completeness, and the impossibility of ungoverned effects, conditional on the pure module constraint: that step executors cannot perform effects. That constraint was enforced by module import graph analysis, which is insufficient against adversarial bypass on the BEAM virtual machine. This paper closes the gap through four mechanisms: (1) a restricted WebAssembly compilation target where effect-producing instructions are structurally absent; (2) purity certificates, cryptographically signed proofs binding executor binaries to their import classifications; (3) a runtime verification gate that rejects uncertified executors before they enter the governance pipeline; and (4) portable governance credentials via remote attestation for cross-organizational verification. We prove four theorems: structural purity by construction, bypass elimination for all five BEAM bypass classes, certificate integrity, and gate completeness. The guarantee holds relative to an explicit Trusted Computing Base. Evaluation on four implemented executors shows verification latency of 39--42 us, full plan cycle under 400 us, runtime overhead under 0.4% of a 100 ms HTTP request, and zero determinism divergences across repeated invocations.

As large language models grow increasingly capable, concerns about their safe deployment have intensified. While numerous alignment strategies aim to restrict harmful behavior, these defenses can still be circumvented through carefully designed adversarial prompts. In this work, we introduce a theoretical framework that formalizes a game between an attacker and a defender. Within this framework, we design a theoretical best-response attack strategy and show that it is closely related to many existing adversarial prompting methods. We further analyze the resulting game, characterize its equilibria, and reveal inherent advantages for the attacker. Drawing on our theoretical analysis, we also derive a provably optimal defense strategy. Empirically, we evaluate a practical instantiation of the theoretically optimal attack and observe stronger performance relative to existing adversarial prompting approaches in diverse settings encompassing different LLMs and benchmarks.

We present an algebraic semantics for governed execution in which governance is axiomatized, compositional, and coterminous with expressibility. The framework, mechanized in 32 Rocq modules (~12,000 lines, 454 theorems, 0 admitted), is built on interaction trees and parameterized coinduction. A three-axiom GovernanceAlgebra record (safety, transparency, properness) induces a symmetric monoidal category with verified pentagon, triangle, and hexagon coherence, where every tensor composition preserves governance. An algebraic effect system constrains the handler algebra so that only governance-preserving handlers can be constructed in the safe fragment; programs in the empty capability set provably emit only observability directives. Capability-indexed composition bundles programs with machine-checked capability bounds, and a dual guarantee theorem establishes that within_caps and gov_safe hold simultaneously under all composition operators. The capstone result is the coterminous boundary: within our formal model, every program expressible via the four primitive morphism constructors is governed under interpretation, and every governed program is the image of such a program. Turing completeness is preserved inside governance; unmediated I/O is excluded from the governed fragment. Governance denial is modeled as safe coinductive divergence. The governance algebra is parametric: any system instantiating the three axioms inherits all derived properties, including convergence, compositional closure, and goal preservation. Extracted OCaml runs as a NIF in the BEAM runtime, with property-based testing (70,000+ random inputs, zero disagreements) confirming behavioral equivalence between the specification and the runtime interpreter.

We present a machine-checked formalization of structurally governed AI workflow architectures and prove that effect-level governance can be imposed without reducing internal computational expressivity. Using Interaction Trees in Rocq 8.19, we define a governance operator G that mediates all effectful directives, including memory access, external calls, and oracle (LLM) queries. Our development compiles with 0 admitted lemmas and consists of 36 modules, ~12,000 lines of Rocq, and 454 theorems. We establishseven properties: (P1) governed Turing completeness, (P2) governed oracle expressivity, (P3) a decidability boundary in which governance predicates are total and closed under Boolean composition while semantic program properties remain non-trivial and undecidable by governance, (P4) goal preservation for permitted executions, (P5) expressive minimality of primitive capabilities (compute, memory, reasoning, external call, observability), (P6) subsumption asymmetry showing structural governance strictly subsumes content-level filtering, and (P7) semantic transparency: on all executions where governance permits, the governed interpretation is observationally equivalent (modulo governance-only events) to the ungoverned interpretation. Together, these results show that governance and computational expressivity are orthogonal dimensions: governance constrains the effect boundary of programs while remaining semantically transparent to internal computation.

Multimodal Emotion Recognition (MER) is critical for interpreting real-world interactions. While Multimodal Large Language Models (MLLM) have shown promise in MER, their internal decision-making mechanisms under modality conflict and missingness remain largely underexplored. In this paper, to systematically investigate these behaviors, we introduce EmoMM, a comprehensive benchmark featuring modality-aligned, conflict, and missing subsets. Through extensive evaluation, we uncover a Video Contribution Collapse (VCC) phenomenon, where MLLM marginalize video evidence due to high token redundancy and modality preferences. To address this, we propose Conflict-aware Head-level Attention Steering (CHASE), a lightweight mechanism that detects modality conflicts and performs inference-time attention steering, effectively mitigating decision bias without retraining the backbone. Experimental results demonstrate that CHASE consistently improves performance across various settings, significantly enhancing the reliability of MLLM in complex affective scenarios.

This paper proposes and evaluates a new performance estimation method that leverages continual learning (CL) algorithms to carry out sequential simulation experiments for a feedback-based molecular communication protocol. As the protocol is sequentially examined in various experimental settings, the proposed CL-based performance estimators incrementally learn a series of unexperienced estimation tasks without compromising those that have been learned in the past. They are designed to work on a standard neural network architecture by customizing regularization and replay strategies in the loss function. Experimental results demonstrate that the proposed estimators can effectively learn on a continuous stream of simulation results and enhance the baseline neural network by improving estimation accuracy at a variety of computational costs. This paper's contribution is to establish the implications of CL in the field of molecular communication.

We introduce Xiaohongshu Social Comparison Reader Elicitation (XHS-SCoRE), a reader-grounded benchmark for detecting if a text-only Xiaohongshu (RedNote) post elicits UPWARD, DOWNWARD, or NEUTRAL/no clear social comparison from a first-person reader perspective. The task targets a socially meaningful relational signal that is behaviorally real yet not reducible to sentiment. Across prompted LLM classifiers and supervised Chinese encoder baselines, we find a consistent mismatch between generation fluency and reliable detection ability: the signal is textually learnable in-domain, but not robustly accessible to prompt-based classification. Prompted LLM classifiers exhibit stable, interpretable failure modes, especially neutralization of comparison-triggering posts and model-specific directional skew. A controlled pilot further shows that LLM-generated Xiaohongshu-style posts can shift perceived standing and comparison-related affect even when prompt-based detection of the same construct remains fragile. XHS-SCoRE contributes both a benchmark for reader-grounded comparison detection and a diagnostic framework for studying when socially meaningful relational cues remain only partially visible to prompt-based inference.

Medical large language model (LLM) evaluations rely on simplified, exam-style benchmarks that rarely reflect the ambiguity of real-world medical inquiries. We introduce the CLinical Evaluation of Ambiguity and Reliability (CLEAR) framework, which assesses how decision-space presentation, ambiguity, and uncertainty affect LLMs' reasoning on medical benchmarks. CLEAR systematically perturbs (1) the number of plausible answer options, (2) the presence of a ground truth or abstention option, and (3) the semantic framing of answer options. Applying CLEAR on three benchmarks evaluated across 17 LLMs reveals three notable limitations of existing evaluation methods. First, increasing the number of plausible answers degrades a model's ability to identify the correct answer and abstain against incorrect ones. Second, this lack of caution intensifies as the framing of abstention shifts from assertive rejection like "None of the Above" to uncertainty admission like "I don't know" (IDK). Notably, just including IDK in the answer space increases incorrect answer selections. Lastly, we formalize the performance gap between identifying the correct answer and abstaining from incorrect ones as the humility deficit, which worsens with model scale. Our findings reveal limitations in standard medical benchmarks and underscore that scaling alone does not resolve LLM reliability issues.

Shor-style quantum algorithms for the elliptic-curve discrete logarithm problem (ECDLP) are highly sensitive to the exact semantics of their group-operation oracles. Consequently, minor implementation choices can invalidate the intended mathematical model and lead to misleading conclusions. This paper introduces a semantics-first verification perspective for an end-to-end, compilable ECDLP implementation built on Qrisp. We specify the implemented oracle at the level of program semantics, derive refinement-style verification obligations for its key components, and provide a high-level complexity argument for the resulting oracle family. A small case study highlights that (i) the core point-update primitive agrees with a classical reference on well-formed inputs, yet (ii) controlled execution may violate the expected control law under the evaluated toolchain, despite a passing trivial control sanity check. These results position semantic auditing as a practical prerequisite for trustworthy ECDLP-oriented quantum software.

Partisan news media erode cross-partisan trust, but large language models (LLMs) offer a potential means of debiasing such content at scale. Across two pre-registered experiments, we tested whether LLM-generated debiasing of liberal news headlines could improve conservative readers' trust-relevant judgments. Study 1 found that subtle lexical debiasing (replacing emotive words with more moderate synonyms) had no effect on any outcome. Study 2 found that a more substantive reframing intervention significantly increased conservatives' perceived trustworthiness, completeness, and willingness to engage with liberal news headlines, without producing a backfire effect among a sample of liberals. In Study 1, the intervention produced robust effects among LLM-simulated silicon participants, whereas it had no impact on human readers. In Study 2, the intervention's effects among silicon participants aligned directionally with human responses but were significantly larger in magnitude for some outcomes. Moderation analyses revealed that the model's implicit theory of who responds to debiasing diverged from the psychological profile that actually predicted human responsiveness. These findings demonstrate that LLM-based debiasing can improve cross-partisan receptivity when targeting ideological framing rather than surface-level language, but that current models lack both the quantitative accuracy and qualitative psychological fidelity to evaluate their own interventions without human oversight.

Statistical change point (CP) detection methods typically rely on likelihood-based inference and ignore contextual information about plausible CP locations beyond the observed sequence. Although informative priors provide a natural way to incorporate such information, general and computationally efficient methods for doing so are lacking, especially for multiple CP detection. To address this gap, we propose a prior-informed CP detection algorithm (Pi-Change) that incorporates prior information on CP locations through a time-varying penalty term. We prove that the proposed penalty can be embedded in the Pruned Exact Linear Time framework while preserving the dynamic programming recursion and pruning rule required for efficient multiple CP detection. Across simulation studies and three time-series applications, Pi-Change discourages spurious CPs unsupported by prior information, remains robust to prior misspecification, and improves detection accuracy. More broadly, Pi-Change extends multiple CP detection beyond purely data-driven fitting by incorporating partial prior knowledge in a computationally efficient and interpretable way. It is particularly useful when CPs arise from heterogeneous mechanisms or are associated with known external events, helping quantify the delay between an event and the resulting structural change.

Finetuning can significantly modify the behavior of large language models, including introducing harmful or unsafe behaviors. To study these risks, researchers develop model organisms: models finetuned to exhibit specific known behaviors for controlled experimentation. Identifying these behaviors remains challenging. We show that a simple perplexity-based method can surface finetuning objectives from model organisms by leveraging their tendency to overgeneralize their finetuned behaviors beyond the intended context. First, we generate diverse completions from the finetuned model using short random prefills drawn from general corpora. Second, we rank completions by decreasing perplexity gap between reference and finetuned models. The top-ranked completions often reveal the finetuning objectives, without requiring model internals or prior assumptions about the behavior. We evaluate this on a diverse set of model organisms (N=76, 0.5 to 70B parameters), including backdoored models, models finetuned to internalize false facts via synthetic document finetuning, adversarially trained models with hidden concerning behaviors, and models exhibiting emergent misalignment. For the vast majority of model organisms tested, the method surfaces completions revealing finetuning objectives within the top-ranked results, with models trained via synthetic document finetuning or to produce exact phrases being particularly susceptible. We further show that the technique can be effective even without access to the exact pre-finetuning checkpoint: trusted reference models from different families can serve as effective substitutes. As the method requires only next-token probabilities from the finetuned model, it is compatible with API-gated models that expose token logprobs.

The record of the beginning of the most widespread legal system in the world is contained in millions of pages of handwritten text. Most of the records of the first centuries of the Anglo-American legal system are hand-written in a highly abbreviated form of medieval Latin which only a few dozen scholars in the world are trained to read. In this interdisciplinary project, we construct a dataset of 4029 lines of text across 193 medieval criminal and civil cases. We then use the dataset to train an open-source end-to-end pipeline for transcribing these manuscripts. We first train standard neural network architectures for line segmentation and handwriting recognition (R-Blla and CNN+LSTM with CTC decoding, respectively) and show that they can already achieve 79% word accuracy, despite the relatively small training set and the challenge of expanding abbreviations. We then demonstrate that simple post-processing significantly boosts accuracy: adding an n-gram language model to the CTC decoder improves word accuracy to 82%, while asking Gemini Pro 3 to correct mistakes boosts accuracy to 88%. Finally, we compare the CNN+LSTM architecture with TrOCR, a transformer-based OCR architecture, demonstrating that TrOCR shows comparable word accuracy but worse character accuracy due to its over-willingness to guess, making it harder for humans to infer the correct reading. We incorporated our pipeline into a web portal (glyphmachina.com), opening up the English legal tradition to legal scholars, medievalists, and students.

LLMs are increasingly equipped with safety alignment mechanisms, yet recent studies demonstrate that they remain vulnerable to jailbreaking attacks that elicit harmful behaviors without explicit policy violations. While a growing body of work has explored automated jailbreak strategies, existing methods face several fundamental challenges, including the lack of systematic utilization of both successful and failed attack experiences, as well as the absence of principled mechanisms for composing and selecting reusable attack rules under diverse constraints. As a result, existing methods struggle to accumulate transferable knowledge over time and to reliably adapt attack strategies across different targets and evolving safety mechanisms. To address these issues, we propose a Self-Evolving Rule-Driven Training-Free Jailbreak (SRTJ) framework that systematically discovers, composes, and refines attack strategies through interaction and feedback, without updating model parameters. Specifically, SRTJ couples experience-driven attack generation with answer set programming (ASP)-based rule selection and constraint-aware composition, where iterative verifier feedback is leveraged to jointly refine successful strategies and analyze failure patterns. The resulting rule memory evolves in a hierarchical multi-level manner, explicitly organizing distilled attack knowledge into long-term, middle-term, and short-term rules, thereby capturing both stable transferable strategies and transient adaptive behaviors to effectively balance exploration and exploitation across attack attempts. Extensive experiments on mainstream jailbreak benchmark (HarmBench) demonstrate that SRTJ achieves strong and stable attack performance across different target LLMs, while exhibiting improved robustness and generalization compared to existing jailbreak methods. The code is available at https://github.com/TheSolkatt/SRTJ.

Biosignals acquired from different locations on the body often provide temporally ordered views of the same underlying physiological process. However, most existing self supervised learning methods treat these signals as interchangeable views, overlooking the directional temporal dynamics that link them. A canonical example is the relationship between electrocardiography (ECG), which captures the electrical activation initiating each heartbeat, and photoplethysmography (PPG), which records the resulting peripheral pulse delayed by vascular dynamics. To capture this structured relationship, we introduce xMAE, a biosignal pretraining framework that leverages masked cross modal reconstruction across temporally ordered biosignals as a training time constraint to encourage physiologically meaningful timing structure in the learned representations. We show that pretraining with xMAE yields representations that outperform both unimodal and multimodal baselines on 15 of 19 downstream tasks, including cardiovascular outcome prediction, abnormal laboratory test detection, sleep staging, and demographic inference, while generalizing across devices, body locations, and acquisition settings. Further analysis suggests that the ECG PPG timing structure is reflected in the learned PPG representations. More broadly, xMAE demonstrates the effectiveness of incorporating temporal structure into multimodal pretraining when signals observe different stages of a shared underlying process. Code is available at https://github.com/hzhou3/xMAE.

Earth system science is producing increasingly large, high-dimensional datasets from physics based Earth system models to AI-based weather and climate models. Embedding-based representations can make these data searchable through similarity search and analog retrieval, but nearest neighbors in latent space are not automatically scientifically meaningful: it may reflect real weather structure, or preprocessing, geography, or model bias. Researchers therefore need ways to inspect how embeddings organize meteorological data, compare representation models, develop retrieval strategies, and verify results against physical evidence. We present an open-source visual analytics workbench for each of these steps. The system links embedding experiments to source data, metadata, spatial context, and model configurations, so latent-space results can be traced back to the physics. Users can explore latent spaces for different models, issue global or localized queries, and inspect analogs through familiar meteorological views. This enables a discovery workflow in which scientists characterize a phenomenon of interest in a well-understood dataset, identifying its signature in latent space, and then use that signature to probe larger, less-labeled archives or ensembles for similar events. We demonstrate the workbench through tropical-cyclone retrieval using ERA5-derived embeddings and IBTrACS metadata, and evaluate its out-of-core retrieval backend to show that large embedding collections can be searched beyond in-memory limits on commodity workstation hardware.

We present MedMosaic, a medical audio question-answering dataset designed to benchmark language and audio reasoning models under realistic clinical constraints. Medical audio data is difficult to collect due to privacy regulations and high annotation costs arising from domain expertise. Thus, existing benchmarks tend to underrepresent complex medical audio scenarios. To address these challenges, MedMosaic features a diverse range of medical audio types, including condition-related physiological sounds, carefully constructed synthetic voices to mimic speech with artifacts as well as real short and long length clinical conversations to model varying context lengths. The dataset also features a total of 46,701 question-answer pairs, spanning categories such as multiple-choice, sequential multi-turn, and open-ended question-answers, enabling systematic evaluation of multi-hop reasoning and answer generation capabilities. Benchmarking 13 audio and multimodal reasoning models reveals that reasoning remains challenging for all evaluated systems, with substantial performance variation across question types. In particular, even state-of-the-art model like Gemini-2.5-pro can only achieve 68.1% accuracy approximately. These findings underscore persistent limitations in medical reasoning and highlight the need for more robust, domain-specific multimodal reasoning models.

Positional encoding plays a pivotal role in determin?ing the extrapolation and generalization performance of wireless foundation models for channel state information (CSI) modeling, latent characterization, and task-specific prediction. However, existing CSI models inherit static or one-dimensional positional priors from natural language and vision architectures, which fundamentally misalign with the intrinsic physics of wireless channels by lacking explicit relative decay, collapsing the 3D spatio-temporal-frequency structure, and remaining scenario?rigid. This paper proposes Adaptive 3D-RoPE, a physics-aligned rotary positional encoding that establishes the structural corner?stone for wireless foundation models. The framework integrates a learnable, axis-decoupled 3D frequency bank to explicitly disentangle multi-dimensional phase dependencies, coupled with a lightweight channel-conditioned controller that dynamically modulates the prior via compact global CSI descriptors. This sample-adaptive mechanism transforms positional encoding from a static transformer component into a dynamic, coherence-aware inductive bias to resolve heterogeneous channel physics. Extensive experiments across 100 datasets demonstrate the superiority of the proposed scheme in both scale extrapolation and zero-shot generalization. Compared to the state-of-the-art, our method achieves up to a 10.7 dB reduction in normalized mean square error (NMSE) under 8 times antenna scale extrapolation. Given the same CSI input scales, our method can also improve zero-shot NMSE by 1.07 dB across unseen mobility scenarios and 0.90 dB in low-frequency-to-millimeter-wave tasks.

Hierarchical Gaussian Filtering (HGF) networks allow for efficient updating of posterior distributions (beliefs) about hidden states of an agent's environment. HGF parent nodes can target the mean or variance of their children. New information entering at input nodes leads to a cascade of belief updates across the network according to one-step update equations for each node's mean and precision (inverse variance). However, the original form of the update equations for variance-targeting parents(volatility coupling) can in some regions of parameter space lead to negative posterior precision, a logical impossibility which causes the updating algorithm to terminate with an error. In this report, we introduce a modified quadratic approximation to the variational energy of volatility-coupled nodes that avoids negative posterior precision. The key idea is to interpolate between two quadratic expansions of the variational energy: one at the prior prediction and one at a second mode whose location is obtained in closed form via the Lambert W function. The resulting update equations are robust across the entire parameter space and faithfully track the variational posterior even for large prediction errors.

Much research on LLMs has focused on increasing benchmark performance. However, the evaluation of such models in real-world collaborative human-AI workflows has stayed behind. This work evaluates a chatbot-style assistant based on Retrieval-Augmented Generation (RAG) in a realistic multi-turn information-seeking scenario inspired by workplace settings where compliance with local legislation and secure handling of sensitive data are often key. Specifically, we examine the performance of humans (N=112) assisted by RAG-assistants compared to LLM-only or LLM+RAG baselines. In this setting, we investigate how underlying model size (3B, 8B, and 70B) shapes the human-AI collaborative dynamic and how it influences perceived usability and satisfaction. Results show that the performance gain of human-AI collaboration over the model-only baselines is significant, irrespective of model size, suggesting that hybrid systems are beneficial in information-seeking scenarios. Interestingly, however, perceived usability and satisfaction among participants showed little difference across model sizes. This demonstrates a nuanced trade-off between model size, performance, and user perception. Our work highlights the added value of evaluating AI applications in actual multi-turn interactions with human users, looking at usability and satisfaction besides accuracy, rather than focusing on benchmark performance only.

This manuscript extends our previous multimodal human-robot interaction system by introducing a controlled ablation study of the three modules that most strongly influence end-to-end performance: the large language model used for action extraction, the perception system used for visual grounding, and the controller used for motion execution. The goal is not to redesign the full pipeline, but to isolate the contribution of each component under a common experimental protocol and then evaluate the best combinations end-to-end. We therefore compare three language models, five perception configurations, and three controllers, followed by a second-stage factorial study over the best candidates. The resulting analysis is intended to clarify which choices primarily affect execution time, which primarily affect success rate, and where the largest engineering gains are likely to come from in future revisions of the system.

We introduce energy-based constraint networks -- a modality-agnostic architecture that learns structural coherence from contrastive pairs. The system processes frozen encoder embeddings through a state-space model with dual-head attention, producing a scalar energy measuring structural consistency alongside per-position energy scores that localize violations. Multiple independently trained branches detect different violation types and compose at inference without interference. We demonstrate the framework in two domains. In text, the system achieves 93.4% accuracy on trained corruption types and 87.2% on 9 unseen types, using frozen BERT and 7.4M trainable parameters. In vision, the same architecture achieves competitive deepfake detection: 0.959 AUC on FaceForensics++ Deepfakes and 0.870 on Celeb-DF without any Celeb-DF training data, using frozen DINOv2 and 3.6M parameters per branch. The framework supports flexible training: branches learn from designer-specified corruptions, real-world paired data, or both. Composable branches require representation compatibility -- a finding validated through extensive experimentation where five incompatible approaches failed before the compatible one succeeded. The architecture is encoder-agnostic and domain-agnostic: changing the domain requires only new corruption strategies; changing the encoder requires only a new input projection layer. To our knowledge, this is the first architecture to learn within-modality structural coherence as an explicit energy landscape with per-position decomposition, and to demonstrate that the same architecture transfers across modalities via corruption respecification alone.

Achieving the right amount of trust in AI systems is important, but challenging. The problem is exacerbated with the rise of Large Language Models (LLMs) as they provide human-level communication capabilities, but potentially hallucinate in the content that they generate. Moreover, they express over-confidence in their answers, making it difficult for users to judge their truthfulness. An important human value that users seek is benevolence, which can be met by LLM's self-reflection leading to reliable and honest answers. Accordingly, this paper proposes conveying appropriate levels of self-reflected certainty to build appropriate trust. Our contributions are twofold: 1) We develop CERTA (Certainty Enhanced RAG for Trustworthy Answers), a specialized Retrieval Augmented Generation (RAG) system that incorporates the relevance between question, context, and answer to reflect its uncertainty in answering questions; 2) We create the Certainty Benchmark with 90 question-context pairs of non-objective questions, divided over four categories (factuality, preference, sycophancy, morality) and three types of contexts (relevant, incomplete, irrelevant). We run experiments with a baseline RAG system and three CERTA settings using two LLMs. Our evaluations indicate that CERTA helps identify answers that are uncertain, decreases the cases of over-agreeing, and provides cautious behavior when prompted for moral judgments.

Retrieval-Augmented Generation (RAG) equips large language models (LLMs) with external evidence by retrieving documents at inference time, but it also turns the retrieval corpusinto a sensitive asset. Under a black-box setting, an adversary given a candidate document can infer whether it has been ingested into the RAG knowledge base (i.e., document-level membership inference) solely from query response interactions, thereby leaking corpus coverage and the existence of sensitive topics. Existing RAG MIA methods either rely on soft signals such as semantic similarity, which often yield overlapping member/non-member score distributions and unstable thresholds, or employ explicit confirmation probes whose intent is conspicuous and thus prone to refusal and detection. We propose E-MIA, which converts verifiable hard evidence in the target document (e.g., fine-grained details, proper nouns/technical terms, definitional statements, metadata cues, and causal/constraint relations) into an exam with four objectively gradable question types (FB/SC/MC/T/F), and uses the aggregated exam score across multiple evidence targeted questions as the membership signal. Experiments across multiple datasets and diverse RAG configurations demonstrate that E-MIA improves member/non-member separability in stringent settings while preserving natural, stealthy queries, and we further analyze the impact of question composition and exam length on attack effectiveness.

Graph Neural Networks are powerful models for learning from graph-structured data, yet their effectiveness is often limited by two critical challenges: over-squashing, where information from distant nodes is excessively compressed, and over-smoothing, where repeated propagation makes node representations indistinguishable. Both phenomena stem from the interaction between message passing and the input topology, ultimately degrading information flow and limiting the performance of GNNs. In this survey, we examine graph rewiring techniques, a class of methods designed to modify the graph topology to enhance information propagation in GNNs. We provide a comprehensive review of state-of-the-art rewiring approaches, delving into their theoretical underpinnings, practical implementations, and performance trade-offs.

Monitoring high-dimensional engineering structures in extreme environments is limited by non-stationary excitation, nonlinear structural kinematics, and stochastic forcing. Traditional model-based and black-box data-driven methods often struggle to resolve these dynamics in real time, particularly under sensor failure or partial observability. This paper introduces a rank-optimized digital twin framework based on Koopman operator theory, Hankel-matrix embeddings, and dynamic mode decomposition. By lifting operational data into a linear invariant subspace, the method enables autonomous, input-blind reconstruction of structural states without requiring a priori mass or stiffness matrices. The framework is validated on an NREL 5MW spar-buoy floating offshore wind turbine, representing a challenging coupled aero-hydro-servo-elastic system. Results show that the rank-optimized Koopman-Hankel manifold separates structural resonances from deterministic 3P rotor harmonics under colored noise, where standard subspace identification can be unreliable. A rolling-horizon virtual sensing strategy achieves high-fidelity reconstruction at critical structural hotspots, with coefficient of determination greater than 0.95 at 1 Hz data assimilation and accuracy exceeding 0.99 at higher sampling rates. By estimating a physical Lyapunov time of approximately 1.0 s, the study defines the predictability horizon associated with the system information barrier. The proposed framework provides a computationally efficient and resilient digital twin approach for real-time identification and virtual sensing of complex structural dynamics.

De novo functional protein design aims to generate protein sequences that realize specified biochemical functions without relying on evolutionary templates, enabling broad applications in biotechnology and medicine. Existing approaches adopt either direct function-to-sequence mapping or decoupled structure-sequence generation strategies but often fail to achieve functionality and foldability simultaneously. To address this, we propose CodeFP, a Co-generative protein language model for de novo Functional Protein design that simultaneously decodes sequence and structure tokens, thereby enabling superior simultaneous realization of functionality and foldability. CodeFP utilizes functional local structures to enrich functional semantic encodings, overcoming the suboptimal translation of flat encodings into structure tokens, while introducing auxiliary functional supervision to alleviate training ambiguity stemming from the one-to-many structure-to-token mapping. Extensive experiments show that CodeFP consistently achieves average improvements of 6.1% in functional consistency and 3.2% in foldability over the strongest baseline.

The growing relevance of multi-agent systems has drawn increasing focus on communication-efficient filters for collaborative perception to alleviate the system's communication burden. While the event-triggered (ET) mechanism can improve communication efficiency in collaborative state estimation, an inevitable trade-off exists between estimation accuracy and communication cost in ET filters. This paper proposes a fast and accurate ET diffusion-based filter for real-time multi-agent collaborative target tracking, aiming to reduce the system's data transmission without compromise in tracking performance. The proposed filter achieves improved tracking accuracy, reduced data transmission, and accelerated convergence using an error-minimized ET cubature information filter (CIF) for local estimation, and a correlation-aware diffusion strategy for global fusion. The experimental results confirm the scalability of the proposed EDC-CIF algorithm and demonstrate its efficacy in simultaneously reducing estimation error and computation time while significantly enhancing communication efficiency.

Sample-level rankings are increasingly used in data-centric NLP for analysis, filtering, debugging, and curation, yet existing pipelines typically score training examples pointwise and rank them as if they were independent. This assumption is fragile in the presence of exact duplicates, near-duplicates, paraphrases, and other redundant structure common in NLP corpora, where stochastic training can make highly similar examples receive unstable relative orderings across random seeds. We study stable sample-level ranking under redundancy and propose \textsc{SCARV}, a modular aggregation framework that operates on top of an existing scoring proxy. \textsc{SCARV} combines robust multi-seed aggregation with a structure-aware aggregation/allocation step over redundancy clusters. Across synthetic redundancy, naturally mined QQP redundancy, multiple proxy families, several NLP tasks, and end-to-end DistilBERT fine-tuning, \textsc{SCARV} substantially improves over bare proxy rankings in global and local stability and yields more reproducible ranking-based decisions such as subset selection and suspicious-example retrieval. Our decomposition and compute-aware frontier sharpen the mechanism: robust multi-seed aggregation is the dominant generic stabilizer, while the structure-aware component adds value mainly under low aggregation budgets or when redundancy clusters are informative, naturally occurring, or sufficiently covered. These results position \textsc{SCARV} not as a universal data selector or a universally dominant replacement for seed-only aggregation, but as a stability-oriented aggregation layer for proxy-induced rankings in redundant NLP datasets.

Developing effective test cases capable of thoroughly exercising large-scale software systems is inherently difficult, especially if such systems have voluminous, complex, and deeply nested source codes. In this work, we present a novel approach for generating test cases using a reinforcement learning-driven agentic framework where Proximal Policy Optimization (PPO) is coupled with an LLM engine to guide prompt selection during test generation. Our approach consists of two phases. In Phase I, the ToT-guided optimization agent partitions and minimizes the source code by removing redundancies without changing the functional behavior of the source code. In Phase II, a PPO-based policy network is trained to solve the problem of selecting prompts among eight different prompting techniques, such as Boundary Value Analysis, Random Fuzzing, etc., based on the inputted 11-dimensional state vector representing the source code complexity metrics and live coverage metrics to direct the LLM engine towards exploring unvisited paths in the program. The PPO agent receives rewards based on a combination of increases in line and branch coverages, penalties for unexplored branches, and rewards for reducing source code length. From experiments conducted on twenty benchmark programs, it is evident that the proposed approach, PPO-LLM, outperforms CBMC, kS-LLM, and kS-LLM++ in terms of branch and line coverage in almost all cases, for various loop bound values ranging from BOUND~1 to BOUND~2000. While at BOUND~1, the coverage of branches is 100\% using PPO-LLM on the PALS suite, in comparison, it is around 86.8\% using kS-LLM++. This confirms that adaptive prompt selection driven by PPO substantially outperforms static prompting strategies on PALS type programs.

Flow matching has become a leading framework for generative modeling, but quantifying the uncertainty of its samples remains an open problem. Existing approaches retrain the model with auxiliary variance heads, maintain costly ensembles, or propagate approximate covariance through many integration steps, trading off training cost, inference cost, or accuracy. We show that none of these trade-offs is necessary. We prove that, for any pre-trained flow matching velocity field, the trace of the posterior covariance over the clean data given the current state equals, in closed form, the divergence of the velocity field, up to a known time-dependent prefactor and an additive constant. We call this the \emph{divergence-uncertainty identity} for flow matching. The matrix-level form of the identity is similarly closed-form, depending solely on the velocity Jacobian. Because the identity is exact and post-hoc, it is computable on any pre-trained flow matching model, with no retraining and no architectural modification. For one-step generators such as MeanFlow, the same identity yields the exact end-to-end generation uncertainty in a single forward pass, eliminating the multi-step variance propagation required by all prior methods. Experiments on MNIST confirm that the resulting per-pixel uncertainty maps are semantically meaningful, concentrating on digit boundaries where inter-sample variation is highest, and that the scalar uncertainty score tracks actual prediction error, all at roughly 10,000$\times$ less total compute than ensembling or Monte Carlo dropout.

A new interpretable experiential learning model based on state history and global feedback is presented. It is capable of learning a behavioral model represented by a transition graph between sets of states, with transitions attributed with utility and evidence count. This model is expected to be suitable for solving reinforcement learning problem in resource-constrained environments. The model was thoroughly evaluated on the OpenAI Gym Atari Breakout benchmark, demonstrating performance comparable to some known neural network-based solutions.

Traditional hallucination detection fails on "Stubborn Hallucinations" -- errors where LLMs are confidently wrong. We propose a geometric solution: Embedding-Perturbed Gradient Sensitivity (EPGS). We hypothesize that while robust facts reside in flat minima, stubborn hallucinations sit in sharp minima, supported by brittle memorization. EPGS detects this sharpness by perturbing input embeddings with Gaussian noise and measuring the resulting spike in gradient magnitude. This acts as an efficient proxy for the Hessian spectrum, differentiating stable knowledge from unstable memorization. Our experiments show that EPGS significantly outperforms entropy-based and representation-based baselines, providing a robust signal for detecting high-confidence factual errors.

Accurate modeling of commuting flows is important for urban governance, traffic planning, and resource allocation. However, the combined influence of individual intentions, geographic constraints, and social dynamics leads to considerable heterogeneity in commuting patterns, making it difficult to develop generation models that generalize across cities. To address this issue, we propose SEDAN, a Structure-Enhanced Diffusion model conditioned on Attributed Nodes for generalizable OD matrix generation. SEDAN models a city as an attributed graph. Each region is treated as a node with demographic and point-of-interest features, and commuting flows are modeled as weighted edges. Adjacency and distance matrices are incorporated to characterize spatial structure. Based on this representation, we design a fusion mechanism within SEDAN to jointly model semantic information and spatial information. Regional semantic attributes are used to model latent travel demand through graph-transformer-based node interactions, while spatial structure is injected into the generation process as explicit constraints. The adjacency matrix guides attention weights to strengthen interactions between neighboring regions. Meanwhile, the distance matrix serves as a diffusion condition to capture spatial proximity and travel impedance. The fusion of urban semantics and spatial constraints enables SEDAN to generate OD matrices that are both behaviorally plausible and geographically coherent. Experiments on real-world OD datasets from U.S. cities show that SEDAN achieves a 7.38\% improvement in RMSE over the state-of-the-art baseline, WEDAN. It also remains robust across heterogeneous urban scenarios and varying structural patterns. Our work provides an effective and generalizable solution for commuting OD matrix generation. The code is available at https://anonymous.4open.science/r/SEDAN.

We propose an arbitrary Lagrangian-Eulerian (ALE)-consistent machine learning framework for long-term fluid-structure interaction (FSI) prediction on deforming unstructured meshes. Specifically, the fluid dynamics are modeled by a surrogate that combines a graph neural operator (GNO) with a vision Transformer (ViT) for spatiotemporal prediction, while a lightweight long short-term memory (LSTM) network predicts structural kinematics at the interface. The two surrogates are coupled through a standard partitioned procedure. Most importantly, kinematic compatibility at the moving interface is enforced via an ALE-consistent boundary-correction step that updates the fluid-side interface velocity with the predicted structural velocity at each coupling update, thereby improving near-interface accuracy and long-term rollout stability. To mitigate autoregressive error accumulation, a two-stage training strategy is adopted, consisting of single-step supervised pretraining followed by long-term autoregressive fine-tuning. The proposed framework is validated on the benchmark problem of a flexible beam vibration in the wake of a cylinder. Results demonstrate accurate phase-consistent predictions over long rollouts and robust generalization under inlet-profile variations in both interpolation and extrapolation settings. Systematic ablation studies further assess the respective contributions of the ViT module, ALE-consistent boundary correction, and long-term training to predictive accuracy and rollout robustness.

Anomaly detection and localization (ADL) is critical for maintaining reliability and availability in cloud systems. Recent ADL developments focus on metric and log data, leaving event data unexplored. To address this gap, we propose EventADL, the first open-box event-based ADL framework for cloud-based service systems. To motivate the design of our framework, we conduct a systematic analysis on 520 real-world incidents, and provide insights into how anomalies and their root causes manifest through event data. EventADL has three phases: offline training, online anomaly detection, and root cause localization. During the training phase, EventADL first learns Event Semantic Patterns (ESPs), which capture normal interactions between system entities using historical event data, and then learns Event Frequency Patterns (EFPs), which capture the normal frequency of known ESPs. In the online anomaly detection phase, any data in the event stream that deviates significantly from either pattern is identified as anomalous. For localization, EventADL constructs an Intervention Graph that models the relationships between recent system interactions and the detected anomalies for automatic root cause localization. The framework is designed to operate efficiently with unlabeled data and to produce interpretable anomalies with their corresponding root causes. Our evaluation on three real cloud service systems and two real-world incidents demonstrates that EventADL outperforms existing methods, achieving F1-scores of at least 90% for anomaly detection and 100% top-3 accuracy in root cause localization.

Diffusion models have become the foundation of modern generative systems, with most research focusing primarily on improving generation efficiency and output quality. The timestep embedding component is a crucial part of the diffusion pipeline, which provides a temporal conditioning signal to the denoising network, enabling it to adapt its predictions across different noise levels throughout the process. Despite their potential to contain substantial information, timestep embeddings remain underexplored in current research, especially for security risks and reliable provenance. To fill this gap, we introduce Shadow Timestep Embedding (STE), a novel mechanism that investigates the underutilized temporal space for malicious information injection into diffusion models. In particular, when zooming in on the timestep embedding space, we find that different timesteps exhibit distinct representational capabilities that can encode side-channel information. Moreover, such encoded information can be utilized for attack and defense purposes through the scheduler interface. We present a theoretical analysis of timestep embeddings as position-encoding mappings and derive a mutual coherence evaluation that explains the separability of disjoint timestep intervals. Our findings reveal the diffusion model's timestep as a powerful side channel for carrying dedicated information, motivating new directions for adversarial generative modeling by understanding the temporal dimension.

We introduce Analytic-CPD, a structured analytic variant of coherent point drift for non-rigid point set registration. The method retains the CPD posterior correspondence layer, but replaces the point-indexed Gaussian-kernel displacement-field M-step with a finite-dimensional structured analytic mapping estimator. Posterior probabilities from the Gaussian mixture model are condensed through a barycentric identity into weighted soft target points, converting the CPD pairwise soft-correspondence objective into a weighted analytic fitting problem. The deformation is represented by a truncated multivariate Taylor mapping of a vector-valued function, so the number of deformation parameters is controlled by the ambient dimension and the analytic order rather than by an M-by-M kernel system over the moving points. A degree-continuation strategy is further introduced to stabilize large-deformation registration by progressively activating higher-order analytic modes. Experiments on two-dimensional analytic deformations and three-dimensional smooth non-analytic deformations show that Analytic-CPD achieves lower final errors and faster convergence than standard CPD in representative large-deformation settings. The results suggest that CPD-style probabilistic correspondences and structured analytic mappings provide a compact and interpretable alternative to kernel-based non-rigid registration. Code is available at https://github.com/monge-ampere/Analytic-CPD.

Continuous Glucose Monitoring (CGM) can detect early metabolic subphenotypes (insulin resistance, IR; $β$-cell dysfunction), but population-scale deployment faces two coupled problems. First, the same physiological state appears through multiple views (CGM time series, venous OGTT, Glucodensity summaries), so single-view representations fail to transfer when deployment shifts the modality or setting. Second, baselines perform inconsistently across these shifts. Both problems point to one remedy: representations that abstract away from any single view to capture higher-level temporal and distributional structure. We propose CGM-JEPA, a self-supervised pretraining framework which predicts masked latent representations rather than raw values, yielding abstraction that transfers across modalities. X-CGM-JEPA adds a masked Glucodensity cross-view objective for complementary distributional information. We pretrain on $\sim$389k unlabeled CGM readings from 228 subjects and evaluate on two clinical cohorts ($N=27$ and $N=17$ public-release subsets) across three regimes (cohort generalization, venous-to-CGM transfer, home CGM) under 20-iteration $\times$ 2-fold cross-validation. X-CGM-JEPA ranks first or second on AUROC for both endpoints across all three regimes while no baseline does, exceeding the strongest baseline by up to $+6.5$ pp in cohort generalization and $+3.6$ pp in venous-to-CGM transfer (paired Wilcoxon, $p<0.001$). Under modality shift, it matches mean AUROC while redistributing toward weaker subgroups (ethnicity AUROC gap shrinks 25-54%); on sparse in-domain venous data, the distributional view lifts label-aware clustering (ARI $+39\%$, NMI $+40\%$). Code and weights: https://github.com/cruiseresearchgroup/CGM-JEPA

This report documents the preparedness assessment of Code World Model (CWM), a model for code generation and reasoning about code from Meta. We conducted pre-release testing across domains identified in our Frontier AI Framework as potentially presenting catastrophic risks, and also evaluated the model's misaligned propensities. Our assessment found that CWM does not pose additional frontier risks beyond those present in the current AI ecosystem. We therefore release it as an open-weight model.

Federated learning (FL) is fundamentally a distributed optimization problem executed by communicating agents with local data, local computation, and partial system visibility. Once FL is viewed through that lens, hierarchy is not merely a scalability mechanism. It becomes the natural place to rethink how distributed optimization should be organized over real multi-tier networks. This article argues that hierarchical federated learning (HFL) should move beyond its common framing as a communication-saving protocol and instead be viewed as an architecture-aware design framework for networked AI. The framework is organized around three coupled design axes: architectural parameters, layer-wise optimization decomposition, and layer-wise communication realization. The first axis determines the coordination geometry of learning through hierarchy depth, layer asymmetry, and layered connectivity. The second determines how the global FL objective is decomposed across layers and highlights modular multi-layer optimization as a major opportunity beyond one dominant method everywhere. The third determines how the distributed optimization is physically realized under heterogeneous communication regimes, from interference-limited lower tiers to reliable upper tiers. A central message is that, in HFL, convergence becomes architecture-dependent: it is directly shaped by the chosen hierarchy, the assigned optimization roles, and the communication mechanisms that connect them. We develop this viewpoint using large-scale wireless edge intelligence as a flagship networked AI setting, then provide a comparative perspective on flat FL, two-tier HFL, and deep HFL together with a regime-oriented design map. The resulting perspective positions HFL as a practical methodology for designing future networked AI systems.

In this work, we introduce CellxPert, a scalable multimodal foundation model that unifies single-cell and spatial multi-omics within a common representation space. CellxPert jointly encodes transcriptomic (scRNA-seq), chromatin-accessibility (ATAC-seq), and surface-proteomic (CITE-seq) measurements, while directly incorporating MERFISH and imaging mass-cytometry data as 2D or 3D spatial-visual layers. CellxPert facilitates four key downstream tasks out of the box: (i) cell-type annotation across a broad ontology of 154 largely overlapping identities -- the largest label space addressed to date and a stringent test of fine-grained discrimination, (ii) efficient fine-tuning using Low Rank Adaptation (LoRA), (iii) genome-wide transcriptomic response prediction to in-silico perturbations (ISP), and (iv) seamless multi-omic integration across various assays and platforms. Unlike current single-cell foundation models, which approximate gene perturbations by deleting or reordering tokenized gene expression ranks, CellxPert employs a Metropolis-Hastings sampler whose proposal kernel uses the model's masked conditional distributions to transition to new transcriptomic states conditioned on the perturbed genes. This Markov-chain procedure mitigates out-of-distribution artifacts introduced by abrupt token manipulation and produces trajectories that are biologically interpretable. Evaluations on PBMC68K, Replogle Perturb-seq, Systema, and BMMC benchmarks show that CellxPert surpasses classical and state-of-the-art baselines in cell-type annotation, perturbation response prediction, and multi-omic integration.

Multivariate time series anomaly detection in ICS has attracted growing attention due to the increasing threat of cyber-physical attacks on critical infrastructure. State-of-the-art methods model inter-sensor relationships from raw time-domain amplitude values, using graph neural networks, Transformers. However, these methods discard the phase spectrum produced by time frequency transformations, We argue that phase information constitutes a complementary and previously overlooked detection modality for ICS anomaly detection. We present PhaseNet++, a frequency-domain autoencoder that operates on the Short-Time Fourier Transform (STFT) of sliding sensor windows, retaining both magnitude and phase spectra. A Phase Coherence Index (PCI), inspired by the Phase Locking Value from neuroscience, summarizes pairwise phase consistency across frequency bins into a continuous adjacency matrix. This matrix guides a graph attention network that propagates information preferentially among phase-synchronized sensors. A sensor-token Transformer encoder captures system-wide structure, and a dual-head decoder reconstructs magnitude and phase jointly via circular and coherence-aware objectives. Evaluated on the Secure Water Treatment (SWaT) benchmark, PhaseNet++ achieves an F1-score of 90.98%, ROC-AUC of 95.66%, and average precision of 91.51%. Ablation studies show that the phase-aware front-end and PCI graph module together add only 264,816 parameters, demonstrating that the phase inductive bias is lightweight. While the absolute F1-score is second best than that of all recent raw-value methods evaluated under different protocols, we position this work as the first systematic study of phase-domain anomaly detection for ICS.

The Receiver Operating Characteristic (ROC) curve of a binary classifier has often been utilized to measure the performance of the classifier. The area beneath this curve is used in particular because of its quoted probabilistic interpretation as being equal to the probability that the classifier will rank a random positive observation above a random negative observation. This paper formalizes this claim, produces a bound on how far away from the truth it is if a hypothesis is not met, and gives a small literature review of the ROC curve.

Integrating molecular, morphological, and clinical data is essential for basic and translational biomedical research, yet systematic frameworks for jointly modeling these modalities remain limited. Here we present Haiku, a tri-modal contrastive learning model trained on multiplexed immunofluorescence (mIF). It comprises 26.7 million spatial proteomics patches from 3,218 tissue sections across 1,606 patients spanning 11 organ types, with matched hematoxylin and eosin (H&E) histology and clinical metadata aligned in a shared embedding space. Haiku enables three-way cross-modal retrieval, improves downstream classification and clinical prediction tasks over unimodal baselines, and supports zero-shot biomarker inference through fusion retrieval conditioned on clinical metadata-only text descriptions. Across tasks, Haiku outperforms competing approaches, achieving cross-modal retrieval (Recall@50 up to 0.611 versus near-zero baseline), survival prediction (C-index 0.737, +7.91% relative improvement), and zero-shot biomarker inference (mean Pearson correlation 0.718 across 52 biomarkers). Furthermore, we introduce a counterfactual prediction framework in which modifying only clinical metadata while fixing tissue morphology surfaces niche-specific molecular shifts associated with breast cancer stage progression and lung cancer survival outcomes. In a lung adenocarcinoma case study, the counterfactual analysis recovers niche-specific shifts characterized by increased CD8 and granzyme B, reduced PD-L1, and decreased Ki67, broadly consistent with patterns reported for favorable outcomes. We present these counterfactual results as exploratory, hypothesis-generating signals rather than mechanistic claims. These capabilities demonstrate that tri-modal alignment via Haiku enables integrative analysis of spatial biology, bridging molecular measurements with clinical context for biological exploration.

AI-generated content (AIGC) detectors are increasingly deployed in high-stakes settings such as academic integrity screening, yet their reliability rests on a fundamental paradox: as language models are trained on human-written corpora, the statistical boundary between AI and human writing will inevitably dissolve as models improve. Commercial incentives have further distorted this landscape -- detection services and "de-AIification" tools often operate within the same supply chain, replacing evaluation of content quality with judgment of content origin. We present StyleShield, the first flow matching framework for conditional text style transfer, operating directly in continuous token embedding space via a DiT backbone with zero-initialized cross-attention adapters conditioned on frozen Qwen-7B representations. At inference, we adapt the SDEdit paradigm from image synthesis to text embeddings, with a single parameter gamma providing smooth continuous control over the evasion-preservation trade-off. On a multi-domain Chinese benchmark, StyleShield achieves 94.6% evasion against the training detector and >=99% against three unseen detectors, maintaining 0.928 semantic similarity. We further introduce RateAudit, a document-level scheduling algorithm that demonstrates detection-rate verdicts can be set to arbitrary values, directly questioning the reliability of score-based evaluation.

There has been intense debate among qualitative researchers about whether generative AI is suitable for qualitative research. In this paper, we summarize the broader ongoing discussion of generative AI in qualitative research and its implications for software engineering researchers. The qualitative research approach, small-q (positivist or post-positivist) or Big Q (non-positivist), is among the major criteria for determining whether generative AI can be used in qualitative research. In addition to research philosophy and research approach, skills, ethics, and personal preferences also play a role in researchers' decisions about whether to use AI in qualitative research.

The standard post-training recipe for large multimodal models (LMMs) applies supervised fine-tuning (SFT) on curated demonstrations followed by reinforcement learning with verifiable rewards (RLVR). However, SFT introduces distributional drift that neither preserves the model's original capabilities nor faithfully matches the supervision distribution. This problem is further amplified in multimodal reasoning, where perception errors and reasoning failures follow distinct drift patterns that compound during subsequent RL. We introduce PRISM, a three-stage pipeline that mitigates this drift by inserting an explicit distribution-alignment stage between SFT and RLVR. Building on the principle of on-policy distillation (OPD), PRISM casts alignment as a black-box, response-level adversarial game between the policy and a Mixture-of-Experts (MoE) discriminator with dedicated perception and reasoning experts, providing disentangled corrective signals that steer the policy toward the supervision distribution without requiring access to teacher logits. While 1.26M public demonstrations suffice for broad SFT initialization, distribution alignment demands higher-fidelity supervision; we therefore curate 113K additional demonstrations from Gemini 3 Flash, featuring dense visual grounding and step-by-step reasoning on the hardest unsolved problems. Experiments on Qwen3-VL show that PRISM consistently improves downstream RLVR performance across multiple RL algorithms (GRPO, DAPO, GSPO) and diverse multimodal benchmarks, improving average accuracy by +4.4 and +6.0 points over the SFT-to-RLVR baseline on 4B and 8B, respectively. Our code, data, and model checkpoints are publicly available at https://github.com/XIAO4579/PRISM.

In this paper an attractor FCM is created, tested, and analyzed. This FCM is neither a hebbian based nor agentic, nor a hybrid; it rather is a gradient descent based, physics constrained, Jacobian version of an FCM. Moreover, this model has several quirks; it uses residual memory, back propagation through time, and a fixed point anchor that is recursively implemented to update its weights. The residuals update the recursive part without losing the system memory. The model's anchor enables it to converge in a fixed point for which back propagation through time unrolls it and ensures that the error minimization is for an accurate gradient. Furthermore, a new learning algorithm is utilized. The Newton's method finds the system's fixed point attractor and then gradient descend is adaptively changing the landscape; an adaptive term is used to directly manipulate the weights through the attractor dynamics. As the adaptive term changes, the descent through the landscape is constantly adjusting according to sigmoid saturation, and that prevents premature convergence to a local minimum. Lastly, the updates are filtered by causal mask that informs the network about the physics, respecting the initial expert based opinions, for which model reduces the error to the target in an efficient way.

Peer review is a multi-stage process involving reviews, rebuttals, meta-reviews, final decisions, and subsequent manuscript revisions. Recent advances in large language models (LLMs) have motivated methods that assist or automate different stages of this pipeline. In this survey, we synthesize techniques for (i) peer review generation, including fine-tuning strategies, agent-based systems, RL-based methods, and emerging paradigms to enhance generation; (ii) after-review tasks including rebuttals, meta-review and revision aligned to reviews; and (iii) evaluation methods spanning human-centered, reference-based, LLM-based and aspect-oriented. We catalog datasets, compare modeling choices, and discuss limitations, ethical concerns, and future directions. The survey aims to provide practical guidance for building, evaluating, and integrating LLM systems across the full peer review workflow.

We present a method for approximating the many-body density of states of a system of quantum identical particles, with a reduction of the computational cost by a combinatorial factor compared to the full calculation. This is carried out by considering an isolated quantum system of identical particles, and studying its non-interacting many-body spectrum through the use of a new approach based on a separation of universal combinatorial properties from the system-specific quantities. In this paper we focus on a practical computation method that leverages our formalism of many-body combinatorics, in order to perform an efficient numerical computation of the many-body density of states. In addition, this method provides further computational improvements by allowing most of the results to be cached in persistent storage and computed incrementally, making way for efficient use of parallelization and dynamic programming techniques. We give an extensive description of the method and provide several detailed examples of approximations of bosonic many-body density of states with tunable accuracy requirements. Lastly, we demonstrate how one such approximation can be used to recover Bose-Einstein-like distributions without any particle statistics assumptions.

We extend our two-scale neural-network method for scalar singularly perturbed problems with one small parameter to dynamical systems with multiple small parameters. To accommodate multiple small parameters, we use a single effective scale parameter defined as the geometric mean of all parameters. We thus augment the network input with a scale-aware feature, enabling it to capture sharp solution transitions intrinsically. Numerical experiments across a range of dynamical systems demonstrate that the proposed framework can handle coupled systems with multiple and high-contrast small parameters and obtain satisfactory accuracy in capturing solution features induced by small parameters.

The realization of integrated, high-speed ultraviolet (UV) modulation is pivotal for the advancement of quantum information processing, portable atomic clocks, and secure solar-blind communications. While mature photonic platforms have facilitated sophisticated system-level integration across visible and infrared spectra, high-speed active modulation in UV remains with traditional bulk crystals. Consequently, a scalable integrated solution that simultaneously combines low insertion loss and extreme compactness with high modulation efficiency has remained challenging. Here, we report the first integrated UV electro-optic modulator on a thin-film lithium tantalate (TFLT) platform. By employing a compact lumped-electrode design, we achieve a record-low VπL of 85 mV\cdot cm at 375 nm, providing an up to four orders of magnitude improvement in terms of bandwidth/Vπ L over bulk technologies. The device demonstrates a robust extinction ratio of 22.7 dB, a low insertion loss of 1.3 dB, and a Vπ of 4.2V. Although the measured 3-dB bandwidth of 922 MHz is currently limited by photodetector performance, the small device footprint of 1.16 mm and electrode design of 200 μm indicate intrinsic potential for high-speed operation beyond 67 GHz which is confirmed by the electrical-to-electrical response. This work establishes TFLT as a disruptive platform for wafer-scale compatible active UV photonics, enabling the next generation of scalable quantum and communication systems.

We demonstrate a delay-programmable two-color femtosecond source based on a chirped-seed noncollinear optical parametric amplifier. Introducing controlled dispersion into the seed enables spectral selection through pump-seed delay, allowing flexible generation of two independently tunable pulse components with adjustable relative timing at high repetition rate. The temporal and spectral properties are characterized using nonlinear optical cross-correlation and dispersion-scan measurements. As a benchmark application, the source is employed in a COLTRIMS-based multiphoton ionization experiment on trapped Li atoms, revealing delay-dependent ionization pathways and demonstrating its suitability for bichromatic ultrafast spectroscopy.

Electromagnetic eavesdropping is a well-established attack vector for remotely monitoring a target activity, most notably displays, over considerable ranges. Other targets have been considered resistant to such attacks or do not exhibit sufficient electromagnetic leakage for practical exploitation. Radio-frequency retroreflector attacks (RFRA) were developed to enable covert, active monitoring of a target by implanting a minimal hardware Trojan. These implants, typically implemented using discrete components such as transistors or diodes, do not betray their presence by emitting signals themselves; rather, they modulate the electromagnetic reflectivity of the target depending on the probed signal line data. Prior RFRA work has demonstrated their viability against video links and low-speed peripheral interfaces. In this work, we extend the applicability of RFRA to high-speed targets by presenting a successful attack on the 100BASE-TX Ethernet standard. We describe the design and realization of a compact implant capable of recovering the MLT-3 encoded signaling used in Fast Ethernet, as well as a dedicated demodulation and interpretation pipeline that mitigates errors introduced by the radio channel and maximizes the amount of recovered information. Experimental results validate the feasibility of covertly monitoring Fast Ethernet traffic using RF retroreflection and highlight the viability of such attacks for high-speed links.

Spectral-based machine learning models have been increasingly deployed in chemometrics and spectroscopy, where predictive accuracy is as important as explainability. Current employed eXplainable Artificial Intelligence (XAI) methods are largely adapted from tabular or generic multivariate domains, assigning relevance to isolated spectral variables rather than to the chemically meaningful spectral zones. Widely adopted tools such as SHapley Additive exPlanations (SHAP), Permutation Feature Importance (PFI), and Variable Importance in Projection scores (VIP) were not designed for the physical continuity and high collinearity of spectral data, and their variable-level outputs require post-hoc aggregation to recover zone-level information. This study introduces the Spectral Model eXplainer (SMX), a post-hoc, global, model-agnostic XAI framework that explains spectral classifiers through expert-informed spectral zones. SMX summarizes each zone via PCA, defines quantile-based logical predicates, estimates predicate relevance with perturbation in stochastic subsamples, and aggregates bag-wise rankings in a directed weighted graph summarized by Local Reaching Centrality. A key component is threshold spectrum reconstruction, which back-projects predicate boundaries to the original spectral domain in natural measurement units, enabling direct visual comparison with measured spectra. The method was evaluated on eight real spectral datasets (six based on X-ray Fluorescence--XRF and two based on Gamma-ray Spectrometry) and one synthetic benchmark with known gr

Optical skyrmions have attracted significant attention across diverse physical systems for their promising scenarios in ultra-precise metrology, optical information processing, and quantum technologies. However, the lack of effective method for their on-chip directional transport and manipulation impedes their applications in photonic integrated devices. Here, we demonstrate a photonic platform that utilizes topologically protected valley edge state to achieve robust on-chip directional transport of optical spin skyrmions. These skyrmions originate from spin-orbit coupling within the evanescent field at the valley photonic crystal surface and exist as eigenstates of the topologically protected edge state, ensuring their robust unidirectional propagation. Leveraging the valley degree of freedom of topological edge states, we further achieve valley-locked spin skyrmions, enabling flexible control over the polarity of spin skyrmions. By endowing spin skyrmions with topological protection in momentum space, our work provides an approach for robust on-chip transport and manipulation of spin skyrmions, thereby paving the way for expanding their application potential in photonic systems.

We study how confinement affects topology and conformations in polymer films of varying thickness $h$. The knotting probability exhibits a maximum at intermediate thicknesses near the bulk radius of gyration $h \approx R_\mathrm{g,bulk}$, vanishes at small $h$ and approaches bulk values for large $h$. Close to walls, the entanglement length increases monotonically and conformations become flatter. A layer-resolved analysis of structural and topological properties allows us to reconstruct the explicit thickness dependencies by integrating layer-resolved properties of a thick film.

Near-infrared (NIR; a.k.a.\ NIRS) deep-learning studies in chemometrics increasingly report mutually inconsistent conclusions regarding convolutional neural network (CNN) design, including small versus large kernels, shallow versus deep architectures, raw spectra versus preprocessing, and single-domain training versus transfer learning. As a result, the same architecture can appear superior in one study and inferior in another, creating a practical impasse for chemometric practitioners. In this review, we argue that these contradictions are not evidence of irreconcilable methods but a structurally expected consequence of uncontrolled moderating variables. Specifically, we trace recurring disagreements to (i) the indirect nature of Vis--NIR measurement in water-dominated matrices, (ii) mismatch between effective receptive field (ERF) and the width of informative spectral structure, and (iii) validation design (including split strategy, hyperparameter tuning budget, and exposure to deployment-like shifts) acting as a hidden hyperparameter that can dominate model ranking. Building on evidence from published chemometrics and spectroscopy studies, we propose a conditional design framework that links architecture and preprocessing choices to spectral physics, dataset regime, and intended deployment scenario. Overall, the proposed perspective moves DL Chemometrics from template-driven architecture selection toward reproducible, physics-aware, and deployment-aligned model comparison.

An ultracompact thin-film polarizer for normal-incidence (0° angle of incidence, AOI) applications is analytically and experimentally investigated. The device is based on Fano resonances in periodically nanostructured dielectric thin films, enabling polarization selective reflection and transmission due to polarization dependent resonance frequencies. The operating principle is analyzed both analytically and numerically, and the optimized structure is fabricated and experimentally characterized. Measurements demonstrate polarization contrast ratios of up to approximately 1:1000 at normal incidence. Laser-induced damage threshold measurements using nanosecond laser pulses further confirm the high damage resistance of the all-dielectric polarizer.

Shape anisotropy of individual building blocks plays a crucial role in creating exotic structures and controlling phase behavior in equilibrium systems. We present a combined experimental and simulation study in which we used light-driven self-propelled rods to investigate when and how shape-induced alignment and steric and hydrodynamic interactions govern self-organization. Varying rod aspect ratio and area fraction causes the system to evolve from active Brownian motion to swarming, active turbulence, flocking, large clusters, and jamming. A state diagram summarizes emergent behaviors, and spatiotemporal analyses reveal distinct giant-number fluctuations across states. This minimal model offers insight into the self-organization of biological rodlike microswimmers, enabling the decoupling of physical from biological mechanisms. Our results provide design rules for programmable synthetic active materials and highlight parallels with bacterial swarms and other biological assemblies.

Unmanned aerial vehicles (UAVs, or drones) are likely to significantly increase the amount of air traffic. If the skies are full of UAVs, they need to interact with each other, for instance by yielding or other evasive maneuvers. The aggregated movements of drones will create traffic patterns. Just like in current road traffic, the interactions will be very frequent, so a centralized computer managing these interactions is expected not to be possible. There is a long history of traffic flow theory and modeling for 1 dimensional (road) traffic; this has been expanded to 2 dimensional traffic (pedestrians). It is unclear how traffic flow theory works for 3 dimensional traffic. In this paper we show how drone traffic can interact in a decentralized way. For the microscopic description, we add asymmetric interaction rules. We show that without centralized control, we can have efficient and safe traffic. Moreover, we provide a framework that directly links microscopic interactions to macroscopic properties. For the macroscopic description, we formulate and apply a numerical scheme that integrates the competition of space by UAVs for multiple classes, directions and dimensions. We apply both the microscopic and macroscopic descriptions to analyze (emerging) patterns which may arise in 3D traffic flow. The current paper provides background to develop interaction rules for drone traffic. Currently, the drone traffic is taking its first steps, but once the aeronautic technique takes off, the legislation regarding drone interactions should be ready. To support so, and be able to assess traffic consequences of decisions, the traffic flow theory framework developed here is essential.

High-order discontinuous Galerkin (DG) methods equipped with subcell finite-volume (FV) limiters provide an efficient framework for the simulation of nonlinear hyperbolic balance laws featuring shocks and complex flow structures. However, for systems with non-conservative terms, the design of hybrid DG/FV schemes that simultaneously guarantee high-order accuracy, robustness, and well-balancedness remains challenging. In particular, for the shallow water equations with variable bottom topography, standard flux-differencing formulations combined with node-wise subcell limiting generally destroy the well-balanced property, even if both the underlying DG and FV methods are individually well-balanced. In this work, we propose a novel flux-differencing formulation for non-conservative systems that enables node-wise subcell limiting while preserving steady states exactly. The key idea is to construct staggered DG fluxes whose non-conservative contributions are in local-times-jump form and vanish individually at equilibrium. To achieve this, we introduce a reformulation of the shallow water equations in which the source term is proportional to the gradient of the total water height. This reformulation allows the design of staggered fluxes that preserve equilibrium locally at the node level, thereby enabling arbitrary nodal blending with low-order FV fluxes. The resulting DG/FV method is high-order accurate, robust, and exactly well-balanced under node-wise limiting. Numerical experiments, including two-dimensional dam-break configurations with wet/dry fronts and complex obstacle interactions, demonstrate the improved stability and accuracy of the proposed approach. Although this work focuses on the shallow water equations, the well-balanced hybrid DG/FV methods developed here are applicable to a broader class of nonlinear systems of balance laws.

We report a significant enhancement of room-temperature photoluminescence from trivalent erbium-doped (Er3+) silicon nitride (Si3N4) metasurfaces at the telecommunication wavelength prepared via ion implantation. The metasurfaces, consisting of periodic nanocylinder arrays, are designed to support Mie-type resonances that tailor the local density of optical states. By systematically optimizing the nanocylinder radii, we achieve a photoluminescence (PL) enhancement factor of ~18 at a radius of 390 nm after thermal annealing, which is in excellent agreement with our simulations. Time-resolved PL measurements reveal a nearly ten-fold reduction in luminescence lifetime, confirming that the enhancement is primarily driven by the Purcell effect. Furthermore, we demonstrate that the PL intensity is strongly dependent on the Er3+ ion implantation depth, with a four-fold increase in emission observed from 20 nm to 80 nm ion range. These results provide a robust pathway for integrating efficient, active light sources into CMOS-compatible photonic device.

We demonstrate direct time-domain observation of l-doubling contributions in molecular rotational dynamics using shaped femtosecond laser pulses. By imposing a tailored spectral phase on the excitation pulse, we pre-compensate centrifugal distortion, which otherwise leads to temporally broadened, multi-cycle revival structures that obscure fine rotational features. A cubic spectral phase [Phys. Rev. A 107, 053108 (2023)] compresses selected revivals into near single-cycle events, in agreement with an analytic expression derived from molecular rotational constants, enabling predictive pulse design beyond numerical optimization. The resulting distortion-free revivals reveal temporally separated l-doubling contributions that remain unresolved in conventional impulsive alignment experiments. The method proves robust against experimental imperfections, including spatial light modulator discretization. While selective control of individual l-doubling components becomes feasible, here we focus on their direct observation in the time domain.

Optical solitons propagating through a multimode fiber represents one of the most fascinating class of objects exhibiting peculiar properties, with widespread potential for applications. We theoretically investigate the effect of the core refractive index distribution, characterized by the index exponent $α$, on the evolution of multimode (MM) soliton beams and their peculiar properties in graded-index multimode fibers. Our analysis reveals an optimal range $α$ = 2.04-2.08, within which MM solitons with minimum pulsewidth and characteristic energy are formed, owing to reduced modal walk-off and enhanced intermodal nonlinear interactions. Within this regime, the MM soliton undergoes efficient spatial condensation into the fundamental mode, resulting in a well-defined quasi-Gaussian output intensity profile. Notably, for some particular values of $α$, we observe a reversal of conventional energy flow associated with MM soliton condensation, leading to the net transfer of energy toward higher-order modes, akin to the thermalization of MM optical fields into negative-temperature equilibrium states. Furthermore, we show that the characteristic Raman-induced spectral redshift of MM solitons can be controlled by tailoring the refractive index distribution. Our results highlight the refractive index distribution as a key control parameter governing MM soliton dynamics and their condensation behavior and are expected to be relevant for the design and optimization of MM fiber-based systems where controlled spatiotemporal dynamics are desired.

One of the most powerful strategies to address properties of real many-body systems is to incorporate data obtained for models, for example, to use data of the homogeneous electron gas in order to build the Local Density Approximation for the Kohn-Sham exchange-correlation potential. In the present work, we examine to what extent we can use model data to design functionals directly for observables of materials. In particular, we study different approximations for the charge density of real inhomogeneous materials expressed as a simple, explicit functional of a given Kohn-Sham potential, using as central building block the Lindhard density-density response function of the homogeneous electron gas. Our increasingly realistic set of approximations includes a fully nearsighted expression equivalent to the Thomas-Fermi approximation, functional Taylor expansions, and different approximations to the Connector Theory developed in [Aouina \textit{et al.}, npj Computational Materials {\bf 11}, 242 (2025)]. In all cases, the charge density is obtained without ever solving the Kohn-Sham Schrödinger equation. Results for cubic helium, a prototypical strongly inhomogeneous material, systematically improve with higher levels of approximation, indicating that this is a promising route to obtain expressions that are relatively simple to calculate and to analyze.

To fulfill the security requirements of quantum cryptography, photon number coherence (PNC) of single photon sources has recently become an important figure of merit. Quantum dots (QDs) embedded in photonic microcavities offer a mature source of single photons, of which many properties can be tuned by the use of different excitation protocols or parameters. We show that the Swing-UP of quantum EmitteR population (SUPER) scheme can significantly decrease the PNC of the emitted photon, compared to resonant excitation. The reason for this is a laser-induced Stark shift, which effectively decouples the QD from the cavity during the SUPER excitation. Our calculations account for environmental effects such as phonons and radiative losses.

Single-event burnout (SEB) in silicon carbide (SiC) power MOSFETs is often characterized by deterministic threshold quantities. Near the boundary between recovery and runaway, stochastic variability can make this threshold description probabilistic rather than sharp. This work introduces a first-passage perspective for stochastic threshold broadening in burnout. The process is described by a reduced electrothermal feedback-relaxation model with an absorbing boundary. The model combines carrier multiplication, avalanche feedback, localized heating, carrier loss, and thermal relaxation. Stochastic carrier and thermal terms represent unresolved event-level variability. The main finding is that finite fluctuations broaden the deterministic burnout threshold into a probabilistic transition band. Noise-induced subthreshold runaway also emerges, where nominally recoverable conditions can still fail through rare stochastic excursions. First-passage-time distributions resolve the time scale of burnout and survival probabilities further distinguish rapid feedback-dominated runaway from delayed stochastic failure. A feedback-relaxation phase diagram organizes recoverable, probabilistic, and rapidly unstable regimes. This framework provides a statistical-physics interpretation of threshold dispersion in single-event burnout of SiC power MOSFETs by linking coarse-grained electrothermal dynamics to probabilistic and time-resolved failure observables.

We present a machine learning framework for testing general relativity (GR) with gravitational wave signals from binary black hole mergers. Using the source parameters of 173 BBH events from the GWTC catalog as a realistic astrophysical population, we generate simulated GR waveforms and construct beyond GR (BGR) waveforms by applying controlled phase deformations. We introduce a response function formalism that provides a systematic framework for quantifying how any observable responds to modifications of GR. We train convolutional neural networks (CNNs) on two input representations: whitened waveforms and a response function type observable derived from the waveform mismatch, which isolates the effect of phase deviations from the bulk signal. Using response functions as the CNN input improves the classification sensitivity by a factor of approximately 33 compared to whitened waveforms, demonstrating that the choice of observable representation is as important as the classifier architecture. We study the fundamental limits of this classification through Bayes optimal error analysis, averaging methods that reveal coherent patterns hidden in noise, and a comparison between CNN accuracy and a single feature classifier as a proxy for human performance. At all deformation scales, the CNN outperforms the best single feature approach. We extend the framework to physically motivated theories using the parameterized post Einsteinian (ppE) formalism and apply it to massive gravity, where the classifier detects deviations for graviton masses of order $m_g \sim 10^{-23}\;\mathrm{eV}/c^2$ with aLIGO design sensitivity.

Frequency locking between coupled laser systems provides a powerful mechanism for stabilizing and controlling coherent emission, yet its implementation and applicability down to the nanoscale remains unknown and unexplored. Here, we demonstrate optical coupling and frequency locking in closely spaced ZnO nanowire lasers operating in the extreme near field (gap < 10 nm). We observe both full and partial frequency locking, manifested as the alignment of all or a subset of the lasing modes, by spatially controlling the optical excitation. We also observe single-mode lasing in a coupled nanowire pair where the multi-mode lasing of individual nanowires is suppressed. In contrast to previously reported coupled-cavity nanowire lasers, where spectral control arises from static filtering mechanisms such as the Vernier effect, our results indicate a dynamically established relationship between actively lasing nanowires. These findings establish frequency locking as a robust and tunable mechanism in nanowire lasers, opening new routes toward stabilized and controllable nanoscale light sources for integrated nanophotonic systems.

Cation-based Ag/Cu filaments formed in an insulating $α$-Si matrix are widely used as memristors in crossbar arrays for efficient in-memory computing. However, the stochastic nature of filament formation and rupture gives rise to device-to-device and cycle-to-cycle variation. Despite successful implementation of large-scale memristor arrays, systematic studies of filament parameters and their spatial distribution in the memristors are scarce. In this work, we use conductive atomic force microscopy (c-AFM) to probe the spatial distribution of conductive filaments in $α$-Si memristors. The charge transport is dominated by a limited number of discrete filaments rather than by uniform conduction across the device area. The systematic analysis of the experiment gives the mean surface density of the conductive filaments $\sim$3200 per $μ\text{m}^2$. Both volatile and non-volatile filaments can be found within one memristor. The experimental data and the nature of volatile and non-volatile filaments may be explained within the model of multiple trap assisted tunnelling. The model yields reasonable estimates for physical properties for both types of filaments.

Optical clocks require an ultra-stable laser to probe and precisely measure the frequency of the narrow-linewidth clock transition. We introduce a portable ultraviolet (UV) laser system for use in an aluminum quantum logic clock, demonstrating a fractional frequency instability of approximately $\mathrm{mod}\,σ_\mathrm{y} = 2 \times 10^{-16}$. The system is based on an ultra-stable cavity with crystalline AlGaAs/GaAs mirror coatings, alongside with a frequency quadrupling system employing two single-pass second harmonic generation (SHG) stages. Its acceleration sensitivity, measured in all three axes, does not exceed $4(2) \times 10^{-12}$/(ms$^{-2}$) and is among the lowest recorded for transportable systems to date. Additionally, partial cancellation between photo-thermal noise and photo-birefringence noise is used to effectively mitigate noise induced by intra-cavity optical power fluctuation at lower Fourier frequencies.

A radiomap, representing the spatial distribution of wireless signal strength within a specific region, is fundamentally determined by the local propagation channel and finds extensive applications in network planning and optimization. The channel model is inherently linked to electromagnetic (EM) wave propagation, and the advent of high-frequency communications presents a new picture - microscopic (and thus negligible) scatterers in lower frequency bands become mesoscopic, rendering non-negligible EM effects. In this paper, we establish a channel model for multiple scatterers based on a spherical wave mode expansion. The source radiation, single scatterer response and multiple scatterer interactions are formed in the superposition of spherical-wave modes, capturing the multi-path effect in wave perspective. Iterative methods are used to handle the massive coupling between scatterers. This forward model is converted to an inverse optimization problem, where the scattering responses and the scatterer locations are jointly learned from sparse field measurements. A simplified approximate model is then introduced, employing fewer and simpler low-order modes while still allowing a larger number of more densely placed scatterers. Simulation results demonstrate that the proposed model accurately reconstructs and extrapolates radiomaps in both the spatial domain and the beam domain. Overall, the proposed framework offers a physically interpretable approach to localized propagation modeling.

Universal scaling near phase transitions is one of the central ideas of physics, linking the growth of spatial correlations to the slowing down of dynamics. So far, direct experimental access to this critical behavior has remained largely confined to equilibrium many-body systems, and especially to static critical behavior. Here we probe how universality emerges in a driven quantum fluid of light formed by exciton--polaritons in a semiconductor microcavity. By probing the fluctuation-dominated disordered phase below the condensation threshold, we directly measure both the static growth of the correlation length $ξ$ and the dynamical slowing down of the relaxation time $τ$. We find that these quantities obey the universal relation $τ\propto ξ^{z}$ with dynamical exponent $z \approx 2$, revealing diffusive dynamics of a non-conserved order parameter. Our results extend the physics of critical dynamics from equilibrium matter to driven optical systems, bridging quantum condensates and lasers.

Deterministic and bright quantum light sources based on scalable semiconductor technologies are a crucial building block for future quantum communication networks. While circular Bragg gratings (CBGs) are highly effective for extracting light from solid-state quantum emitters, conventional architectures rely on complex multi-layer processing or flip-chip bonding, which introduce detrimental strain and limit scalability. Here, we present a fabrication-minimal approach to realize monolithic, free-standing CBG cavities with deterministically positioned single GaAs quantum dots (QDs). By utilizing aspect-ratio-dependent etching (ARDE) in a single-step top-down process, we achieve the necessary vertical structural asymmetry for directional emission without requiring bottom reflectors. Finite-difference time-domain (FDTD) simulations validate this geometry, predicting free-space extraction efficiencies up to $68 \, \%$ and coupling efficiencies of $40 \, \%$ into a lensed single-mode fiber ($\text{NA} = 0.6$). Experimentally, the deterministically coupled QD-CBG devices yield a photoluminescence intensity enhancement of up to $\times 700$ compared to unprocessed planar QDs, reaching integrated count rates of $45 \, MHz$. Furthermore, the suspended membrane architecture effectively relaxes residual strain, significantly reducing the average exciton fine-structure splitting from $7.3 \, μeV$ in planar QDs to $1.3 \, μeV$ in the CBGs. Interferometric measurements confirm that the fabrication process preserves the optical quality of the emitters, with average coherence times of $70 \, ps$. By bridging optimized FDTD design with precise nanofabrication and robust optical performance, these results establish free-standing GaAs CBGs as a highly scalable platform for bright and coherent entangled photon pair sources.

Compositional inference - the decomposition of observations into an unknown number of latent components - is central to perception and scientific data analysis. Attention-based models perform well when components are approximately separable, as in object-centric vision. Under additive superposition, however - where multiple components contribute to every observation - we identify a structural failure mode we term slot collapse: multiple slots converge to the same dominant component while weaker ones remain unrepresented. We trace this to a general limitation: attention is memoryless with respect to explained evidence. All slots repeatedly operate on the same input without accounting for what has already been explained, so gradients are dominated by the strongest component, inducing shared fixed points across slots. As a result, attention fails to enforce non-redundant allocation under additive superposition. We address this by introducing residual evidence modeling, instantiated via evidence depletion - a minimal modification combining multiplicative depletion with an attention bias. Controlled ablations show that parallel attention, sequential processing alone, and loss-based regularization fail to resolve collapse; evidence depletion, which adds residual state to sequential attention, consistently succeeds. Across synthetic benchmarks and real-world audio mixtures (FUSS), evidence depletion reduces slot collapse by up to an order of magnitude, generalizing beyond synthetic settings. On gravitational-wave source inference for the ESA/NASA LISA mission, under identical architectures, data, and losses, standard attention fails while evidence depletion prevents collapse and enables multi-source posterior estimation. These results show that under additive superposition, residual evidence tracking is the operative ingredient for preventing collapse and enabling compositional inference.

The quasiparticle effective mass $m^*$ of the three-dimensional uniform electron gas (UEG) is a fundamental Fermi-liquid parameter whose value and density dependence have remained controversial for decades. Using renormalized perturbation theory with explicit counterterms, we determine $m^*$ in the metallic regime ($r_s \le 6$) from first principles by two complementary routes -- the self-energy and the forward-scattering four-point vertex via the $p$-wave spin-symmetric Landau parameter $F_1^s$ -- that agree within uncertainties at each density through sixth renormalized order. The resulting $m^*/m$ remains close to unity throughout the metallic regime, with a shallow non-monotonic density dependence -- a minimum near $r_s\approx 1$ followed by a gentle upturn -- reflecting the interplay of exchange and dynamical screening in the self-energy, and disfavoring strong monotonic suppression. This finding supports a physical picture for the metallic UEG in which dominant charge correlations are concentrated in nearly forward scattering and generate only a weak $F_1^s$ component.

We introduce a composition-weighted symbolic regression framework for interpretable prediction of materials properties directly from chemical composition. The method jointly learns analytical functional forms and task-dependent elemental weightings without predefined descriptors. By incorporating max/min operators, it naturally enforces constraints such as non-negative band gaps and bounded classification probabilities, unifying regression and classification tasks. Efficient search is achieved through a hybrid Monte Carlo tree search--genetic programming algorithm with gradient-based refinement and parallel computation. Benchmarks on MatBench tasks show competitive accuracy relative to state-of-the-art black-box models while yielding explicit analytical expressions. Applied to III--V semiconductor alloys, the model produces smooth composition-dependent trends and learned elemental weights with chemically meaningful periodic behavior. This framework provides a scalable and interpretable route for materials discovery and property screening.

In far-field thermal radiation, electromagnetic waves emitted by an object propagate to infinity, requiring the far region to be modeled as an effective thermal bath. This bath was proposed as infinitesimal environmental "dust", but explicit calculations with such distributed dust involve integrals over infinite space that are difficult to evaluate. In this work, we map this environmental dust to a photonic bath at infinity within the nonequilibrium photonic Green's function formalism. By explicitly evaluating the spatial integral over the dust, we show that its contribution reduces to a simple local self-energy, for which we derive analytical expressions for both three-dimensional objects and planar systems. We further demonstrate that the bath behaves as a black body and clarify its role in far-field thermal radiation. An alternative derivation based on the surface Green's function framework is also provided in Appendix B, demonstrating the theoretical consistency of the results without invoking the dust model. The photonic bath at infinity provides a convenient framework for both analytical and numerical calculations in far-field thermal radiation.

For a society to remain healthy and prosperous, people must collectively behave and act to contribute to the common good, even if there is often a tradeoff against their individual benefit. Paradigmatic examples include the adoption of sustainable behaviors and technologies to combat the climate crisis, and the mobilization for collective action to promote the rights and freedoms of repressed minorities. In this tutorial, we illustrate how game theory and network systems theory can be powerful tools to model and study this collective decision-making problem. We provide examples of how awareness of this tradeoff can impact collective change toward the societal good, exploring different problem contexts such as sustainable behavior and collective action. Finally, we review recent developments using systems and control-theoretic approaches to generate awareness and guide the emergent population dynamics towards a desired outcome, and conclude by highlighting new research and application frontiers.

Kolmogorov-Arnold networks (KANs) offer a potent combination of accuracy and interpretability, thanks to their compositions of learnable univariate activation functions. However, the activations of well-fitting KANs tend to exhibit pathologically high-curvature oscillations, making them difficult to interpret, and standard regularization penalties do not prevent this. Here we derive a basis-agnostic curvature penalty and show that penalized models can maintain accuracy while achieving substantially smoother activations. Accounting for how function composition shapes curvature, we prove an upper bound on the full model's curvature relative to the curvature penalty, and use this to motivate richer forms of penalties. Scientific machine learning is increasingly bottlenecked by the trade-off between accuracy and interpretability. Results such as ours that improve interpretability without sacrificing accuracy will further strengthen KANs as a practical tool for both prediction and insight.

In this study we analyze the topological invariants and edge states of transverse magnetic wave propagation in continuum photonic systems at a finite-width interface between two gyrotropic matrials with different magnetic bias. Where previous studies have almost exclusively considered sharp transitions between two different electromagnetic media, we consider the more general geometry where the magnetic field bias is allowed to vary arbitrarily in a finite-width interface between to bulk regions. We find that when the magnetic field bias varies continuously between the two bulk regions, the Bulk Edge Correspondence (BEC) holds robustly with respect to well-defined Chern invariants. However, discontinuities in the magnetic field bias introduce edge modes which are highly localized at the associated discontinuity and whose spectral properties alter the BEC. We analyze the spectral properties of these edge modes and define a new anomalous BEC in continuum photonic systems which includes contributions from topological invariants and discontinuities in magnetic field bias.

Sapphire optical fiber shows great promise for remote sensing in extreme environments approaching 2000 degC, by using laser-processing to form a single-mode waveguide within it. However, for practical application, longer devices with high manufacturability and reliability are required. We report the design, modeling, fabrication, and optimization of an index-guiding sapphire photonic crystal fiber Bragg grating temperature sensor. The device is fabricated using femtosecond laser direct writing to inscribe both the photonic crystal waveguide and the Bragg grating. A spatial light modulator was used to compensate for the mismatch between the immersion objective and the high-index oil used. This improved the aspect ratio and suppressed cracking during fabrication, for higher reliability. The design results in a 6-fold reduction in fabrication time over an equivalent depressed cladding waveguide, significantly reducing the cost of manufacture. Devices up to 7 cm long were fabricated and spliced to standard single-mode fiber. The propagation loss was estimated to be 0.7 dB/cm and the Bragg gratings had a bandwidth of approximately 0.12 nm. Devices were tested in a furnace showing a temperature sensitivity of between 19.0-32.3 pm/degC over a range 25-1200 degC. These longer devices have the potential to enable practical high precision extreme temperature monitoring in many applications, with lower manufacturing cost and higher reliability.

Variational quantum algorithms are promising for near-term quantum computing, but are severely limited by hardware noise and the substantial circuit overhead required for error mitigation methods such as Zero-Noise Extrapolation (ZNE). We propose a Physics-Informed Denoising Network (PIDN) that reduces the cost of ZNE by learning a surrogate model of its optimization dynamics. By viewing the variational update as a trajectory in the parameter space, PIDN is trained to reproduce ZNE-mitigated expectation values and gradient directions while incorporating a physics-informed loss that preserves the gradient descent dynamics. Once trained, PIDN replaces repeated multi-noise evaluations with denoised expectation and gradient estimation directly from the current noisy observation and the historical trajectory, significantly reducing circuit executions. We benchmark the approach on the quantum approximate optimization algorithm for 3-regular graphs, Sherrington-Kirkpatrick, and transverse-field Ising models, as well as the variational quantum eigensolver for LiH, BeH$_2$ and H$_2$O. Across all tasks, PIDN attains performance comparable to ZNE, while reducing the number of circuit executions by a factor of approximately 4 to 6. Gradient cosine similarity with ZNE remains above 0.95 throughout training. Robustness analysis shows that PIDN fails only when ZNE itself becomes unreliable, and ablation studies confirm the necessity of the physics-informed loss for maintaining directional consistency. We further find that PIDN tracks optimization dynamics most accurately when the effective loss landscape retains strong low-frequency structure. These results establish PIDN as a scalable, resource-efficient strategy for noise-resilient variational optimization in the noisy intermediate-scale quantum regime.

Cobalt ferrite (CoFe$_2$O$_4$) is a prototypical ferrimagnetic spinel oxide whose exceptional magnetic anisotropy, magnetoelastic coupling, and thermal stability underpin applications in spintronics, magnetic hyperthermia, energy harvesting, and catalysis. This chapter presents a comprehensive computational framework that integrates electronic$-$structure calculations with atomistic spin modeling, statistical mechanics, and continuum micromagnetics to predict magnetic functionality across length and time scales. Starting from density functional theory with Hubbard corrections (DFT$+$U), we derive exchange constants J$_{ij}$, magnetocrystalline anisotropy K$_1$, and magnetoelastic coefficients B$_1$, accounting for cation inversion, strain, and correlation effects. These parameters feed into generalized Heisenberg Hamiltonians, enabling Monte Carlo and Landau-Lifshitz-Gilbert simulations of finite-size effects, hysteresis, coercivity, and hyperthermia response in nanoparticles and thin films. Coarse-graining strategies bridge to micromagnetic modeling, ensuring consistent parameter flow without empirical fitting. Computational case studies demonstrate size-dependent anisotropy enhancement, surface spin disorder, strain-tunable switching, and doping trends, revealing design principles inaccessible to experiment alone. Validation against benchmarks, e.g. Curie temperature, anisotropy constants, coercivity, magnetostriction, confirms predictive accuracy. Current challenges, e.g., U$-$parameter sensitivity, realistic surface chemistry, spin-lattice coupling, and large-scale integration are discussed alongside emerging directions including DFT$+$DMFT, coupled dynamics, and machine-learned potentials.

Photon upconverters are anti Stokes emitters capable of generating photons with higher energy than those absorbed. This behavior can be achieved through different mechanisms, which are extensively studied for applications in imaging, anticounterfeiting, information encryption and most importantly, solar energy technologies. Among these mechanisms, photon upconversion based on sensitized triplet-triplet annihilation (sTTA-UC) is particularly attractive because it operates under low intensity, incoherent light. In sTTA UC, two optically dark triplet states of a conjugated annihilator fuse upon collision to populate a higher energy fluorescent singlet, with the triplets initially generated via energy transfer from a lower energy sensitizer. Here we introduce a general molecular design strategy to maximize singlet generation through TTA. By selective substitution, we engineered a naphthalene derived annihilator with an excited state energy landscape that strongly favors singlet formation, achieving yields up to 0.83. When combined with an appropriate triplet sensitizer, the system delivers stable UV/visible upconversion peaking at 390 nm, with an absolute upconversion yield about 0.19 and an activation excitation intensity threshold lower than 0.1 suns under non-coherent broadband excitation fully compatible with the requirements of solar powered technologies.

The exceptional nonlinearity of alkali-metal vapors enables highly efficient nonlinear optical processes even at relatively low optical intensities. However, such processes have traditionally relied on centimeter-scale vapor cells. Here, we utilize a versatile chip-scale Rb vapor platform to generate coherent blue and mid-IR light in continuous-wave mode by means of resonant four-wave mixing. Optimized optical overlap with the atomic medium enables blue light generation of $\sim$20 $μ$W over a very short interaction length, while maintaining a directly measured linewidth of $\sim$1 MHz, which is presently limited by the measurement apparatus. Comparison with a conventional glassblown vapor cell further shows that the micromachined platform can achieve higher coherent blue-light generation efficiency despite its substantially shorter interaction length. Moreover, an anodically bonded Si window enables to detect coherent mid-IR emission with collected powers of $\sim$50 nW. We further characterize the temperature dependence and input-power scaling of the blue emission, confirming efficient nonlinear conversion within these compact vapor cells. This chip-scale platform provides a versatile foundation for a range of nonlinear optical functions, from precise wavelength references and quantum light sources to next-generation quantum sensors.

In non-relativistic mechanics, the total (orbital) angular momentum (AM) of a spatially-distributed system can be decomposed into intrinsic and extrinsic contributions. In relativistic quantum mechanics, intrinsic AM is typically associated with spin, which can be described using the Pauli-Lubanski four-vector. Here, we develop a unified formalism that combines the main features of both approaches and describes the intrinsic AM of a relativistic wavepacket, including both spin and orbital contributions. Our approach is based on the "expectation Pauli-Lubanski vector" constructed from the expectation values of the wavepacket's momentum and AM. Equivalently, it defines the intrinsic AM relative to the wavepacket's energy centroid. In contrast to the conventional Pauli-Lubanski formalism, the zero-mass singularity does not occur for the expectation Pauli-Lubanski vector. Consequently, the intrinsic AM of a wavepacket may have an arbitrary orientation with respect to its momentum, even for massless particles. We illustrate the general theory with a number of examples of relativistic wave beams and packets carrying spin and orbital AM.

The rising complexity of cutting-edge cryogenic systems is currently imposing challenging technical constraints to the monitoring of ultra-cold temperatures through standard commercially available sensors. Among different alternative technologies, superconducting microwave resonators have been recently investigated as ideal candidates for performing on-chip cryogenic thermometry, in reason of their intrinsically low power dissipation, typically large temperature sensitivities and excellent sub-mK resolution below 10 K. In such a framework, through this study we aim at demonstrating the possibility to enhance the temperature performance of superconducting microwave resonators by means of surface nanostructuring. More specifically, different arrays of nanogaps are strategically patterned on the inductive line of a 1.3 GHz planar resonator, by partially etching a Nb50Ti50 thin film, in order to tune the critical transition of the material and, therefore, increase the curvature of the fres(T) response. Although the presence of such weak-links introduces larger microwave losses, a 1.5 K decrease of TC is recorded, which directly translates into an enhancement of the temperature sensitivity by a factor 10, with respect to a reference non-nanostructured sensor. In particular, a maximum value of dfres/dT = 62 MHz/K, at 4.2 K, is achieved for the device showing the largest nanogap width of about 350 nm, demonstrating that the surface nanostructuring of superconducting thin-films can be effectively engineered to enhance the temperature response of microwave resonators for high-performance cryogenic thermometry. We believe that similar approaches might be investigated and, eventually, adopted for the near-future development of the next generation of low-temperature sensors.

Modeling the atomic structure of amorphous materials has long been a critical challenge in materials science. Recent advances in monolayer amorphous materials enable direct observation of their atomic structures, paving the way for a better understanding of their atomic-scale models. Here, we investigate amorphous multielement monolayers using machine learning potential from first-principles total energies via energy-driven kinetic Monte Carlo based active-learning framework. A polymorphic crystallite model is proposed to describe the atomic configuration of monolayer amorphous boron nitride, as it consists of coexisting crystallite of $o-B_2N_2$ and $o-B_4N_4$ structural motifs. Generality of the polymorphic crystallite model is further validated in two other multielement monolayer amorphous systems. Monolayer amorphous LiCl shows coexisting hexagonal and tetragonal crystallites, while monolayer amorphous BCN contains a combination of graphene-like, h-BN-like, and borophene-like crystallites. These findings expand the classical picture of amorphous structure models and offer new insight into the microscopic structure of amorphous materials.

We propose Local Optima Networks (LONs) as a formal framework for modeling innovation dynamics. A LON is a directed weighted graph in which nodes represent locally stable technological configurations and edges encode transition probabilities between their basins of attraction. We construct LONs from fitness landscapes and model innovating agents as stochastic walkers exploring the adjacent possible on the resulting network. We show that this model simultaneously generates the four main empirical regularities of the discovery-process tradition: sublinear novelty growth (Heaps' law), heavy-tailed frequency distributions (Zipf's law), anomalous fluctuation scaling (Taylor's law), and power-law distributed inter-event times. The exponents fall within empirically observed ranges and are jointly constrained by LON topology. Communities in the LON provide an operational definition of technological paradigms grounded in basin-level accessibility. The LON framework thus bridges the discovery-process and adaptive-search traditions of innovation modeling within a single, parsimonious, and empirically testable representation.

Criticality has been proposed as a key principle underlying complex behavior in biological and artificial systems; however, how criticality translates from individual dynamics to collective behavior remains unclear. We study this question using a multi-agent system with spatially constrained interactions in which agents sense neighboring light signals through exteroceptors and act by switching their own light on or off, thereby forming a dynamical interaction network at the macroscopic level. The agents' internal states are themselves governed by a reservoir dynamical system at the microscopic level. By varying the microscopic parameters around dynamical criticality, together with the macroscopic interaction topology, we systematically investigate the relation between the two levels. We find that near-critical dynamics within individual agents is not sufficient to produce collective critical-like avalanche statistics. Instead, scale-free behavior depends on the effective connectivity of the macroscopic interaction network, which controls activity propagation. As a result, macroscopic critical-like dynamics are enabled by microscopic regimes that deviate from criticality, with the required deviation depending on the properties of the interaction network. Investigating this relation, we find that slightly subcritical micro-level regimes support near-critical dynamics across a wider range of macroscopic parameters. These results show that in this multi-agent system, collective near-critical behavior depends on the interplay between internal dynamics and the interaction structure that governs activity propagation.

Cathode particle fracture is widely recognised as a major degradation mechanism in lithium-ion batteries, yet cracking also permits electrolyte wetting of newly exposed internal surfaces, modifying interfacial reaction pathways. The mechanistic role of electrolyte wetting in redistributing reactions within cracked particles remains unclear. Here, we isolate this effect through a controlled comparison between (i) a fully coupled electro-chemo-mechanical model resolving lithium concentration, electrostatic potential, and stress fields in both the active material and the electrolyte inside and outside cracks, and (ii) a single-particle chemo-mechanical model employing the conventional uniform flux assumption. The coupled model predicts strong spatial heterogeneity in interfacial reaction rates, with flux amplification approximately 8x relative to the imposed uniform flux at the crack tip. Reaction redistribution, and thus lithium flux, is governed predominantly by local solid-state lithium concentration and stress variations, while electrolyte potential gradients inside cracks remain secondary under the conditions considered. Uniform flux models can underpredict delivered capacity by 25% at 1C-rate; this discrepancy increases at higher rates. They also underestimate tensile stresses throughout the delithiation process by 10%, directly affecting crack driving conditions. These results demonstrate that neglecting crack-electrolyte coupling leads to systematic underestimation of both utilisation limits and fatigue-relevant stress histories.

An optomechanical oscillator undergoes a Hopf bifurcation that connects two dynamical regimes with different information-processing capabilities: thermal Brownian motion and coherent self-sustained oscillation. Below threshold, the oscillator occupies a stable fixed point around which thermal fluctuations drive stochastic Brownian motion - a regime dominated by linear response, with only short-lived memory and negligible usable nonlinearity. Above threshold, radiation pressure, free-carrier dynamics, and thermo-optic relaxation act together to sustain a stable limit cycle that simultaneously provides both nonlinear transformation and dynamical memory. Here we show that this coherent regime can be used as a physical reservoir for computation: by perturbing the phonon-lasing attractor, the cavity performs nonlinear input-output transformations and retains short-term memory without any external feedback mechanism. Using only a single chip-integrated device with 20 virtual nodes, we reconstruct nonlinear functions, predict the evolution of chaotic time series, and perform spoken digit classification on a two-digit sub-task. The mechanical resonance frequency sets the intrinsic dynamical timescale of the reservoir and therefore its processing speed; while the present device operates near 0.4 GHz, optomechanical and nanomechanical systems can be engineered to reach multi-GHz and sub-terahertz frequencies, directly translating into a scalable path toward ultrafast integrated physical computing.

Phase-transforming pyroelectric materials have emerged as promising candidates for low-grade thermal energy harvesting. However, whether first-order transformations with large pyroelectric coefficient or second-order transformations with better reversibility are preferable remains unclear. Here we report compositionally tunable phase transformations in Ba$_{1-x}$Sr$_x$TiO$_3$ ($x \in [0, 0.3]$), revealing evolution from first-order to second-order character. We identify a transitional regime between Sr$_{0.15}$ and Sr$_{0.22}$ where transformation mechanism fundamentally changes. Within this regime, Sr$_{0.19}$ achieves optimal lattice compatibility, exhibiting electrical leakage suppressed by over two orders of magnitude while retaining substantial polarization response. Energy conversion demonstrations show the multilayer Sr$_{0.19}$ device delivers pyroelectric current of $\sim$1.6 $μ$A at 64$~^\circ$C with an energy density of 1.6 mJ/cm$^3$ per cycle and 5.5\% conversion efficiency. Remarkably, this composition operates stably over 10,000 full energy conversion cycles without external bias field or recharging, demonstrating that transitional regime compositions provide the optimal balance between energy density and operational durability for practical low-grade heat harvesting.

Coupled spin-lattice dynamics (SLD) underlie a wide range of magnetic phenomena, yet a unified first-principles framework that propagates both degrees of freedom without empirical parameterization has remained elusive. We present a fully ab initio SLD approach integrated into VASP, in which interatomic forces and effective magnetic fields are obtained at each time step from self-consistent constrained-moment density-functional calculations. The method is validated on four materials spanning ferromagnetic, non-collinear, and geometrically frustrated orders, recovering the correct magnetic ground state in every case from random initial conditions. SLD trajectories also provide physically correlated training data for magnetic machine-learning potentials, as demonstrated for BiFeO$_3$ by a reduction of up to approximately one order of magnitude in energy MAE over training on randomized spin configurations. This framework opens a practical first-principles route to finite-temperature spin-lattice coupled phenomena in magnetic materials.

The plaque reduction assay (PRA) remains the gold standard for antiviral susceptibility testing, evaluating drug potency by measuring reductions in plaque-forming units (PFUs). However, the traditional PRA is time-consuming, labor-intensive, prone to manual counting errors, and offers limited scalability. Moreover, its reliance on destructive fixation and chemical staining reduces the assay to a static, endpoint observation, obscuring the dynamic, time-resolved kinetics of dose-dependent viral inhibition. Here, we introduce a label-free, time-resolved PRA platform that transforms the conventional assay into a continuous, high-dimensional measurement of viral infection dynamics. Our system integrates a compact lens-free imaging setup with a custom-designed ultra-large-area (100 cm^2) thin-film transistor (TFT) image sensor and deep learning-based algorithms to autonomously quantify PFU dynamics within an incubator. Validated using herpes simplex virus type-1 (HSV-1) treated with acyclovir, the platform matched chemically-stained ground truth measurements with zero false positives while accelerating readout by ~26 hours. Crucially, our system revealed that increasing drug concentrations induce temporally distinct delays and suppress new PFU formation, enabling conclusive drug efficacy evaluations within ~60 hours post-infection. This scalable, label-free framework redefines antiviral susceptibility testing as a rapid, time-resolved and information-rich measurement framework, providing a generalizable platform for virology research, high-throughput drug screening, and clinical diagnostics.

Intelligent reflecting surfaces (IRSs) have attracted considerable attention because of their ability to dynamically control electromagnetic wave propagation. While most existing IRSs have been developed for low-power communication and sensing applications, their extension to high-power wireless power transfer (WPT) environments remains largely unexplored, as the high induced currents can damage or saturate the sensitive control elements, disrupting their tuning functionality. Here, we propose a metal-oxide-semiconductor field-effect transistor-based (MOSFET-based) binary control circuit for IRSs operating at 2.4 GHz that can withstand input power levels exceeding 1 W per unit cell. The control circuit employs a back-to-back MOSFET switching topology with series and parallel capacitors to suppress impedance variations arising from device nonlinearity while maintaining a reflection phase difference of approximately 180 degrees between the ON and OFF states. A theoretical model based on transmission lines is developed and validated against full-wave co-simulations incorporating nonlinear SPICE device models. The dynamic range is evaluated with respect to both the rated current and the reflection phase difference, demonstrating stable operation up to 1.25 W. Supercell-level beam steering is further demonstrated through far-field simulations, confirming active control of the reflection angle via switching pattern reconfiguration. These results establish a foundation for the deployment of IRSs in high-power WPT scenarios.

The approximate master equation (AME) provides a highly accurate description of dynamical processes on networks, yet its steady states are generally analytically intractable. In this study, we develop an analytical framework to derive the steady states of the AME by introducing a controlled approximation that enables closure of the moment equations. This framework reproduces the steady state of the pair approximation by achieving closure with the minimum required order of moments, and can be systematically refined to approach the exact steady states of the AME. We apply this to the SIS model, the voter model, and evolutionary games, demonstrating that the steady states can be derived. In particular, for evolutionary games, we show that combining our framework with the singular perturbation method enables the analytical derivation of the time evolution.

This paper is the direct-formulation companion to [Burbano-Gallegos, P'erez-Arancibia, and Turc, ESAIM: M2AN, 60(1):273--315, 2026], which developed indirect combined-field-only boundary integral equations (BIEs) for time-harmonic electromagnetic scattering by smooth perfectly electrically conducting (PEC) obstacles, relying entirely on Helmholtz boundary integral operators. Here we exploit the same equivalence between the Maxwell PEC scattering problem and a pair of vector Helmholtz boundary value problems -- one for the electric field and one for the magnetic field -- to derive direct BIE formulations whose unknowns are the Dirichlet and Neumann traces of the total fields, decomposed into their normal and tangential surface components. These unknowns carry direct physical meaning: in particular, the magnetic-field formulation yields the surface electric currents as part of its solution. The mixed regularity of the two field-trace components requires introducing a tailored product H"older space, a distinctive feature absent from the indirect approach. We prove that the resulting Direct Electric and Magnetic Combined-Field-Only Integral Equations (D-ECFOIE and D-MCFOIE) are uniquely solvable at all frequencies, and introduce Calder'on-type regularizations (RD-ECFOIE and RD-MCFOIE) that render them of the Fredholm second kind. We further examine the low-frequency breakdown affecting the electric-field formulation and introduce a modified equation that enforces the physical charge-conservation constraints, which restores numerical accuracy and well-conditioned linear systems for frequencies arbitrarily close to zero. Numerical experiments, performed using a high-order Nystr"om solver based on the Density Interpolation Method and implemented in the Julia package Inti.jl, validate the accuracy and robustness of the proposed formulations across a range of geometries and frequencies.

Lumens are cavities enclosed by polarized cells that are essential for organ function, from nutrient transport in the gut to gas exchange in the lungs. Defects in lumen formation are associated with severe diseases, including polycystic kidney disease and respiratory malformations. The emergence, growth, and maintenance of lumens involve a rich set of phenomena that can be framed within out-of-equilibrium physics and biological active matter, including osmotically driven hydraulic flows, coarsening-like dynamics, morphological instabilities, and mechanochemical feedbacks linking luminal pressure to tissue response. Yet experimental and theoretical efforts to study these phenomena have largely developed within specific biological systems, complicating the identification of shared physical principles across them. In this review, we bring these efforts together and present lumenogenesis within a biological physics framework in which lumens are viewed as active balloons: pressurized cavities that are inflated, sculpted, and maintained through tightly coupled active processes.

On the grounds of nonadditive entropies -- appropriate for complex systems -- we investigate the electroencephalogram amplitudes of typical and ADHD children. The corresponding probability distributions are $q$-Gaussians, i.e., $ρ(x) \propto e_q^{-βx^2} \equiv [1+(q-1) βx^2]^{1/(1-q)}$, where $(q,β)$ are, respectively, the entropic index characterizing complexity and the inverse width. We show that $q$ tends to monotonically vary with $β$ for both typical and ADHD subjects, thus revealing critical behavior of the brain. Moreover, we verify that ADHD subjects have a higher complexity than the typical ones. Consistently, biomarkers for objective phychyatric diagnosis could emerge along this path. We show that $q$ tends to monotonically vary with $β$ for both typical and ADHD subjects, thus revealing critical behavior of the brain. Moreover, we verify that ADHD subjects have a higher complexity than the typical ones. Consistently, biomarkers for objective phychyatric diagnosis could emerge along this path.

Strong light-matter coupling gives rise to polaritons - hybrid excitations whose mixed photonic and matter character enables control over optical, electronic and chemical properties. This Feature Article surveys the main architectures supporting polariton formation, including photonic microcavities, plasmonic nanostructures, open cavities and metasurfaces, and outlines how inorganic semiconductors, organic aggregates and hybrid systems access strong and ultrastrong coupling. Key phenomena such as coherent dynamics, vibronic interactions, dark-state reservoirs and polariton-mediated energy and electron transport are discussed, together with the experimental and theoretical tools used to study them. We highlight examples where strong coupling modifies charge transport, energy flow and chemical reactivity, and we summarize emerging regimes, including intermediate and dark-strong coupling, that broaden the landscape of hybrid light-matter physics.

Imperfections in X-ray imaging systems can limit their performance, especially in High Energy Density (HED) or Inertial Fusion Energy (IFE)-relevant experiments that are typically single shot, by introducing structured, non-stationary features that overlap with the signal of interest. When the X-ray transmission is reconstructed by typical flat-field normalization, even small shot-to-shot drift of structured features imprints residual patterns onto transmission maps, degrading signal visibility and biasing measurements such as electron density, velocity and feature sizes. We investigate this limitation by modeling the artifacts as a separable feature layer and training a U-Net architecture to estimate and infer them directly from the experimental data. We compare our method against Fourier filtering and more advanced procedures like Dynamic Flat-Field Normalization (DFFN) to evaluate artifact suppression capability and signal preservation in the reconstructed transmission maps. In multiple synthetic injection tests, our Physics-Guided Deep Learning approach is able to obtain an improvement in mean Structural Similarity Index (SSIM) from 0.345 to 0.906 and from 0.0679 to 0.945, while better preserving filament profiles and reducing degradation of the filament signal during artifact suppression. Additionally, we utilize deep ensembles to obtain predictive epistemic uncertainty estimates for the U-Net based reconstruction, to ensure Out Of Distribution (OOD) robustness for this procedure.

Programmable folding of elastic sheets typically relies on predefined flexible creases or active materials-enabled hinges, which lack intrinsic bistability and limit reprogrammability within a single structure. Here, we present a dimple-encoded origami platform that converts bistable dimple snapping into spatially addressable hinges with prescribed folding angles in a continuous sheet. This interaction-enabled mechanism enables the design of distributed hinge networks through the arrangement and selective inversion of dimples. We establish folding-angle design charts that can be directly used to select local dimple arrangements for target fold angle, forming a practical hinge library without altering the underlying unit geometry. Using this approach, a single dimpled sheet can be reprogrammed to realize multiple distinct configurations, such as triangle, square, and pentagon shapes. We further extend the method to flat-to-3D morphing of polyhedral origami and validate the results through experiments and finite element simulations. As demonstrations, we realize self-supporting cubic shells with enhanced impact resistance and partially deployable cube configurations that remain stable upon opening, highlighting their potential for protective enclosures and deployable architectural structures. The proposed strategy provides a fabrication-friendly route to reprogrammable shape-morphing and adaptive mechanical systems.

Autonomous vehicles (AVs) are widely known to follow conservative, rule-based motion policies that surrounding drivers can learn to anticipate. A direct consequence is that human drivers may accept shorter longitudinal gaps when cutting in front of an AV than when targeting another human-driven vehicle (HDV). We test this hypothesis using the Waymo Open Motion Dataset (WOMD), which provides 25,906 real-world highway scenarios at 10 hertz. An eight-criterion lane-change detector extracts 706 HDV-to-AV and 3,172 HDV-to-HDV cut-in events from the same traffic environment. The median accepted gap in front of the Waymo AV is 7.58 meters versus 9.57 meters for HDV targets, a 1.99 meter reduction that is statistically significant (p equals 5.76 times 10 to the negative eighth power, d equals negative 0.224) and persists under speed-matched resampling. Cut-in speeds toward the AV are 37 percent higher (51.7 versus 37.7 kilometers per hour, d equals 0.502), and 68.0 percent of AV-targeted cut-ins occur below the 10 meter gap boundary versus 51.8 percent of HDV-targeted events (chi-squared equals 60.5, p is less than 10 to the negative thirteenth power). These results reveal a systematic and safety-relevant asymmetry in human gap-acceptance behavior that warrants AV-specific calibration of both motion-planning safety envelopes and traffic simulation models.

Molecular dipoles near interfaces emit highly directional radiation due to near-field interactions, making surface-bound fluorophores sensitive probes of local physicochemical changes. We introduce smart coverslips, stably coated with uniform, brightly fluorescent nanobead films, that exploit refractive-index-dependent emission shifts for sensitive micro-refractometry in small volumes. Supercritical-angle fluorescence refractometry uses single back-focal-plane images to allow us real-time RI sensing and nanometric thin-film height measurements without the need for multi-angle or multi-wavelength acquisition. Our fast, label-free, and non-invasive approach allows measurements of thin-film properties and monitoring of interfacial dynamics on a standard inverted microscope and is broadly applicable to nanobiophotonics, chemical sensing, and in-situ materials analysis.

We design a system with a binary suspension of colloids and a surface that triggers the self-assembly of crystallites with a finite thickness. The proposed design allows controlling the number of layers forming the aggregate and constrains the two types of particles to lie on different planes. These functionalities are achieved by decorating the colloids and the surface with multiple DNA oligomers featuring specific interactions. The surface triggers a chain of reactions between DNA oligomers, leading to localized self-assembly. Equilibrium principles control the thickness of the aggregates. Instead, compositional order is achieved by engineering the reaction kinetics between DNA oligomers in a way that limits interactions between colloids of the same type. We validate our design using theory and reaction-diffusion simulation algorithms, which capture the multibody nature of the interactions. This work demonstrates how engineering the kinetics provides a new avenue for controlling the morphology of aggregates assembled by DNA.

Nonlinear energy exchange between vibrational modes underlies phenomena ranging from internal resonance to wave mixing, yet modal interactions are typically inferred from frequency-domain signatures rather than directly observed in space. Here, we present real-space imaging of nonlinear modal energy routing in a near-mirror-symmetric microelectromechanical resonator using phase-locked multi-harmonic stroboscopic interferometry. By reconstructing the spatial eigenmode content of individual harmonic components, we directly resolve the energy transfer pathway between interacting modes. Our measurements reveal that nonlinear energy exchange is not governed by frequency proximity alone. Even when harmonic frequencies lie closer to an opposite-symmetry mode, energy transfer remains strongly suppressed unless the interacting modes share identical spatial symmetry. A reduced two-mode model incorporating geometric nonlinearity shows that the intermodal coupling terms factorize into a single symmetry-determined modal-overlap integral, establishing spatial parity as the fundamental admissibility condition for nonlinear coherent energy exchange. These results demonstrate that symmetry, rather than spectral detuning alone, governs nonlinear modal coupling and introduce real-space nonlinear modal imaging as a general framework for controlling energy flow in nonlinear wave systems.

We demonstrate a paraffin-coated planar reflective multipass vapor cell for compact optical atomic magnetometry. The cell has an internal volume of $12 \times 12 \times 8~\mathrm{mm}^3$ and supports 20 optical passes with a total transmittance exceeding $65\%$, while maintaining a longitudinal spin-relaxation time of $^{87}\mathrm{Rb}$ longer than $1~\mathrm{s}$. The planar geometry provides spatially separated input and output beams, enabling compact optical integration. A single-cell free-induction-decay (FID) magnetometer reaches $10~\mathrm{pT}/\sqrt{\mathrm{Hz}}$ in the geomagnetic-field range, presently limited by current-source noise in the field coils. A two-cell differential configuration achieves a sensitivity of $\sim 28~\mathrm{fT}/\sqrt{\mathrm{Hz}}$ over the $1$--$15~\mathrm{Hz}$ band for bias fields of $0.3$--$0.7~\mathrm{G}$. These results establish that paraffin-coated planar multipass cells offer high optical depth, long coherence times, and an integration-friendly platform for ultrasensitive magnetometry.

Surfactant self-assembly at solid-liquid interfaces governs interfacial stability, transport, and reactivity across many technologies, yet resolving interfacial surfactant phases and their transition kinetics in situ remains challenging. Here, we establish an atomistically grounded plasmonic framework that quantitatively maps interfacial surfactant phases and phase transitions onto optical signatures. Distinct morphologies differ in packing and hydration, modifying the effective permittivity within the optical near field and producing surfactant phase-specific plasmonic extinction peak shifts. Using cetyltrimethylammonium bromide on silica as a prototypical surfactant-surface system, we combine atomistic simulations, electronic-structure calculations, and continuum electrodynamics to translate molecular morphologies into predicted spectral shifts for literature-reported surface phases. We experimentally confirm the predicted ordering and magnitude of steady-state peak shifts during stepwise concentration changes, and extract transition kinetics from exponential relaxations of the time-resolved peak shift. A key mechanistic signature is reversal of the spectral shift direction upon transition from an impermeable bilayer to a water-accessible, channel-containing phase, consistent with hydration-driven reduction of the local effective permittivity. Because the approach relies on dielectric contrast in the plasmonic near field and works through a dielectric overlayer, it provides a broadly applicable route for real-time identification of interfacial surfactant phases and their kinetics in aqueous conditions.

High-harmonic generation (HHG) in strongly correlated Mott insulators is investigated using exact diagonalization and time-dependent density-matrix propagation of a laser-driven one-dimensional Hubbard chain. By projecting onto equilibrium Hubbard bands, we use the doublon population and its dynamics as a diagnostic to analyze intraband (spin-wave-like) and interband (doublon-holon creation) excitation channels. A filling-dependent crossover emerges: Bloch-like intraband response at dilute filling, mixed dynamics at intermediate filling, and interband-dominated HHG with plateau and cutoff near half filling. In the considered parameter range, increasing interaction strength $U$ strongly suppresses interband contributions through the enlarged Mott gap and correlation-induced localization. Intra- and interband current decomposition reveals opposing flows below the Mott gap (Δ_{\mathrm{Mott}}) and selective dephasing suppression of interband coherence, enhancing net doublon accumulation. Time-frequency analysis uncovers the filling-dependent features of quantum trajectories, manifesting in distinct below-Δ_{\mathrm{Mott}} emission. This doublon-based analysis provides a transparent link between equilibrium spin-charge separation and nonequilibrium strong-field response, and clarifies how dephasing modifies interband coherence and doublon accumulation.

Traditional technical analysis indicators, although widely used by market participants, are often not sufficiently effective. We propose the Visibility Graphs Relative Strength Index (VGRSI), based on backward visibility relations in the price of a financial instrument. Rescaled to the 0--100 range, it can generate profitable trading signals. The performance of the indicator was evaluated using an automated trading strategy based on a 30-day optimisation window and a 7-day test window for three instruments representing different asset classes: DJI30, EUR/USD and XAU/USD over the 2024--2025 period (503 trading days). The strategy based on VGRSI signals generated a profit of USD~146,000 for DJI30, USD~69,000 for EUR/USD, and USD~125,000 for XAU/USD. This gives a total result of USD$\sim$340,000, which corresponds to an average profit of USD$\sim$676 per trading day, with a fixed investment of USD~1,000 to open a single trade. For all three assets, the strategy generated substantial profits while maintaining a moderate drawdown (10--18\% relative to a portfolio value of USD~10,000), a relatively low trading intensity (3.3--4.8 trades per day) and high Sharpe ratio values (2.55--3.6). These results indicate that VGRSI constitutes a promising technical analysis tool that goes beyond the classical trend-following approach by exploiting the geometric properties of asset price fluctuations.

Grating interferometry is a promising diagnostic technique that enables simultaneous acquisition of three complementary, synergistic X-ray images: transmission, differential phase, and dark-field. Its key advantage over other setups is its ability to use large pixels and, hence, large-area detectors, as well as its compatibility with low-coherence, compact X-ray sources, both of which are key factors for human-scale imaging. It has already demonstrated strong potential for chest imaging applications, including the diagnosis of pulmonary emphysema, fibrosis, and cancer. To retrieve transmission, differential phase, and dark-field images from data, an algorithm is required to separate the distinct mechanisms contributing to measured contrast. Since its realization, this image-retrieval step has remained fundamentally unchanged. In this work, we develop a novel transmission- and dark-field retrieval algorithm for grating-interferometry derived from the X-ray Fokker-Planck equation. To demonstrate and validate our Fokker-Planck algorithm, we apply it to experimental measurements of a test sample and to data from a mouse chest acquired with varying exposure times and added Poisson noise. The retrieved images were qualitatively and quantitatively compared with those retrieved using a conventional sinusoidal-fitting approach. Across both samples, the Fokker--Planck method produced images consistent with conventional retrieval, with a comparable signal-to-noise ratio. Notably, our Fokker-Planck method suppresses artefacts arising in the conventional approach under grating perturbations (e.g., structural defects like scratches) and reduced flux or visibility, yielding smoother and more reproducible images. Additionally, we demonstrate that our Fokker-Planck method has an advantage over the conventional dark-field retrieval method for fast sample imaging with short exposure times and high noise.

We require decomposition methods for the ABCD-matrix formulation in rotationally symmetric paraxial geometric optics when designing a multi-component optical system from a given single paraxial specification (represented by an ABCD matrix) to optimize non-paraxial specifications (e.g., optical aberrations). In this study, we propose two kinds of three-matrix decomposition of ABCD matrices by focusing on the fact that the ABCD matrices have three real-number degrees of freedom. In addition, we formulate a transformation between the two kinds of decomposition for a single matrix, which can increase or decrease the number of refraction surfaces in the optical configuration while keeping the paraxial specifications fixed. This nature is useful for the optical design of multi-component systems with optimized non-paraxial characteristics.

We present a unified refractive index (RI) sensing platform that integrates THz-comb-based frequency multiplication with dual-comb active-dummy temperature compensation. In conventional RI-sensing optical frequency combs (OFCs), sensitivity, stability, and measurement speed are fundamentally coupled, limiting overall performance. In the proposed system, RI-induced shifts in the repetition frequency are amplified in the terahertz domain, while temperature-induced fluctuations are suppressed through common-mode rejection in a dual-comb configuration. Experimental results demonstrate a sensitivity of 5.05 * 10^7 Hz/RIU, high linearity (R^2 = 0.9979), improved resolution (1.07 * 10^-4 RIU), and high accuracy (5.50 * 10^-5 RIU). The RI-induced frequency shift is expanded from tens of hertz to hundreds of kilohertz, enabling rapid and precise readout with short gate times. This approach overcomes the conventional trade-off between sensitivity and stability. More fundamentally, it establishes orthogonal control of signal scaling and noise suppression as a design principle for high-performance RI sensing.

Reliable real-time 3D shape sensing is essential for robust control and interpretation of deformable systems during motion. Existing vision-based approaches require line-of-sight and complex instrumentation, limiting operation in occluded and space-constrained settings. Here, we introduce a scalable, skin-like sensor that reconstructs its continuous 3D deformation in real time from distributed strain measurements. The device embeds a 2D array of mirror-stacked, printed oxidized eutectic gallium-indium (o-EGaIn) strain gauges within an elastomeric film to measure off-neutral-axis strains. Combined with a mechanics-informed observation model and a fast optimization routine, the system estimates local curvature, elongation, offset, and orientation under concurrent stretching, bending, and indentation, enabling reconstruction of complex surfaces. A 5-by-5 array with a 12 mm pitch achieves a mean surface reconstruction error of 0.62 mm with 0.1s latency across all tested scenarios. When conforming to complex surfaces, the sensor provides fast 3D shape mapping of the underlying geometry. Demonstrations involving palm gesturing, finger indentation, and contact-induced balloon deformation highlight utility for epidermal motion tracking, haptic interaction, and intraoperative monitoring.

Synchronized chaos has previously been predicted and observed in a small number (3) of mutually coupled lasers. In this work, we demonstrate that this phenomenon can theoretically persist in significantly broader scenarios, extending to complex coupled arrays of up to 11 lasers and arrays with finite built-in disorder. We quantify the resulting high-dimensional dynamics by computing Lyapunov spectra and the associated Lyapunov dimension, confirming that the observed states are chaotic rather than quasi-periodic. Furthermore, we uncover a regime of transient synchronized chaos where the system eventually escapes from perfectly synchronized chaotic state into an asynchronous state. We find that the lifetime of these transient states follows a bi-exponential distribution.

Ab initio wavefunction methods provide accurate molecular simulations but their computational scaling restricts applications to small systems. We develop a workflow combining quantum embedding to decompose a molecule into fragments with a heterogeneous quantum-classical (HQC) method to simulate fragments. We sample fragment electronic configurations on two 156-qubit quantum processors (ibm$\_$cleveland, ibm$\_$kobe), using up to 94 qubits, running 9,200 circuits for over 100 hours, collecting $1.3 \cdot 10^9$ measurement outcomes - the most resource-intensive HQC computation for quantum chemistry to date. We compute fragment wavefunctions via optimized subspace diagonalization on two supercomputers (Fugaku, Miyabi-G), achieving 72.5$\%$ parallel efficiency with scalable distributed linear algebra kernels. We simulate two protein-ligand complexes spanning dispersion- and electrostatics-dominated regimes (11,608 and 12,635 atoms), demonstrate $>40\times$ increase in system size and up to $210\times$ improvement in accuracy over the previous state-of-the-art, with HQC matching coupled-cluster (CCSD) accuracy in fragment energies, and establish a scalable pathway for systematically improvable biomolecular simulations.

Rapid and controllable reduction of graphene oxide (GO) remains a critical challenge for realizing its full technological potential. Here, we report efficient reduction of GO by a synergistic electron-beam-assisted single-pulse near-infrared (NIR) laser process. Time-resolved electron energy-loss spectroscopy measured with a dynamic transmission electron microscope (DTEM) is used to locally track the oxygen concentration evolution after NIR laser pulse irradiation. This finds an oxygen diffusivity of 1.6 +/- 0.4 x 10$^{-8}$ m$^2$/s, which corresponds to 90% reduction of a 46-nm thick film within 960 ns. Electron beam irradiation is found to change the optical absorptivity of GO in the NIR region and the thermal heating cycle resulting from the laser pulse is simulated. Structural characterization via selected-area electron diffraction (SAED) and high-resolution transmission electron microscopy (HRTEM) finds localized restoration of sp$^2$ bonding accompanied by turbostatic disorder in the reduced GO. Together, these results point to a mechanism involving the creation of defects and vacancies produced by electron beam irradiation, which increases the efficiency of NIR light absorption and oxygen diffusion normal to the layers. This study demonstrates the important role of such defects in controlling the photochemistry of GO and its response to NIR illumination.

Full-f global gyrokinetic simulations of the plasma boundary have until now required heroic computational efforts and case-by-case expert intervention, precluding systematic parameter scans. Here we demonstrate a paradigm shift: hundreds of independent, concurrent, and unsupervised full-f boundary gyrokinetic simulations in a geometry inspired by the Tokamak à Configuration Variable (TCV), covering both the closed flux surface region and the open-field-line scrape-off layer (SOL) while scanning triangularity, elongation, and heating power. All simulations are evolved much longer than the turbulence relaxation time until the steady state is reached. Analysis of the steady-state profiles reveals that the impact of plasma shaping on confinement is strongly power dependent: at low power, triangularity primarily controls the SOL ion temperature, while at high power it mostly affects the edge ion temperature gradient. The low-power hot SOL observed for positive triangularity is explained by a neoclassical trapped-ion mechanism in which triangularity modifies the field-line arc length between banana turning points and the high-field-side limiter, altering the interaction with cold neutral-ionization regions. Fingerprint analysis of turbulent transport categorize the simulations in a regime dominated by ion temperature gradient (ITG) or trapped electron modes (TEMs), confirmed by dedicated local linear gyrokinetic calculations. The generated open data represents a previously unobtainable resource. It can serve both as a benchmark for boundary transport models, and as a training dataset for data-driven methods in fusion foundation and surrogate models.

Excited-state electronic structure in strongly correlated systems remains challenging due to the exponential scaling of the many-body Hilbert space and the difficulty of constructing systematically controlled active spaces. Building on the stochastic cluster expansion (SCE) framework previously developed for ground-state correlation energies, we extend the formalism to excitation gaps by expressing energy differences directly as a hierarchy of orbital-space cluster contributions. In this formulation, excitation energies are reconstructed from reduced-rank calculations involving a minimal frontier chemical subspace (FCS), treated exactly, together with stochastic sampling of the remaining orbital environment. This approach eliminates the need for large or chemically preselected active spaces. We demonstrate the method on charge-transfer complexes and polyacenes, where accurate singlet-triplet gaps are obtained that agree with full-system results. The method converges with low-order cluster terms and provides a systematically improvable framework for excited states in correlated systems.

Chalcogenides have recently emerged as an important class of phase-change materials (PCMs) for nanophotonics, owing to their very high refractive index (RI) and low optical loss in the visible to near-infrared range. They exhibit an ultralarge RI change (> 0.7) upon phase transition, which can be triggered by multiple stimuli such as electrical bias, laser illumination or thermal heating. These properties make them highly appealing materials for flat optics and metasurface applications. Current nanophotonic implementations of chalcogenide PCMs mostly rely on two-dimensional (2D) or quasi three-dimensional (3D) thin film patterning based on the coating of chalcogenide materials from a solid-state target. This limits fast prototyping of 3D freeform micro- and nanostructures, thus restricting geometric design freedom and device functionality. Here, we demonstrate a solution-phase direct printing of chalcogenide PCMs into functional structures. The method is based on dip in two photon-induced solidification (DITPS) of a specially synthesized antimony trisulfide (Sb2S3) precursor solution. Direct printing with DITPS is simple, maskless, fast and cost effective, enabling true freeform 3D printing of photonic devices with sub micron resolution. We show direct writing of Sb2S3 helices with different wire cross section profiles on gold and ITO substrates, as well as functional planar Fresnel zone plates (FZPs) and computer generated hologram metasurfaces (CGHMs) in a single printing step. This freeform DITPS approach thus enables rapid 3D prototyping of high index metasurfaces and opens a route to integrating high-index PCMs into existing photonic architectures and device platforms.

We show that time-reversed Young interferometry reorganizes, rather than reverses, optical entropy. A fixed detector conditions the reciprocal source--detector Green function and produces a source-label probability distribution. Marginal entropies in the standard and time-reversed geometries are generally unequal; the reciprocal invariant is instead the mutual information between source and detector coordinates. Near a destructive response, the conditioned source-label entropy can decrease while Fisher information for small phase, tilt, or defocus perturbations increases. The result identifies time-reversed Young interferometry as a source-space information processor with no analogue in ordinary detector-plane fringe readout.

Tree tensor network states (TTNSs) combined with the density matrix renormalization group (DMRG) are emerging as powerful tools for vibrational and vibronic structure simulations in molecules with strong coupling and fluxionality. In this Perspective, we discuss how TTNS methods enable accurate, full-dimensional computations of thousands of eigenstates for molecular systems ranging from quartic-force-field benchmarks to molecules with strong vibronic coupling and protonated water clusters as large as the 33-dimensional Eigen ion, H$_3$O$^+$$\cdot$(H$_2$O)$_3$. We emphasize the close connection and interoperability between DMRG-based TTNS methods and the multilayer multiconfiguration time-dependent Hartree method (ML-MCTDH), which share the same underlying ansatz. We also highlight practical challenges of predictive simulations, including robust error estimation, convergence of observables such as infrared intensities, and optimization of tensor network tree structures. Finally, we outline recent advances toward direct targeting of excited states and discuss opportunities for broader applications in molecular spectroscopy and quantum dynamics.

As the AAS Working Group on Graduate Admissions (WGGA) we are sharing brief recommendations for improving and standardizing key elements of the graduate admissions process in astronomy. Most astronomy graduate programs have large areas of overlap in their admissions processes; however, the existing small variations in requirements and mismatches in communication and transparency make admissions more challenging for students and programs alike. To improve this situation, and building on the work presented in the AAS Graduate Admissions Task Force (GATF) report we recommend a few simple and straightforward changes for application content, communication, and timelines. These include an application format that consists of 1) two 500-word recommendation letters, 2) one 1500-word application essay, 3) an applicant CV, and 4) unofficial transcripts; and an admissions timeline that includes effective and transparent communication from programs and encouraging an April 1st "down-select date" for applicants.

Complex contagions describe systems where the probability or rate of contagious transmission is a nonlinear function of the exposure to contagious agents. These models were first studied theoretically but have since been used to capture effects such as nonconformism, social reinforcement or peer pressure in empirical data. However, recent studies have shown that local correlations (e.g., group structure or temporal burstiness) and heterogeneity (e.g., diversity of parameters or covariates) can give the illusion of nonlinear effects even when the dynamics is actually linear. We briefly review these studies to inform a new model and explanation for these effective models of complex contagions. We find global threshold dynamics and superlinear complex contagions even in populations where agents are distributed across social groups described solely by linear or even sublinear contagions. This effect can be understood as a manifestation of Simpson's paradox. Incidence data from heterogeneous groups can look superlinear once averaged over all groups, since the sampling of groups represented at high incidence is biased towards those with stronger local transmission. We then define what we call a Simpson's contagion: a contagion process that looks superlinear when observed over an entire population, but is mechanistically linear or even sublinear in all of its subgroups. By exploring these Simpson's contagions over mathematical case studies, our work contributes to the growing body of literature on the ubiquity of threshold and complex contagions as effective models, and our results stress the pitfall of model selection that ignores correlations and heterogeneity in populations.

Characterizing quantum many-body phase structure is a major goal for quantum simulation. Here, we employ an unsupervised learning approach based on diffusion maps to learn phase transitions in bosonic lattice systems described by Bose-Hubbard type models, which can be realized in ultracold atoms and related quantum simulation platforms. We demonstrate that this approach identifies phase structure across distinct settings without prior knowledge of order parameters or handcrafted observables, including ground-state transitions involving symmetry-protected topological phases and nonequilibrium regimes distinguishing ergodic and many-body localized behavior. Our results indicate that the approach has the potential for direct application to experimentally accessible measurement data for learning quantum phases in current quantum simulators.

Information sharing between individuals is crucial to improve performance in collective tasks. However, in a competitive world, individuals may be reluctant to share information with the others, and it is still unclear how the presence of strategic behaviors affects the collective performance of a group. In this study, we introduce an evolutionary game modeling the dynamics of individual behaviors in a collective estimation task. The individuals are organized in a network and have to guess the distribution of ball colors in a box. Each of them samples a given number of balls and can strategically decide whether to share or not this information with its neighbors. We develop a framework that allows to investigate analytically how the collective performance depends on the network structure. We find that the optimal network results from a trade-off between the sharing rate and the way the information is integrated in the network. We further reveal that there exists an intermediate average degree for each type of network maximizing the collective performance. In addition to the uniform case, we consider the case of non-homogeneous allocations of the number of individual samples, showing that the largest collective performance is obtained when the number of ball extracted by an individual is inversely proportional to its degree.

We report and analyse the classification performance of an experimental hybrid spatio-temporal photonic reservoir computer based upon a free-space VCSEL array. We demonstrate experimentally the enhancement of spatial-only reservoir operation, featuring the diffractive coupling of lasers in an external cavity, by exploiting up to 88 virtual nodes with time multiplexing. We analyse the dependance of performance on the spatial and virtual node number, and achieve an improvement for both spatial- and temporal-only reservoirs with a reduced test error of 0.026 in a classification task. Further, given the high bandwidth of the non-linear laser transformation, we demonstrate the expansion of a 12 spatial node network to a 968 node network, operating at an input time of 17.6ns, maintaining high processing speed and improving network scalability and performance.

We propose a laser isotope separation mechanism in which selectivity arises from nuclear spin rather than isotope shifts, enabling the use of broadband ultrafast lasers. A Ramsey pulse sequence is applied to paramagnetic molecular isotopologues possessing two electronic states coupled by a dipole transition. For even-even isotopologues (nuclear spin $I = 0$), each electronic state is a single level and the time-reversed sequence returns all population to the ground state exactly. For even-odd isotopologues ($I > 0$), the hyperfine interaction splits each state into multiple levels with coupling amplitudes set by Wigner $6j$ symbols; incommensurate phase evolution during the dark interval prevents the echo from closing, trapping a fraction $P_m$ of the population in the excited manifold. In the impulsive limit ($Ω\gg A_{\rm HF}$), $P_m$ depends only on the angular momentum quantum numbers $(J_g, J_m, I)$ and is independent of laser intensity or bandwidth. Density matrix simulations confirm $P_m = 0$ for $I = 0$ and $P_m \approx 0.23$-$0.47$ for $I > 0$ across representative systems including ${}^{235}$U, ${}^{87}$Sr, and ${}^{57}$Fe. Under realistic collisional conditions, single-pass enrichment exceeding 90% from natural feed is achievable without cascading.

Diamond quantum sensors offer high precision and spatial resolution as magnetic probes, making them promising for a wide range of applications. While diamond anvil cells (DACs) can generate extremely high pressures, techniques for magnetometry under such conditions remain limited. By fabricating an ensemble of NV centers directly on the anvil diamond surface, we enable precise magnetic measurements under high pressure. In this work, we employ this NV ensemble to image the stray magnetic field of iron up to 30 GPa, enabling the observation of the magnetic transition ($α$-$\varepsilon$ transition) in iron.

Second-harmonic generation microscopy is a powerful technique capable of probing local crystal symmetries and electric fields at interfaces. However, it often suffers from weak signal strength and is difficult to understand in multilayer systems where many materials can give competing signal contributions. In this work we present direct observation of delocalized, surface plasmon polariton-mediated second-harmonic generation on gold monocrystalline surfaces and structures. We generate second-harmonic light up to 35 um from the excitation spot and, excitingly, we obtain signal from atomically flat surfaces without a fundamental excitation beam present in the same region. We reveal that this process arises from the interaction of two counter-propagating surface plasmon polaritons, which we believe to be the first observation of this process at the microscale. This signal has the same polarisation dependence as localised second-harmonic generation and is emitted in a collimated beam travelling perpendicular to the sample surface. In part due to local electric field enhancements, we were able to observe these signals on a CMOS camera with 1 s exposure and no gain using an industrial-grade pulsed laser. Our results enable wide area multilayer samples to be probed using a single excitation beam, with applications including in energy, catalysis and single particle surface sensing.

The nuclear charge radius is one of the most fundamental quantities of the atomic nucleus. It can be deduced from a combination of experimental measurements of muonicX-raytransitionenergieswithmodellingofthoseX-raytransitionenergies. In thisworkwepresentMuDirac (1.3.0), whichisanopen, publiclyavailable, sustainable and computationally efficient software tool that will be at put the disposal of the negative muon community. With MuDirac (1.3.0), the community will be able to accurately and efficiently estimate nuclear properties, such as the nuclear charge radius, by assuming a 2-parameter Fermi distribution of the nuclear charge.

While vertical GaN-on-silicon architectures promise a transformative leap in cost-effective power electronics and high-resolution micro-LEDs, their deployment remains bottlenecked by the high electrical resistance of conventional epitaxial buffer layers. Here, a universal and straightforward sputtering-based strategy is presented to realize high quality GaN epitaxial films on Si(111) substrates characterized by exceptionally low vertical resistance, ohmic behavior, and robust thermal stability. This technique centers on the in-situ formation of a sub nanometer (0.5 nm) silicide-based template via rapid thermal annealing method demonstrating unprecedented versatility across 25 different metallic species. Scanning transmission electron microscopy (STEM) reveals that a unique amorphous like interlayer (AL-IL) effectively accommodates lattice mismatch and relaxes epitaxial strain. These AL-IL templates further serve as high performance platforms for metalorganic chemical vapor deposition (MOCVD) overgrowth, successfully bridging the gap between scalable, low-cost fabrication and device-grade vertical performance.

Detailed-balance limits for transition-metal dichalcogenide (TMD) solar cells have been reported, but existing TMD-specific limits do not simultaneously resolve thickness-dependent optics, carrier multiplication (CM), hot-carrier (HC) extraction, and finite cooling leakage. Here, we develop a generalized detailed-balance theory that provides an upper-bound framework. The model combines energy- and thickness-dependent absorptance a(E,d), exciton-resolved monolayer absorbance, an experimentally available CM quantum-yield limit (eta_CM <= 0.97), and an endoreversible HC engine with ideal energy-selective contacts and finite heat-leak coefficient kappa. The framework shows that CM and HC draw on the same above-gap photon-energy reservoir; therefore, CM does not raise the reversible HC thermodynamic limit. Instead, CM can protect finite-kappa performance only by shifting excess-energy utilization from a cooling-sensitive voltage channel into collected current. For optically thick TMDs under AM1.5G illumination, the SQ optimum lies near E_g = 1.3 eV, whereas the CM/HC-favored envelope shifts toward E_g = 1.0 eV with reversible efficiencies above 50%. For monolayer TMDs such as WSe2 (E_g = 1.63 eV), CM is essentially inactive because only about 3.7% of above-gap AM1.5G photons satisfy E > 2E_g, giving an idealized short-circuit-current gain of only about 0.6% before device nonidealities. Bulk-like TMDs can show large HC-related gains at d = 10-50 nm, but even kappa = 0.2 W m^-2 K^-1 implies about 100 W m^-2 heat leak for Delta T = 500 K. Thus, high-E_g monolayer TMDs are not promising one-sun CM candidates, whereas narrow-E_g, bulk-like TMD absorbers remain plausible beyond-SQ candidates only if energy-selective extraction and phonon-engineered cooling suppression are realized together.

Achieving long-term stability in high-capacity lithium-ion battery anodes remains a critical challenge. In this study, we present a materials-intrinsic strategy for extending the cycle life of Ge, a promising next-generation anode material, through trace doping with metal elements. We systematically investigated the effects of small additions of various metals and found that elements with large atomic size, particularly Yb, markedly improved the cycling stability without sacrificing the initial capacity, while appropriate Yb doping enhanced the anode lifetime by approximately a factor of three. Structural and electrochemical analyses revealed that this improvement originates from mechanical softening of the Ge anode, which suppresses lithiation-induced damage such as cracking and delamination. Nanoindentation measurements further showed a strong negative correlation between dopant atomic size and film hardness, establishing anode softening as a new design principle for damage-tolerant electrodes. Although Yb doping reduced the rate capability at high C-rates, the present results demonstrate a clear shift in design strategy from volume-change suppression to mechanical compliance. These findings provide a useful framework for stabilizing high-capacity alloy anodes through atomic-scale mechanical control.

Integrated circuits are a cornerstone of modern information technology, and analog wave-based architectures could enable fast and efficient processing beyond conventional charge electronics. In magnonics, spin waves provide a highly tunable, compact and energy-efficient medium for on-chip microwave signal transport and processing. However, progress has been limited to isolated elements or short devices, severely limiting the overall functional complexity and scalability. Here we realize the key elements of universal magnonic circuitry, using a single-step direct laser writing process in yttrium iron garnet, and monolithically cascade them in multi-stage programmable devices and networks. Using magneto-optical Kerr effect microscopy, we show efficient spin-wave propagation and preserved phase coherence in waveguide structures for hundreds of wavelengths. In coupled waveguides, we observe complete and periodic power transfer over several coupling lengths, and in phase shifters we achieve arbitrary, tunable phase delays. By cascading these elements, we realize programmable splitters, frequency demultiplexers, and phase-controlled 2x2 routers, where output power and relative phase can be programmed on demand via external fields. Finally, we realize programmable magnonic interferometric meshes for on-chip radio-frequency signal routing, with up to six magnonic inputs and outputs and seven cascaded stages, without the need for intermediate amplification. These direct-write cascaded networks bridge a long-standing gap in magnonic scalability, offering a viable pathway toward integrated, large-scale architectures for both classical and quantum processing.

Controlled experimental studies of percolation are challenging due to difficulties in tuning site connectivity, isolating local interactions, and mitigating finite-size effects. In this work, we experimentally investigate percolation with a platform of coupled lasers, where connectivity, interaction strength, and system size can be controlled. Using a square array of 100 lasers with astronomical number of possible cluster configurations, we show that the emergence of a percolating cluster corresponds to the onset of phase locking among the lasers. We also show that the percolation probability undergoes a second-order alike transition as a function of the site-occupation probability, with a threshold consistent with classical theoretical predictions. Surprisingly, we find that at low pump level, amplified mode competition (nonlinear regime) alters the effective behavior of the lasing sites and modify the nature of the percolation transition. The experimental results are interpreted by the means of a theoretical toy model with connectivity rules to the classical percolation.

We develop a structure-preserving solution framework for the optimal control of the time-dependent Maxwell's equations. Building on a well-posedness theory for a weak form of the forward problem, we first analyze a forward solver that couples Nédélec and Raviart--Thomas finite elements with Crank--Nicolson time stepping. The solver preserves the de~Rham structure, enforces a discrete Gauss law, exactly satisfies a per-time-step energy balance, and converges to the weak solution under low regularity assumptions on the problem data, which are dictated by the optimal control setting. To control the Maxwell system, we add the curl of a space-time current density as a source to Ampére's law. The curl form yields charge conservation without auxiliary constraints. We prove the well-posedness and continuity of the control-to-state map, derive the adjoint system and a gradient representation for a tracking-type objective functional, and formulate a discrete optimization scheme that inherits structure preservation from the forward solver. Our discrete stationarity conditions are consistent with their continuous counterparts, and the discrete optimal controls converge, with mesh and time refinements, to the continuous optima. We demonstrate the merits of our optimal control formulation and the theoretical developments by numerically solving a series of source-cloaking model problems.

Reliable parameter extraction from experimental data is central to quantitative analysis in spectroscopy, diffraction, photoluminescence, chromatography, microscopy, and time-resolved measurements. We present FitED, a Python-based desktop application for interactive and automated nonlinear fitting of one-dimensional scientific data. FitED combines an accessible graphical workflow with a numerical backend capable of fitting both conventional peak profiles and arbitrary user-defined analytical functions. The software supports Gaussian, Lorentzian, Pseudo-Voigt, and exact area-normalized Voigt profiles, together with custom functions such as exponential decays, stretched exponentials, saturation curves, and spectroscopy-specific response functions. It integrates robust text-file import, region-of-interest selection, background modeling, parameter bounds, weighting strategies, automated pre-fit search, iterative peak refinement, residual visualization, session persistence, and structured export of fitted curves, components, reports, and metadata. By combining mathematical transparency with practical usability, FitED aims to make nonlinear fitting more reproducible and accessible while preserving the parameter-level control required by experienced experimental researchers.

Treatment-related lymphopenia is a frequent and clinically significant consequence of cancer therapy that can compromise immune-mediated tumor control and worsen patient outcomes. Despite its importance, no mechanistic framework exists to accurately predict the severity of lymphopenia from patient-specific data. Here, we present a biokinetic model that quantitatively describes lymphocyte depletion and recovery during and after radiotherapy, integrating radiation dose-volume distributions, blood circulation dynamics, and distinct kinetics of fast- and slow-recovering lymphocyte populations. The model was calibrated and validated using 56 independent clinical datasets encompassing various tumor sites and treatment modalities. It reproduces observed lymphocyte counts and enables prediction of individual severity of lymphopenia from baseline or early-treatment counts. Applying this framework, we demonstrate that particle therapy reduces lymphocyte depletion by ~30% compared with photon therapy, providing a quantitative explanation for its observed immune-sparing benefit. By linking radiation physics, immune kinetics, and clinical outcomes, our model establishes a mechanistically grounded predictive approach for anticipating systemic immune toxicity. Beyond radiotherapy, this framework offers a generalizable strategy for integrating early biological markers into treatment optimization, advancing personalized and immune-preserving cancer therapy.

Urban planners need up-to-date, global, and consistent street network models and indicators to measure resilience and performance, model accessibility, and target local quality-of-life interventions. This article presents up-to-date street network models and indicators for every urban area in the world. It uses 2025 urban area boundaries from the Global Human Settlement Layer, allowing users to join these data to hundreds of other urban attributes. Its workflow ingests 180 million OpenStreetMap nodes and 360 million OpenStreetMap edges across 10,351 urban areas in 189 countries. The code, models, and indicators are publicly available for reuse. These resources unlock worldwide urban street network science beyond samples as well as local analyses in under-resourced regions where models and indicators are otherwise less-accessible.

We present a minimal agent-based model of interacting agents characterized by their wealth to study taxation and inequality in a non-conservative economy. Wealth evolves through an extremal stochastic replacement process in which the poorest agent has its wealth replaced by a new random value, financed through a collective taxation mechanism. We explore taxation regimes ranging from regressive to progressive schemes and tune the overall redistribution strength. Under regressive taxation, the system self-organizes into two distinct stationary phases when changing the total tax collected: a non-ergodic, high-inequality regime characterized by wealth condensation in a subset of agents that permanently escape replacement, and a more homogeneous ergodic phase in which all agents participate in the dynamics. Increasing taxes drives an abrupt transition between these phases. The transition is discontinuous and exhibits hysteresis and bistability, consistently detected through the Gini index, the Top $1\%$ wealth share, the entropy, and the Binder cumulant. In contrast, neutral and progressive taxation suppress persistent wealth concentration, preventing the emergence of strongly unequal states and eliminating hysteretic behavior. These results show that minimal stochastic redistribution mechanisms alone can produce discontinuous transitions, metastability, and non-ergodicity, demonstrating that taxation structure can determine the emergence and stability of macroscopic inequality.

In biological systems, neural circuits compute through directed, short-latency interactions whose effects unfold across multiple time scales and behavioral contexts. We address the problem of inferring these local, lag-specific interactions from sampled neural population activity under varying stimuli, without assuming a parametric form for the underlying dynamics. Our approach leverages denoising score models by estimating joint-window scores over consecutive activity snapshots (i.e., brain states) and converting these scores into calibrated, directed edge tests via cross-block score products. The key insight is that these products recover the Jacobian of the transition map between brain states under nonlinear dynamics. To cleanly separate lag-specific effects, we introduce minimal multi-block windows that condition on intermediate time points, avoiding the omitted-lag bias inherent in pairwise analyses. The resulting method, Score--Block Time Graphs (SBTG), identifies lag-specific directed interactions in sampled neuronal population data. We specifically apply SBTG to whole-brain C. elegans calcium imaging data to recover lag-specific circuit structure not resolved by current methods, including improved alignment with independent connectomes, cell-type-specific temporal organization, and neuromodulatory profiles consistent with known receptor kinetics. These findings highlight the potential for SBTG to serve as a practical ``AI for science'' tool by turning high-dimensional neural population recordings into statistically testable circuit hypotheses.

This paper introduces an extension of generalised filtering for online applications. Generalised filtering refers to data assimilation schemes that jointly infer latent states, learn unknown model parameters, and estimate uncertainty in an integrated framework -- e.g., estimate state and observation noise -- at the same time (i.e., triple estimation). This framework appears across disciplines under different names, including variational Kalman-Bucy filtering in engineering, generalised predictive coding in neuroscience, and Dynamic Expectation Maximisation (DEM) in time-series analysis. Here, we specialise DEM for ``online'' data assimilation, through a separation of temporal scales. We describe the variational principles and procedures that allow one to assimilate data in a way that allows for a slow updating of parameters and precisions, which contextualise fast Bayesian belief updating about the dynamic hidden states. Using numerical studies, we demonstrate the validity of online DEM (ODEM) using a non-linear -- and potentially chaotic -- generative model, to show that the ODEM scheme can track the latent states of the generative process, even when its functional form differs fundamentally from the dynamics of the generative model. Framed from a neuro-mimetic predictive coding perspective, ODEM offers a biologically inspired solution to online inference, learning, and uncertainty estimation in dynamic environments.

We establish a genealogical framework for an existing analytical moment duality between a Wright--Fisher type SDE and a counting process with interaction. To achieve this, we construct a finite-population Moran model featuring interactive neutral reproduction as a novel mechanism. In the corresponding events, an individual, regardless of its own type, can only reproduce if a randomly encountered partner is of the ``fit'' type. This Moran model has a relatively simple counting process as its factorial moment dual, whose genealogical meaning appears to be cryptic: after all, the line-counting process of the natural genealogical process of the model, namely the ancestral influence graph (AIG), exhibits a complex hierarchical structure not reflected in the factorial moment dual. Since moment duality is a property in expectation, we are allowed to systematically remove information from the AIG and merge different realizations of the ancestry. We call the result the \emph{Frankenstein process}. Based on this, we establish the factorial moment duality from a genealogical perspective. The moment duality in the diffusion limit follows in a natural way.

In this work, we present a mathematical model for cyclic and sequential patterns of brain activity, combining heteroclinic dynamics with discrete neural-field models. We first show that spatial-discrete neural-field equations with biologically realistic equilibria cannot support heteroclinic cycles. On the other hand, heterocline dynamics often arise in Lotka-Volterra-type systems, but these equations do not directly correspond to neuronal processes. To address this, we use a version of the Universal Approximation Theorem to approximate any target dynamics by a neural network interpretable as a high-dimensional Amari-type neural-field system. When the target dynamics contains a heteroclinic cycle, the approximating vector field generates a periodic trajectory that closely follows the heteroclinic connection. As a case study, we consider the cognitive processes underlying focused-attention meditation. We show how the model reproduces sequential transitions among cognitive states and we conclude providing a neural interpretation of the approximating dynamics.

Quantitative susceptibility mapping (QSM) has been increasingly applied in longitudinal studies of neurodegenerative diseases and aging to assess temporal alterations in brain iron and myelin. The accuracy of such investigations depends on the repeatability and sensitivity of measurements. However, the ill-posed nature of the QSM processing steps makes the reconstruction vulnerable to background field changes, head orientation changes, noise, and imperfect registration, which compromise repeatability and sensitivity and hinder reliable detection of true changes. To address these limitations, we propose Longitudinal QSM, a simultaneous reconstruction framework that jointly estimates susceptibility maps across time points while enforcing spatial sparsity of temporal changes. The method was evaluated through simulations and in-vivo experiments and compared with conventional reconstruction methods. Longitudinal QSM consistently reduced inter-scan variability and accurately recovered simulated lesion changes. Application to stroke patient and multiple sclerosis patient data further demonstrated that the framework stabilizes non-lesion variability while preserving lesion-related temporal changes. This approach offers a promising tool for monitoring subtle temporal changes in brain iron and myelin in various neurodegenerative diseases as well as throughout aging and development.

We extend the $N$ branching Brownian motions model of population invasion to higher-order asexual reproduction. Increasing reproduction order leads to qualitative changes: invasion fronts generically cease to exist beyond binary reproduction; and in the binary case itself, their speed becomes diffusion-independent. Ternary reproduction shows critical behavior, with collapse into a strongly localized `invasion bullet' in the supercritical regime, diffusive spreading in the subcritical regime, and a continuous family of fronts at criticality. These results suggest that the dominance of division and binary reproduction in nature reflects fundamental constraints on invasion dynamics.

Regeneration of the nervous system after injury remains an important therapeutic objective, especially in the central nervous system (CNS), in which regeneration is restricted by both neuronal limitations as well as adverse extracellular environments. Conversely, the peripheral nervous system (PNS) displays enhanced regenerative capability in the presence of supportive Schwann cells (SC) and pro-growth stimuli. While the structure and molecular mechanisms are thoroughly understood, functional biomarkers that can non-invasively monitor regeneration in real time are limited. In this review, we discuss the promise of electroencephalography (EEG) as well as electromyography (EMG) as real-time, non-invasive biomarkers to monitor damage to nerves and regeneration in both CNS and PNS contexts. First, we contrast biological and electrophysiological indicators of CNS/PNS injury, showing how EEG signs, including oscillatory power, connectivity, and evoked potential changes, reflect dysfunction due to injury as well as neuroplastic reorganization. Also, EMG provides direct insight into muscle activation and peripheral output, providing useful EEG complementation in neuromuscular pathway integrity and reactivation. In CNS injuries (e.g., stroke, spinal cord injury (SCI)), EEG typically shows global slowing, disrupted interhemispheric coherence, and partial recovery of higher frequencies. For PNS injuries, EEG can capture cortical remapping and return of somatosensory evoked responses with re-establishment of the peripheries' connectivity. EMG, in turn, enables monitoring of reinnervation and restoration of functional motor output. This review presents a dual-system perspective, positioning EEG and EMG not only as diagnostic tools but also as functional biomarkers of neural regeneration, thereby bridging electrophysiology, plasticity, and clinical recovery.

Human societies continuously transform scattered information into collective judgments and coordinated action, whether through markets discovering prices, governments allocating resources, communities enforcing norms, or science converging on reliable claims. Importantly, the computational difficulty of collective decision-making, particularly the time and communication required to reach solutions, imposes fundamental constraints on social organization. While theoretical computer science offers formal tools for analyzing such problems, for instance, by analyzing resource requirements, including time and memory, surprisingly, there is no domain of social science that focuses on the nature of computation in the human world. This perspective argues that we now have the opportunity to deploy these computational frameworks to study human social organization, opening research directions at the intersection of computer science and social science. We highlight core social phenomena that can be framed as computational, including (i) distributed consensus and coordinated action, (ii) societal restructuring with scale, (iii) hierarchical and modular structure, and (iv) externalized memory systems. We identify several concepts from theoretical computer science that may provide insight into these phenomena, especially emphasizing more recently developed approaches beyond the paradigm of Turing~Machines and worst-case computational complexity.

How do we measure genuine understanding in artificial cognitive systems? Current approaches face a measurement gap: probabilistic systems refine confidence gradually, practice-based systems compile knowledge through repeated execution, and neural systems distribute understanding across opaque embedding spaces. We propose that making understanding measurable requires architectures where understanding formation produces discrete, inspectable structural signatures. This paper presents hierarchical automata built from finite state machines representing patterns and higher-order automata representing compositions. Constrained inference constructs automata from single observations. Similarity detection clusters related automata, making concept robustness quantifiable. Graph memory makes compositional knowledge directly inspectable. Metacognitive mechanisms enable observable reconfiguration. We demonstrate understanding measurement in a simple geometric domain. Graph evolution tracking reveals five measurable signatures: immediate representation formation, structural knowledge, generalization capacity, compositional awareness, and metacognitive access. These measurements distinguish structural understanding from statistical correlation. Our contribution is a framework for making understanding measurable through discrete compositional knowledge structures. This measurement capability complements perceptual learning in neural systems and task execution in neurosymbolic architectures.

Indoor lighting affects cognition, affect, and behavioural regulation, but these effects are often treated as isolated findings rather than as parts of a unified process. This paper proposes an active inference account of shared indoor lighting in multi-user environments such as offices, classrooms, and libraries. It argues that lighting shapes behaviour through three distinct channels: illuminance modulates perceptual precision, correlated colour temperature modulates arousal relative to circadian optimum, and spectral composition biases behavioural disposition toward engagement or rest. The paper formalises this hypothesis through a proof-of-concept POMDP model of agents performing sustained reading over five hours, using both reading performance and eye-tracking observations. The model generates six falsifiable predictions, all confirmed across 20 Monte Carlo simulations.

Phylogenetic trees are the primary framework for conveying evolutionary relationships. While many tools exist for visualizing phylogenetic trees, most are limited to static graphics, require coding expertise, or are developed for a specific website and not easily reusable or extensible. To address these limitations, we developed heat-tree, a collection of software packages in JavaScript, R, and Python for interactive visualization, manipulation, and editing of phylogenetic trees and their associated metadata. Heat-tree allows for the creation of customizable, web-compatible tree visualizations that can be easily embedded in R Markdown, Jupyter Notebooks, and Quarto documents, as well as directly in HTML/JavaScript applications and websites. The package supports radial and rectangular tree layouts, automated translation of metadata values into visual encodings on the tree, interactive tree editing, and export capabilities for publication-quality figures. All visualization parameters are definable programmatically or interactively using the comprehensive graphical user interface included with each visualization. Heat-tree was designed to be a user-friendly software package for interactive tree viewing, manipulation, editing, and self-contained, embeddable visualization across software environments.

Hill functions, the standard tool for modelling gene regulatory networks, carry three structural flaws when the cooperativity exponent is non-integer: loss of global smoothness, silent complex-valued arithmetic corruption of ODE trajectories, and an identically zero basal production rate that traps bistable models in off-states. Logistic functions $f^\pm$, being globally $C^\infty$, real-valued for all arguments, and strictly positive at zero, resolve all three simultaneously. For a two-gene negative-feedback oscillator, local asymptotic stability is established for all positive parameters via the Routh--Hurwitz criterion, and no Hopf bifurcation is possible without time delays. For bistable positive autoregulation, saddle-node thresholds are characterised through explicit transcendental equations; with biophysically grounded \textit{E.~coli} parameters, basal logistic production drives off-state escape in $\approx 44$~min while the Hill model remains permanently trapped. The 11-gene Traynard cell-cycle Boolean network is translated automatically via the product-of-logistics De~Morgan formalism and integrated without warnings, all variables remaining bounded and non-negative. The De~Morgan framework places every repressor threshold at a positive measurable concentration, whereas the weighted-sum formulation of Samuilik et al.\ places repressor critical points at negative concentrations, rendering them biologically inert. On an 80-gene Boolean-derived ODE system with $n = 3.509$, the Hill solver entered silent complex-valued contamination at $t \approx 52.64$ and terminated near $t \approx 63$--$65$; the logistic formulation completed $t \in [0, 200]$ without a single warning. The always-positive production rate ensures full controllability, enabling sliding mode, model predictive, and feedback-linearisation strategies where Hill-based formulations fail.

Identifying transient high-activity episodes in astronomical time series requires partitioning data into regions of distinct statistical behavior. A widely adopted approach combines Bayesian Blocks with a hill-climbing procedure to isolate high-activity regions, but carries $\mathcal{O}(N^2)$ complexity -- a scalability challenge for wide-field surveys like ZTF and the upcoming Rubin Observatory (LSST), where light curves routinely contain thousands of irregularly sampled observations. We present Peak-Driven Region Segmentation (PDRS), a linear-time $\mathcal{O}(N)$ algorithm for rapid extraction of high-activity regions in irregularly sampled data. PDRS seeds candidate regions at statistically significant local maxima and expands them via a gradient-aware multi-source breadth-first search. Saddle-point merging and a median-based filter suppress spurious detections. Functioning as a computationally efficient pre-processing stage, PDRS isolates candidate transient events for downstream analysis. We demonstrate its efficacy on quasar light curves from SDSS Stripe~82 and AGN light curves from ZTF DR23, showing that PDRS identifies candidate high-activity regions comparable to those from Bayesian Blocks at substantially reduced cost. Its domain-agnostic formulation and physically interpretable parameters make PDRS broadly applicable beyond astronomy, including biomedical signals, seismic recordings, and industrial sensor monitoring.

In this paper, we use a highly sensitive telescope to characterize solar X-ray transients ranging from microflares in active regions down to weakly energetic brightenings in the quiet Sun. X-rays are closely linked to the initial energy release and immediate heating of solar flares, making them invaluable in understanding their driving processes. NuSTAR is the first long-term, direct focusing hard X-ray observatory to have observed the Sun, offering a unique opportunity to search for and characterize X-ray events from inside and outside active regions that would be otherwise unobservable. We present the first statistical survey of NuSTAR solar observations, characterizing the thermal and possibly nonthermal properties of 113 weakly energetic transients down to $10^{26}$ erg, making this the first to directly compare events from the quiet Sun to those in active regions. Relative to RHESSI microflares, our NuSTAR transients are generally cooler, dimmer, and have slightly steeper spectra. Thermal energy content of active region transients appears to be independent of the volume of emitting plasma for transients produced by active regions. This is in contrast to those from the quiet corona, which on average have lower energy content, smaller emission volumes, and appear cool but bright rather than hot but dim, suggesting a break in trends from traditional microflares. We found no quiet Sun transients with a thermal energy content above $3^{27}$ erg, implying an upper limit on the amount of energy released in plasma above 3 MK by quiescent processes.

We introduce an analytic approach to study gravitational lensing in the presence of a distribution of hadrons. The situation is analogous to the propagation of photons in a medium with a nontrivial Cooper-pair condensate, where the photon acquires an effective mass term that may depend on the coordinates if the condensate is not homogeneous. As a result, photons generally do not follow null geodesics in the hadronic medium. In this setup, hadrons are described by the nonlinear sigma model minimally coupled to Maxwell theory. The modified Raychaudhuri equation, including hadronic corrections, is derived, along with the integral curves of probe photons in the eikonal approximation. These results are consistent with the theory of gravitational lensing in plasma media, with the advantage that transport properties, such as the refractive index, can be expressed analytically in terms of the hadronic density without assuming a phenomenological modeling thereof. As an example, we study the hadronic lensing produced by an analytic black hole sourced by superfluid pionic vortices, and we obtain the hadronic correction to the deflection angle in the weak-field limit.

RX J1553.0+4457 (TMTS J15530469+4457458) is a nearby detached post-common-envelope binary containing a white dwarf and an active late-type companion. We present a multi-wavelength study of its short-timescale optical activity, orbital modulation, X-ray behaviour, and broadband spectral energy distribution. The analysis combines high-cadence BOOTES multi-band photometry, six sectors of public TESS full-frame imaging, Einstein Probe/FXT X-ray observations obtained after the WXT detection, CAFOS optical spectroscopy, and archival UV-to-mid-IR photometry. The BOOTES data reveal two short optical flares separated by about 3 h, with amplitudes of roughly 1-1.5 mag and faster decay at shorter wavelengths. The combined TESS light curve shows a stable signal at P = 0.083782 d, consistent with the first harmonic of the known spectroscopic orbital period, and the TESS flare sample lies in the energetic regime of active M-dwarf flares. During the same activity window, the EP/FXT spectra show a factor of about four decline in the 0.3-10 keV flux, mainly associated with decreasing emission measures. The broadband SED is well reproduced by a cool white dwarf plus a late-type M dwarf, with no clear mid-infrared excess. RX J1553.0+4457 is therefore best interpreted as a detached post-common-envelope binary whose rapid optical variability is dominated by magnetic activity on the late-type companion. A weak wind-fed or intermittent accretion contribution remains possible, but the current data do not require a luminous accretion disc or a dominant accretion-powered optical component.

We present \texttt{CosmoPostProcess}, a simulation-based forward-modelling algorithm calibrated to reproduce Euclid optical cluster observables. Its main deliverable is a correction for stacked surface-density profiles, binned in richness and redshift, accounting for selection systematics in richness-selected samples relative to unbiased references. We focus on the Euclid richness definition foreseen for cosmological analyses, which does not apply a colour selection; red-sequence richness is not considered. The algorithm processes $N$-body simulations by painting galaxies with a halo-occupation model and emulating survey detection and richness assignment. We also implement a novel estimate of optical cluster centres from projected galaxy densities, validated against Euclid pipelines. Baryonic effects are included through a correction calibrated on hydrodynamical simulations; the baryon-corrected excess surface density agrees within \(2\,\%\) over \(r\in[0.1,\,5]\,h^{-1}\,\mathrm{Mpc}\). Selection-bias contributions are assessed by varying cosmology and the mass--richness relation. Projection-induced selection bias follows a robust pattern: correlated large-scale structure projected along the line of sight enhances the stacked profile near the one-halo to two-halo transition, peaking at about \(1\,h^{-1}\,\mathrm{Mpc}\) with an amplitude of \(20\!-\!40\,\%\), depending on richness and redshift. The effect is mild at low and intermediate redshift ($z\lesssim0.7$), at the few-percent level, but becomes more relevant at higher redshift ($z\gtrsim0.7$). Baryonic modifications remain sub-dominant outside the core, at about \(2\,\%\) beyond \(r\gtrsim0.3\,h^{-1}\,\mathrm{Mpc}\). The framework delivers radial profile corrections with uncertainties, combining projection-induced selection bias, baryonic physics, and miscentring, to control systematics in Euclid DR1 cluster cosmology. (abridged)

Jet-cloud interactions are a key manifestation of Active Galactic Nucleus (AGN) feedback on nuclear scales, distinct from the large-scale radio-mode feedback that suppresses gas cooling in galaxy halos. On these smaller scales, radio jets can inject energy and momentum into the interstellar medium (ISM), shaping the physical and kinematic properties of the nuclear and circumnuclear regions of galaxies. Using JWST MIRI/MRS observations of seven nearby radio-loud AGN (3C293, 3C305, Centaurus A, Cygnus A, IC5063, NGC1052, and M87), we investigate jet-driven turbulence in both the warm molecular and ionized gas phases. By combining spatially resolved H$_2$/PAH flux ratios with diagnostic line ratios of the ionized gas, we constrain the dominant H$_2$ excitation processes and assess the impact of radio jet--ISM interactions on the multiphase gas. We find that radio jets drive enhanced turbulence in both molecular and ionized (traced by [FeII], [NeII] and [NeIII] lines) gas, not only along but also perpendicular to the jet axis, indicating that jet--ISM interactions extend beyond the collimated jet channel and affect the nuclear environment. Strong correlations between the H$_2$/PAH ratio, the H$_2$ excitation temperature, and shock-sensitive ionized-gas tracers indicate that jet-driven shocks dominate the excitation of the H$_2$ rotational lines in most sources. These results indicate that radio jets are a key driver of multiphase ISM kinematics and excitation in nearby radio-loud galaxies.

X-ray spectroscopy of AGN offers unique insights into the reprocessing of radiationand gas dynamics near SMBH. The Sey 1 galaxy Mrk 509 is an ideal laboratory for these studies since its complex FeK$α$ in emission and the past evidences of transient and fast flows. We present the first high-resolution 2-12 keV spectrum of Mrk 509 obtained with the Resolve calorimeter on-board XRISM, complemented with XMM-Newton and NuSTAR observations to constrain the broadband continuum. We modeled the spectra using self-consistent reflection models for the continuum and emission lines, and photoionized plasma models for the absorption components. The XRISM/Resolve spectrum reveals a narrow FeK$α$ core resolved with $σ\sim 10 eV$ (v$_{FWHM} \sim$ 1100 km/s) and a broader component with $σ\sim 450 eV$. We also find tentative evidence (3.6$σ$) for a ionized absorber. The data suggest that this component is infalling with a velocity of $v_{in} \sim 11000$ km/s and that it is located within few thousands gravitational radii. The narrow FeK$α$ emission is consistent with an origin in the dusty torus, while the broad component arises from the inner BLR or in the accretion disk (R$\sim 30--120 r_g$). Relativistic reflection modeling indictaes the inner edge if the emitting disk to R$\geq 27 r_g$. If confirmed, the high velocity inflow would likely represent fragmented clumps of a "failed wind" raining onto the accretion disk. providing potential direct evidence that non-standard accretion processes coexist with canonical disk-like flows in the inner regions of AGNS.

We use narrow-band Halpha+[NII] imaging data gathered during VESTIGE, a blind survey of the Virgo cluster carried out with MegaCam at the CFHT, to identify HII regions in 385 galaxies showing ionised gas emission. We identify 76645 HII regions in 322 star-forming galaxies and study their physical properties for those above the completeness limit (L(Ha)>=10^37 erg s-1). The present work is focused on perturbed cluster galaxies, identified as those having a reduced amount of HI when compared to similar objects in the field. We derive composite luminosity functions, diameter and electron density distributions, and several scaling relations, and compare them to those already derived for gas-rich, unperturbed systems identified during the VESTIGE survey. The analysis shows that the statistical and physical properties of HI gas-deficient cluster galaxies are different from those of unperturbed systems, with perturbed objects having a steeper faint-end slope and a brighter characteristic Ha luminosity than gas-rich galaxies. The difference in the two distributions comes principally from the outer disc (outside the effective radius). The analysis of the scaling relations indicates that perturbed objects have a lower number of HII regions per unit stellar mass and disc surface than unperturbed systems, with differences increasing with the HI-deficiency parameter, principally in the outer disc where HII regions are less present in gas-poor systems. All these differences can be explained in the framework of galaxy evolution in rich environments, where their hydrodynamic interaction with the surrounding ICM (ram pressure) removes the gas outside-in quenching the star formation activity in the outer disc once the HI is removed.

The candidate gravitational wave (GW) event S250818k was one of only three non-retracted LIGO-Virgo-KAGRA public alerts issued during the fourth observing run of the network (O4) with a binary neutron star (BNS) merger classification probability exceeding one percent. This triggered a prompt search for a potential electromagnetic (EM) counterpart in the large localisation error region (949 deg$^2$ projected in the sky at 90% credible level). The transient SN2025ulz, discovered by the Zwicky Transient Facility (ZTF) during the search, attracted a great deal of attention due to a potential spatial and temporal coincidence, and due to its initial fast decay and featureless spectrum. Here, we report on the follow up of this transient by the Electromagnetic counterparts of gravitational wave sources at the Very Large Telescope (ENGRAVE) Collaboration. We conducted an extensive multi-wavelength observational campaign, which led to the spectral classification of the transient as a type IIb supernova (SN), indicating that it is unrelated to the candidate GW event. In this article, we describe our observing strategies, data reduction, and interpretation. All of our results confirm and strengthen our classification of the source, and also show that shock cooling tails associated with type IIb SNe are one of the most prominent contaminants in kilonova searches.

We present a study of intergalactic medium (IGM) transmission at $4.5 < z < 6.5$ using high-signal-to-noise JWST/NIRSpec prism spectroscopy of 143 galaxies at $5<z<7$ from the CAPERS and JADES surveys. By comparing the observed flux blueward of Ly$α$ emission line to the prediction of spectral energy distribution modeling, we directly measure the IGM transmission along the individual galaxy sightlines. The average transmission measured from these galaxy sightlines is consistent with previous measurements based on luminous quasars. Current NIRSpec spectroscopy is sufficiently deep to probe IGM transmission on single sightlines. We find evidence for a highly ionized structure, \bubble, at $z\sim 5.75-6$ in the GOODS-S field based on the analysis of a high-S/N spectrum of one galaxy, GS-18846, at $z=6.335$. The IGM transmission of GS-z6IS is $0.17\pm0.02$, an order of magnitude higher than the average of previous measurements at this redshift. This structure has a line-of-sight scale of $\sim110$ cMpc and spatially extends over at least $21\times17$ cMpc$^2$. GS-z6IS is associated with a known large-scale galaxy overdensity at the same redshift, whose member galaxies show enhanced Ly$α$ visibility and a broader Ly$α$ equivalent width distribution compared to field galaxies at similar redshift. This result supports the interpretation that Ly$α$ overdensity can trace bubbles of increased IGM transmission, although environmental effects on galaxy properties may also contribute. Our study demonstrates that high-S/N galaxy spectra offer a powerful new approach to tracing ionized structures during the epoch of reionization.

Using a large sample of Type 1 AGN spectra, we investigated the complex structure of the Fe II emission region in order to understand the atomic processes responsible for the enhancement of the Fe II emission. We explored correlations between Fe II features and other spectral parameters, with special focus placed on the quasar main sequence, whose underlying physics is crucial for understanding the origin of the strong Fe II emission. The Fe II emission was modelled using the flexible Fe II template that decomposes the optical Fe II lines into several line groups. According to the atomic properties of transitions, the Fe II lines were divided into inconsistent and consistent groups (Fe II$_{incons}$ and Fe II$_{cons}$), while Fe II$_{cons}$ lines were additionally decomposed into components originating from different parts of the broad-line region (Fe II$_{ILR}$ and Fe II$_{VBLR}$). We traced the behaviour of these line groups and components along the quasar main sequence. Anti-correlation between the equivalent width (EW) of Fe II and the FWHM of Fe II appears to be a more fundamental relation underlying the quasar main sequence. The increase in the EW Fe II for smaller line widths is primarily caused by the strengthening of the EW Fe II$_{incons}$ lines and, with a smaller contribution, by the enhancement of the EW Fe II$_{ILR}$ components, while the EW of Fe II$_{VBLR}$, on average, does not significantly change along the quasar main sequence. The results indicate a possible stratification of the Fe II emission region occurring in sources with strong Fe II emission. An increased Eddington ratio may modify the broad-line region structure, leading to specific physical conditions suitable for triggering additional atomic processes. This may result in the appearance of Fe II$_{incons}$ lines and FeII$_{ILR}$ components, which consequently increase the optical Fe II strength.

We analyze the star formation and morphological properties of massive quiescent galaxies at cosmic noon ($2 < z < 3$) in the Abell 2744 field, using deep JWST NIRCam broad-band and medium-band imaging from the UNCOVER Treasury program and the MegaScience survey, complemented by archival HST data. Using BAGPIPES SED modeling, we select 14 unique massive quiescent galaxies ($M_* \gtrsim 10^{10}$ M$_\odot$, $\mathrm{sSFR} < 0.2/t_\mathrm{age}$). Morphological analysis with statmorph and pysersic reveals that most galaxies are intermediate type or S0s with a median Sérsic index $n \sim 4$, consistent with bulge-dominated systems. This value remains constant over $z \sim 1.5$--$4$, indicating that the morphology of massive galaxies is linked to their quiescence since at least $z \sim 4$. Spatially resolved SED modeling with piXedfit shows that $\sim 79\%$ of galaxies exhibit positive radial sSFR gradients, providing direct evidence for inside-out quenching, with the mean sSFR increasing by $\sim2$ dex from $R/R_e = 0.5$ to $4.5$. Formation time ($t_{50}$) profiles confirm that inner regions formed $\approx 0.5$ Gyr earlier, on average, than the outer regions, and quenching timescale profiles show that the cores were quenched more rapidly than the outskirts. Some galaxies show weak indications of possible AGN activity. Most galaxies are compact, with a mean half-mass radius of $R_e = 1.95 \pm 0.13$ kpc. The observed inside-out quenching pattern and possible AGN signatures are consistent with AGN feedback playing a role in star formation cessation, while the bulge-dominated morphologies suggest morphological quenching may also contribute.

Many pulsars exhibit a peculiar behaviour in their pulse profile of a sudden increase in their rotational period, which is popularly known as a pulsar glitch. Some of them show giant glitches with relative amplitude $ΔΩ/Ω\sim 10^{-6}-10^{-5}$. With the model of pinned neutron vortices inside the neutron star (NS) crust, this large glitch cannot be explained so far. However, the increasing evidence of massive pulsars indicates the appearance of exotic degrees of freedom in the inner core of the pulsars. Given this, we consider the pulsar as a hybrid star (HS). This model opens up the possibility of vortex-pinning inside the core. Under the Gibbs equilibrium conditions, it is possible for hadrons and the quark phase to coexist. Due to the global charge neutrality condition, quark pasta structures are formed in the background of hadronic matter. We consider these pasta structures as pinning sites of superfluid vortices. We show that considering the core contribution, our calculations come to be of the order of $ΔΩ/Ω\sim 10^{-6}$, which is close to the observations shown by the Vela-like pulsars.

We investigate deviations from the cosmic distance duality relation adopting model-dependent and -independent approaches using i) a Taylor expansion, ii) a power-law parameterization, iii) a logarithmic correction, iv) a (2;1) Padé polynomial and v) a second order Chebyshev parameterization. We derive constraints on all parameters using observational Hubble data, galaxy clusters, type Ia supernovae, DESI data and gamma-ray bursts. Through Monte-Carlo Markov chain analyses adopting the Metropolis Hastings algorithm, we find no significant violation of duality, then model selection criteria favor flat scenarios even though a slight curvature is not totally ruled out. For the $H_0$ tension we find a preference at $1$-$σ$ for $h^R_0=0.730\pm0.010$ from supernovae when dropping DESI data and for $h^P_0=0.674\pm 0.005$ from Planck when using DESI and gamma-ray bursts.

The Gaia BH3 system hosts the most massive known stellar-origin black hole and a low-mass metal-poor companion whose chemical composition may constrain early explosive nucleosynthesis processes. We investigate the chemical abundances of the companion in order to constrain the formation of this remarkable system. We perform a detailed analysis of high-resolution ESO-UVES spectra of the companion. 51 elements from lithium to uranium were investigated through spectral synthesis, including 15 treated in NLTE. We compare the resulting pattern to r-process enriched stars, to nucleosynthesis models and to stars of the ED-2 stream. The abundance pattern of the BH3 companion is consistent with that of r-I stars and is well reproduced by a combination of core-collapse supernova yields and an r-process component. The chemical patterns of four ED-2 stars closely match that of the companion especially when a dilution is taken into account. The present analysis provides the most detailed chemical characterisation of a metal-poor star associated with a stellar-mass black hole. The chemical similarity with ED-2 stars argue against local pollution across the binary system. The abundances instead reflect early spatially inhomogeneous enrichment of the progenitor cluster.

The origin of ultra-high-energy cosmic rays (UHECRs) remains an open question. Extragalactic magnetic fields can modify their propagation and, at sufficiently low energies, suppress the observed flux through the magnetic horizon (MH) effect.} {We quantify the impact of the MH on the propagation of UHECR protons using cosmological simulations and a dedicated numerical framework that follows cosmic rays in a time-evolving background.} {We use \texttt{UMAREL}, a parallel code developed for this study, to propagate UHECR protons through a cosmological volume simulated with ENZO. The magnetic-field configurations are chosen to be consistent with recent radio constraints on magnetic fields in cosmic-web filaments. Unlike stationary approaches, we follow particle trajectories through a sequence of time-evolving snapshots and compare the resulting arrival properties with those in an unmagnetised reference model.} {We find that observationally motivated extragalactic magnetic fields progressively suppress the flux of arriving protons below \(E \lesssim 3 \times 10^{19}\,\mathrm{eV}\) through an effective Magnetic Horizon (MH). We estimate \(R_{\mathrm{MH}} \sim 50\,\mathrm{Mpc}\) for protons with \(E = 10^{18}\,\mathrm{eV}\) and \(R_{\mathrm{MH}} \sim 150\,\mathrm{Mpc}\) for protons with \(E = 10^{19}\,\mathrm{eV}\).} {The MH generated by extragalactic magnetic fields must be taken into account when modelling UHECR propagation and interpreting the spectrum observed in the local Universe.}

This work presents a model for generating nonthermal power-law tails of particles' energy probability density functions in turbulent collisionless plasmas, applicable to both non-relativistic and relativistic scenarios. We propose that strong Alfvénic turbulence energizes plasma particles through curvature acceleration, particularly for particles with Larmor radii comparable to the scales of turbulence. When the energy density of the energized particles increases, the efficiency of the energy exchange process diminishes. As a result, the acceleration process saturates, leading to power-law distributions of particle momentum and energy. In the non-relativistic case, the momentum probability density function scales as $f(p) dp \propto p^{-3} dp $, while in the ultrarelativistic case, the energy probability density function scales as $ f(γ) dγ\propto γ^{-3} dγ$. This model provides a unified framework for understanding particle acceleration in both energy regimes, complementing existing analytical approaches. Its predictions are consistent with available observations of energetic ion distributions in the heliosphere and with the findings from numerical simulations of ultrarelativistic particle acceleration in magnetically dominated plasma turbulence.

LHAASO has reported 43 sub-PeV $γ$-ray sources, which are promising candidates for cosmic-ray (CR) accelerators above the PeV energy, commonly called as PeVatrons. Multi-wavelength observations are crucial for identifying the underlying particle species and estimating the CR energy content of these sources. In this work we investigate the region around LHAASO J1912+1014u (and HESS J1912+101) using Fermi-LAT $γ$-ray data and FUGIN CO data. We analyzed 15 years of Fermi-LAT data in the 0.4--409.6 GeV energy range. By improving the standard Fermi-LAT diffuse emission model, we significantly reduced the large residuals around the source in the 1.6-12.8 GeV band. We detected a statistically significant excess above the diffuse background, which likely represents $\ge$10 GeV emission associated with the LHAASO/H.E.S.S. source. The GeV excess exhibits a hard spectrum (photon index of about 2.1) and is well reproduced by interstellar gas templates with systemic velocities of about 25 $\mathrm{km~s^{-1}}$ or 60 $\mathrm{km~s^{-1}}$. We performed a comprehensive fit to the GeV--TeV spectral energy distribution. Although a leptonic scenario can reproduce the observed spectrum, a hadronic scenario is favored once electron cooling is considered. The inferred CR proton spectrum has an index of $\sim$2.2, and the total CR proton energy above 1 GeV is (1--5) $\times 10^{49}~\mathrm{erg}$, depending on the assumed velocity range of the associated interstellar gas. A stringent upper limit on diffuse X-ray emission further supports the proton PeVatron scenario.

Each strongly lensed image of a quasar behind a lensing galaxy (or galaxy cluster) is composed of a swarm of micro-images. This is a result of microlensing due to stellar-scale substructure in the lens. The presence of microlenses forms a network of micro-caustics, and a source transiting these micro-caustics gives rise to variation in observed strongly lensed images. These micro-image swarms are currently observable only through collective intensity fluctuations, which hide the underlying information on the stellar (and compact dark matter, if any) mass functions within the swarm. To unlock the information present in micro-image swarms, it is necessary to explore new techniques. In this work, we study the prospects of determining the micro-image swarm size in lensed quasar images using the intensity interferometry (i.e., the Hanbury Brown & Twiss effect). We consider QSO 2237+0305 and PS J0147+4630, two of the brightest lens quasars in the sky, and study micro-image swarm features in visibility space for both macro-minimum and macro-saddle-point images. At the end, we argue that, with ongoing and expected technical advances, observations of micro-image swarms are plausible, at least for the brightest lensed quasars.

The cold dark matter model successfully describes the Universe on large scales, yet faces challenges at sub-galactic scales. Ultralight dark matter (ULDM), with particle masses around $10^{-22} \mathrm{eV}$, offers a promising solution to these small-scale issues. Pulsar Timing Arrays (PTAs), designed to detect nanohertz gravitational waves, can also provide a sensitive probe for ULDM signals. In this work, we perform a Bayesian search for ULDM using PTA data sets, focusing on two types of signals: the oscillatory gravitational potential from scalar ULDM and the fifth-force interaction mediated by dark photon dark matter (DPDM). We incorporate pulsar distances in the analysis to better model the ULDM density. No statistically significant evidence for ULDM has been found, therefore we place 95% confidence-level upper limits on the relevant parameters. For scalar ULDM, our analysis does not exclude the scenario in which ULDM constitutes all of dark matter. The constraints from PPTA-DR3 show significant improvements over the earlier PPTA-DR2 (2018 Preview) across most of the mass range, and are consistent with the recent uncorrelated limits from other PTAs. We also present for the first time the DPDM constraints using EPTA data. The obtained bounds on the DPDM from the EPTA-DR2 and PPTA-DR3 are comparable to existing constraints.

We present a population study of Atacama Large Millimeter/submillimeter Array (ALMA) Cycle 6 observations of the 100 most gravitationally unstable dense cores in Aquila using a simple mass versus size analysis. We identify 66 continuum sources from ALMA 12m observations at 106GHz and through comparisons with known protostellar catalogs; two of these detected dense cores appear to be completely starless, without any accompanying/nearby protostar detections. Additionally, we find nine other starless ALMA 12m detections within protostellar cores that have fragmented into a mixture of starless and protostellar substructures. We test the turbulent core collapse model by conducting synthetic observations of turbulent magnetohydrodynamical simulations of collapsing starless cores in order to predict how many starless cores should be detected given their central density and density profile. The simulations predict at least one (1.19) detection, consistent with our two detections of ALMA 12m emission within completely starless cores. We also use a combination of ALMA Compact Array Cycle 4 observations and the Herschel Gould Belt Survey data to analyze how mass is distributed on three distinct spatial scales, in order to understand how turbulence shapes the evolution of substructure development as dense cores collapse to form new star systems. We find an increase in multiplicity at the smallest scales when the parent larger-scale structure also has a higher degree of fragmentation.

Black-hole shadows are purely geometric in the leading-order geometric-optics approximation: their boundary is set by null geodesics and carries no information about the polarization of the probing radiation. This changes at subleading order. We show that the gravitational spin Hall effect of light shifts the critical impact parameter governing photon capture by a helicity-dependent amount, causing polarized radiation with opposite helicities to trace slightly different shadow boundaries -- even in static, spherically symmetric spacetimes. The correction is analytic, universal, and scales as $1/ω$: it depends only on a single geometric function evaluated at the photon-sphere radius. We derive this result from the spin Hall equations of motion, confirm it numerically through ray-tracing calculations, and extend the analysis to Reissner-Nordström black holes, where electric charge amplifies the effect by up to a factor of $2.5$ at extremality. We further develop a perturbative treatment for slowly rotating (Kerr) spacetimes, showing that frame dragging introduces a $\cos\varphi$ modulation of the shadow splitting that can reverse its sign on one side of the image for spins $χ\gtrsim 0.21$. Although the magnitude of the effect is small, the conceptual implication is clear: black-hole shadows are not purely geometric observables.

Explosive events from magnetars are likely due to the catastrophic release of stress in their crusts, but the behavior of crustal matter beyond linear elasticity is poorly understood. We argue here that seminal results from molecular dynamics informing crust breaking calculations are non-converged, and must be revisited. We estimate the criteria for quasi-static, rate-independent flow by comparing imposed deformation timescales to grain boundary diffusion in polycrystals. We argue that convergence in this regime should be observed at strain rates slower than $10^{-5}\,ω_p$ (plasma frequency $ω_p$) in simulations of $N\approx10^5$ particles across order 10 grains at a quarter of the melting temperature. Though computationally expensive, this is tractable with modern methods and GPU supercomputers.

We study the dynamics of spinning charged test particles orbiting a Schwarzschild black hole immersed in a test uniform magnetic field. This setup provides a simple but physically relevant framework for modeling particle motion in magnetized astrophysical environments near compact objects, where both spin-curvature coupling and electromagnetic interactions can play a significant role. The particle trajectories are obtained numerically in both equatorial and off-equatorial configurations, allowing us to examine the influence of spin-curvature and Lorentz forces on the motion. In the equatorial plane, assuming the particle's spin vector is orthogonal to the orbital plane, we derive analytical expressions for the conserved energy and angular momentum, as well as for the radial and orbital frequencies as functions of spin parameter and magnetic parameter. We also construct the corresponding effective potential to determine the allowed regions of particle motion. The equatorial dynamics remain integrable due to the existence of conserved quantities associated with the spacetime symmetries and the alignment of the magnetic field. In contrast, the off-equatorial motion constitutes a non-integrable dynamical system. While limiting subcases of the system, i.e., the spinning neutral and non-spinning charged cases, can be analyzed using two-dimensional Poincaré surface of sections (PSs), the combined system can be reduced only up to three degrees of freedom. Hence, to investigate the resulting complexity, we analyze the phase space using four-dimensional PS along with recurrence analysis, revealing the presence of chaotic behavior for particular choices of parameters and initial conditions. Finally, we compare the dynamics of spinning charged test particles with the limiting cases, thereby distinguishing the respective contributions of spin-curvature and electromagnetic interactions.

Galactic magnetic fields regulate star formation and cosmic-ray (CR) transport, and understanding their three-dimensional structure, particularly in star-forming late-type galaxies, is key to constraining galactic CR transport. We explore the validity of stacking Stokes $Q$ and $U$ spectra, to infer about the intrinsic polarisation characteristics of star-forming galaxies. To prepare the stacking experiment, we align, scale, convolve, and reproject $C$-band (6 GHz) Stokes $Q$ and Stokes $U$ cubes of 27 star-forming late-type edge-on galaxies. On the stacked cubes, we perform RM-synthesis and discuss the derived polarised intensity (PI), polarisation angle ($χ_0$), and RM maps. Synthetic data tests demonstrate that stacking Stokes $Q$ and $U$ spectra is valid for tightly constrained underlying distributions of PI, $χ_0$, and RM. For underlying PI, $χ_0$, and RM distributions that represent star-forming galaxies, stacking introduces a systematic uncertainty of $δ_\mathrm{RM}^\mathrm{sys}=90 \mathrm{rad m^{-2}}$ and significantly underestimates the recovered PI. Stacking results reveal a clear X-shaped pattern in the polarisation plane, consistent with prior findings, detecting polarised emission up to 9 kpc above the galactic disc. We find stronger PI on the approaching side of galaxies. Furthermore, we find a decrease in PI in the galactic halo of $\sim 60$% near the galaxy's minor axis. A global RM pattern, as reported in a previous study, cannot be confirmed. Based on our analysis, we present stacking of Stokes $Q$ and Stokes $U$ cubes as an effective tool to recover faint polarised emission in the halo of nearby galaxies, if the underlying distributions of PI, $χ_0$, and RM are tightly constrained. Our findings motivate future studies using broader-band data to increase the resolution in Faraday depth.

We present predictions from the recent COLIBRE cosmological hydrodynamical simulations of galaxy formation for the present-day galaxy luminosity functions (LFs) at wavelengths ranging from the far-ultraviolet (FUV) to the submillimetre. The simulations are post-processed with the radiative transfer code SKIRT, accounting for dust attenuation and emission using the distribution and properties of dust grains predicted directly by COLIBRE. Results from simulations varying in mass resolution by a factor of $\sim 10^2$ ($\sim 10^5 - 10^7\,\mathrm{M_{\odot}}$) show very good convergence over most luminosity ranges. The COLIBRE-SKIRT LFs match the data remarkably well from the FUV to the near-infrared ($3.4\,\mathrm{μm}$) and also in the far-infrared and submillimetre wavelength range ($70-850\,\mathrm{μm}$). In the mid-infrared (MIR; $8-24\,\mathrm{μm}$), COLIBRE-SKIRT matches the data well at low luminosities but significantly underpredicts the luminosities of MIR-bright galaxies, with the discrepancy increasing towards longer wavelengths. The total infrared LF, obtained by integrating the spectral energy distributions over $8-1000\,\mathrm{μm}$, also matches observations well at the faint end but underpredicts the number of very bright galaxies. The unprecedented agreement at all other wavelengths indicates that COLIBRE, coupled with this calibration-free SKIRT post-processing framework, successfully predicts the properties of stellar populations at the present day and the amount and distribution of interstellar dust.

The pristine underdense patches of the Universe, cosmic voids, are powerful cosmological laboratories, uniquely sensitive to dark energy, modified gravity, and neutrino masses, yet their baryonic content remains uncharacterized. We present the first observational constraint on baryon underdensity in void interiors, exploiting the dispersion measures (DMs) of Fast Radio Bursts (FRBs) as tracers of the free electron column, independent of gas phase, temperature, and metallicity. By stacking 3,455 sightlines from CHIME/FRB on 1,288 SDSS BOSS voids over redshifts $0.2 < z < 0.7$, we measure a DM deficit toward void centers at $3.2σ$ significance, establishing that diffuse baryons inhabit the emptiest corners of the cosmic web at a suppressed level. The measured signal amplitude is consistent with an effective Universe model built directly from the observed galaxy underdensity in these voids, and a baryonic model calibrated to the FRB DM-redshift relation ($α_v = 1.80 \pm 0.87$). A uniform-density void model yields an electron density contrast of $δ_\mathrm{e,v} = -0.58 \pm 0.30$, implying a $\sim 60$% underdensity of baryons in void interiors relative to the cosmic mean. Jointly interpreting our FRB measurement with existing stacks of the thermal Sunyaev-Zel'dovich effect on voids further constrains the mean void gas temperature to $T_\mathrm{e} \lesssim (1.1 \pm 0.7) \times 10^6$ K, pointing to a warm-hot diffuse phase, consistent with hydrodynamical simulation predictions. With forthcoming FRB (CHORD, DSA, SKA) and galaxy (DESI, LSST, Euclid, PFS-Subaru, SPHEREx, Roman) surveys, set to expand both samples by orders of magnitude, this approach opens a new window onto tomographic baryon mapping, with direct implications for feedback models governing gas expulsion into low-density environments, and for the use of cosmic voids to extract cosmological constraints.

Recent ACT data favor a higher scalar spectral index $n_s$, placing models such as $α$-attractor T-models and natural Inflation in tension with current observations. We propose a K-inflation framework with a field-dependent non-canonical kinetic term $G(φ)$ that reconciles these models with the latest Planck-ACT-LB-BK18 constraints. Our analysis includes a refined calculation of the reheating equation-of-state parameter $w_{\rm re}$, avoids standard power-law approximations, and tests consistency with the Swampland Distance and de Sitter Conjectures. We find that the additional friction from the non-minimal kinetic coupling shifts both models into the favored observational regions. For the $α$-attractor T-model with $n=2$, viable solutions occur for $β\sim \mathcal{O}(10)$, with Swampland consistency favoring $α\gtrsim \mathcal{O}(10^{-3})$. This case predicts matter-like reheating and a red-tilted gravitational-wave background that is unlikely to be detected soon. In contrast, natural Inflation with $n=4,5$ is compatible with CMB constraints for $α\lesssim 7,8$ and $β\lesssim -1$, respectively, leading to stiff reheating and a blue-tilted gravitational-wave background potentially observable by LISA, Cosmic Explorer, Einstein Telescope, DECIGO, and BBO while satisfying BBN and $ΔN_{\rm eff}$ bounds. Combining gravitational-wave probes with Swampland criteria may therefore help distinguish possible UV completions of inflation.

In this study, we use a sample of 130 well-localized fast radio bursts (FRBs) to constrain the physical baryon density $Ω_{\rm b}h^2$, and the astrophysical contribution from host galaxies. The cosmological dependence entering the intergalactic dispersion measure is described through a non-parametric reconstruction of the Hubble parameter $H(z)$ obtained from cosmic chronometer data using the \texttt{ReFANN} neural-network framework, independently of the FRB sample. Within a Bayesian analysis, we jointly infer $Ω_{\rm b}h^2$ and the parameters of a log-normal host-galaxy distribution, namely its median $e^μ$ and logarithmic scatter $σ_{\rm host}$, using both real FRB data and a mock catalog. For the real sample, we obtain $Ω_{\rm b}h^2=0.02236\pm0.00090$, $e^μ=178.15^{+16.51}_{-16.97}~\mathrm{pc}\,\mathrm{cm}^{-3}$, and $σ_{\rm host}=0.794^{+0.064}_{-0.067}$. For the mock catalog, we find $Ω_{\rm b}h^2=0.02248\pm0.00018$, $e^μ=182.36^{+6.83}_{-6.48}~\mathrm{pc}\,\mathrm{cm}^{-3}$, and $σ_{\rm host}=0.711^{+0.024}_{-0.025}$. The baryon density constraint from the real FRB sample is in excellent agreement with both Big Bang Nucleosynthesis and Planck CMB determinations, differing from their central values by only $\simeq 0.05\%$. The mock analysis further illustrates the potential of future FRB samples, reducing the uncertainty on $Ω_{\rm b}h^2$ to the sub-percent level while remaining statistically consistent with early-Universe constraints. Our findings show that combining FRB dispersion measures with a non-parametric reconstruction of the expansion history provides a robust pathway to constrain both cosmological and astrophysical parameters, establishing FRBs as a complementary low-redshift probe of the baryon density.

4HWC J2029+3641 is a newly discovered point source detected by HAWC, with no previously identified TeV counterpart. The gamma-ray pulsar PSR J2030+3641, located 0.1 degree from the source center, is a middle-aged pulsar showing spin parameters similar to Geminga. Using Fermni-LAT data spanning from August 2008 to February 2026, we performed binned maximum likelihood spectral analysis in the energy range from 300 MeV to 1 PeV. A phase-resolved analysis was conducted to separate the off-peak and on-peak emissions. No significant spatial extension was found for the off-peak component. The off-peak spectrum exhibits strong curvature and is best described by an exponentially cutoff power-law model. The observed radio-to-gamma phase lag and narrow peak separation favor an outer-gap model for the gamma-ray emission.

Gas flows between galaxies and the circumgalactic medium (CGM) play a central role in galaxy evolution and can become observable as giant nebulae when illuminated by the quasars. We present an ensemble study of nebulae around 30 UV-luminous quasars at $z\approx0.4{-}1.4$ from the CUBS and MUSEQuBES surveys, 27 of which are detected in extended [O II] and/or [O III] emission. Based on a joint analysis of nebular morphology and surrounding galaxy environments, we introduce three morpho-kinematic classifications. We identify eleven irregular, large-scale (>50 kpc) systems, many of which are likely interaction-related; twelve compact host-galaxy-scale nebula, likely tracing CGM/ISM gas; and four systems with complex morphologies of uncertain origin. We introduce a quantitative measure of the spatial and kinematic association between nebulae and quasar-host group galaxies, finding a statistically significant association for ten nebulae, most of which are irregular, large-scale nebulae, consistent with qualitative analysis. Radio jets are detected in six systems, with no strong correlation found between radio activity and nebular emission. The [O II] nebulae are more asymmetric than their Lyalpha counterparts at $z>2$, but bear more similarity to H I gas observed in 21 cm around local elliptical galaxies. Blueshifted-redshifted patterns, likely tracing gas rotation, are observed in roughly 30% of the systems, though disturbed kinematics suggest that feedback may also be important. These results show that giant quasar nebulae are not a uniform class of objects, but instead arise through multiple pathways shaped by host-galaxy gas, galaxy interactions, group environments, and quasar activity, with the most striking cases associated with galaxy interactions.

This paper presents a necessarily incomplete review of the evolution of cosmology since the first Astro/Cosmo Moriond meeting in 1981. I trace the journey from the classical Big Bang model based on three pillars -- universe expansion, primordial nucleosynthesis, and the cosmic microwave background -- to the modern $Λ$CDM paradigm and the discovery of cosmic acceleration. I discuss major observational milestones: the COBE discovery of CMB fluctuations, the CMB measurements of the flat universe, the pivotal discovery of accelerated expansion through Type Ia supernovae and the emergence of precision cosmology with Planck. I review current tensions in cosmological parameters, particularly the Hubble tension and $\s8$ discrepancies, and discuss future prospects from large-scale structure surveys like DESI. The emergence of ``Big Bang 2.0'' reflects the profound paradigm shift from a model based on standard physics to a dynamical cosmos dominated by dark matter and dark energy, the description of which requires a physics that has yet to be developed and validated.

We investigate the pole structure of Kerr black-hole perturbations in the frequency domain, focusing on the building blocks of the Green's function for the radial Teukolsky equation: the homogeneous radial solutions, the connection coefficients, and the Green's function itself. We show that the homogeneous solutions and the local connection coefficients develop simple poles at the Matsubara frequencies, thereby establishing the Matsubara pole structure explicitly within the Teukolsky formalism for asymptotically flat subextremal Kerr black holes. At the level of the local fixed-sector connection formula, the explicit Matsubara-pole factors cancel in the ratio of connection coefficients entering a decomposed Green-function contribution. We also identify higher-order zero-frequency singularities in the decomposed Green-function contributions, which scale as $ω^{-2l-1}$ and cancel collectively in the total radial Green's function. These results clarify how Matsubara poles and sectoral zero-frequency singularities arise in the Teukolsky formalism and provide a frequency-domain foundation for understanding prompt response in time-domain ringdown waveforms in Kerr spacetime.

We measure the kinetic Sunyaev-Zel'dovich (kSZ) signal through a joint analysis of the pairwise kSZ effect and galaxy clustering using CMASS galaxies and ACT DR6 maps. This approach breaks degeneracies between the optical depth and nuisance parameters, enabling a reconstruction of the halo optical depth profile as a function of aperture scale. The kSZ signal reaches its highest signal-to-noise ratio of 7.2 at an aperture radius of $θ_{\rm AP} = 2$ arcmin, while the full profile rejects the no-kSZ hypothesis at $8.7σ$. Applying the same analysis pipeline to the Websky simulation, we find that the observed optical depth profile is somewhat more extended than the simulated one. This difference suggests that baryonic feedback in the real Universe may be stronger and redistribute gas to larger radii more efficiently than modeled in the simulation, although residual systematic effects and modeling uncertainties remain to be further investigated.

Cosmic microwave background anisotropies encode crucial information about the early Universe and fundamental cosmological physics. Although the standard $Λ$CDM model provides a successful description of cosmic evolution, persistent cosmological tensions and subtle small-scale anomalies still challenge its internal consistency. In this paper, we investigate six phenomenological amplitude parameters $A_{\rm{new}}$ (new=L, SW, Dop, eISW, lISW, Pol) corresponding to the key effects related to CMB anisotropy: the Lensing, Sachs-Wolfe, Doppler, early Integrated Sachs-Wolfe, late Integrated Sachs-Wolfe, and Polarization effects, respectively. Using modified CAMB and Cobaya packages, we constrain the $Λ$CDM$+A_{\rm{new}}$ models with two data combinations: Planck+DESI+PantheonPlus (PDP) and Planck+ACT+DESI+PantheonPlus (PADP). Only the $Λ$CDM+$A_{\rm{L}}$ is favored by AIC, with $A_{\rm{L}}=1.0656_{-0.0303}^{+0.0304}$ from PDP and $A_{\rm{L}}=1.0795_{-0.0289}^{+0.0260}$ from PADP, which implies 2.16$σ$ and 3.06$σ$ deviation from the $Λ$CDM model; values of $A_{\rm{SW}}$ show 1.21$σ$ and 1.96$σ$ deviations to 1; $A_{\rm{lISW}}$ is poorly constrained because the lISW effect has negligible influence at $\ell \geq 30$; and others are consistent with the $Λ$CDM model. Moreover, no noticeable improvement on the Hubble and $σ_8$ tensions is found within these one-parameter extended scenarios. ACT DR6 high-$\ell$ data strengthens the $Λ$CDM$+A_{\rm{L}}$ preference over the $Λ$CDM model, and reduces $A_{\rm Pol}$ uncertainty by more than one order of magnitude, highlighting the importance of ground-based high-$\ell$ observations for future CMB analyses.

We reconstruct a static and spherically symmetric black hole geometry originally proposed as an effective metric by identifying a consistent matter source derived from a fundamental action. The space-time is supported by a magnetically charged nonlinear electrodynamics (NED) field non-minimally coupled to a scalar field. Dimensional consistency reduces the parameter space to a single magnetic charge, and the inverse construction formalism yields a one-parameter family of electromagnetic Lagrangians $\mathcal{L}(F)=F^{n+1}/(n+1)$, encompassing both linear and nonlinear electrodynamics. We analyze the horizon structure and determine the critical magnetic charge separating black hole and horizonless configurations. The photon sphere and the corresponding shadow radius are computed, and observational bounds from the Event Horizon Telescope for Sagittarius A* constrain the allowed range of the magnetic charge. In the extended phase space thermodynamics, the solution satisfies the first law and the Smarr relation, exhibits a Hawking-Page phase transition, and presents a single change in stability without van der Waals-type critical behavior. We also investigate the topological properties of both the photon sphere and the thermodynamic parameter space. The photon sphere carries a total topological charge $Q_{\text{tot}}=-1$, while the thermodynamic vector field yields a global winding number $W=0$, placing the solution in the same topological class as the one of the Reissner-Nordström black hole. We finally discuss the versatility of this non-minimal coupling framework in both providing theoretical support to previously introduced solution and also to connect them to observational settings within strong-field gravity.

The considerable difference between early and late universe measurements of the Hubble constant, called the Hubble tension, poses a potential challenge to the standard $Λ$CDM cosmological model. We examine an interacting dark matter-dark energy model, $Λ_s$CDM, characterized by a gauge-invariant coupling $Q = ξHρ_{\mathrm{de}}$ and an effective pressure dynamically induced within the dark matter fluid. Using the CLASS Boltzmann code modified in this work, we analyze both the background and perturbation observables and compute an extensive Markov Chain Monte Carlo analysis with the latest cosmological datasets, including observational Hubble parameter data, Planck 2018 CMB compressed likelihood, BAO (from DESI DR2), Pantheon+ Type Ia supernovae, and redshift-space distortion measurements. The model predicts $H_0 = 71.8_{-0.3}^{+0.4}\mathrm{kms^{-1}Mpc^{-1}}$, reducing the tension with the SH0ES local measurement from about $5σ$ in $Λ$CDM to $1.2σ$ in $Λ_s$CDM. In contrast to the early dark energy model, the resolution emerges from late-time modification of the expansion history induced by the energy transfer from dark matter to dark energy. Moreover, the model suppresses late-time structure growth, providing $σ_8 = 0.744 \pm 0.0185$, lying below the $Λ$CDM value and moves in the direction preferred by weak lensing surveys. Since the interaction term is suppressed at high redshift, the pre-recombination sound horizon departs by less than $1\%$ from its $Λ$CDM value, suggesting that the alleviation of the tension dominantly originates from the late-time expansion rather than early-universe effects. We conclude that $Λ_s$CDM constitutes a phenomenologically viable interacting dark sector framework that addresses key cosmological tensions while remaining consistent with current precision data. }

Continuum reverberation mapping (RM) is a powerful technique for constraining the accretion disk structure in active galactic nuclei (AGNs). In typical cases, the shorter-wavelength emission is used as the reference, and a positive time lag is observed since the inner, hotter regions of the accretion disk respond earlier than the cooler outer regions at longer wavelengths. However, we detect a short-timescale negative inter-band lag in SDSS~J083717.88+191647 using RM techniques, where the \textit{g}-band lags behind the \textit{r}-band emission. The light curves from the Zwicky Transient Facility reveal two distinct phases, a stabilizing and a declining phase, in which the time lags show opposite signs. Using \texttt{JAVELIN} with the $g$-band as the reference, we obtain time lags of $3.68^{+1.94}_{-2.78}$~days during the stabilizing phase and $-1.60^{+0.69}_{-0.54}$~days during the declining phase. Although negative continuum lags have been reported in a few previous studies, the present case is distinguished by its clear phase dependence and the accompanying color evolution. We attribute the observed lag reversal to a moving dust-cloud obscuration scenario, in which the cloud crossing the line of sight preferentially obscures emission from the outer longer-wavelength regions of the disk, causing the $r$-band to decline earlier than the $g$-band and thus producing the observed negative inter-band lag. Our results indicate that AGN variability may be more complex than previously thought. Future high-cadence, multi-band observations will be essential to test this dust-obscuration model and to further explore the interplay between the accretion disk emission and dust in AGNs.

Blazars are promising sources of extragalactic high-energy astrophysical neutrinos, detected at energies $\gtrsim 10$ TeV by the IceCube neutrino observatory. Here, we report the first-ever broadband timing and spectral study of the flat-spectrum radio quasar PKS 1725+123, which has recently emerged as a compelling multimessenger target following its spatial association with the IceCube event IC-201021A. This triggered extensive follow-up observations from radio to VHE $γ$-rays, and a multi-episode flare was identified at a later time. During this period, the source exhibited high flux variability across all wavelengths. The {\it Fermi}-LAT analysis suggests rapid variability on timescales of less than 6 hours, implying a compact emission region with a radius of $\sim10^{16}$ cm. Our one-zone leptohadronic model shows that the high-energy $γ$-ray flux is produced by a combination of inverse-Compton scattering of external photons from the hot accretion disk and the broad-line region, while the X-ray emission is dominated by synchrotron self-Compton radiation from relativistic electrons. The secondary radiation from the hadronic cascade is found to be sub-dominant in the $γ$-ray regime, and the X-ray data constrain the maximum proton energy to $\sim 20$ PeV in the observer frame. Photopion production occurs predominantly with accretion-disk photons, resulting in an estimated muon-neutrino event rate of $\approx 0.3~\mathrm{yr}^{-1}$ during the flaring state with the flux peaking at $\sim1$ PeV. Future observations of TeV $γ$-rays by CTA and LHAASO will further constrain cosmic-ray production in this source.

We investigate the impact of nonequilibrium ionisation (NEI) and the metagalactic radiation-field on the thermal evolution, virial shock stability, and absorption signatures of gas surrounding galaxies. Using 1D, spherically symmetric hydrodynamical simulations with an extended version of the hydra code, we follow dark-matter growth, gas dynamics, time-dependent ionisation and cooling in the presence of the UV background. We explicitly track all ions of H, He, C, N, O, Ne, Mg, Si, S, and Fe in haloes of mass 1e11-1e13Msun from z=100 to z=0. Without a UV background, NEI enhances post-shock cooling due to underionised gas, reducing pressure support and raising the minimum mass for stable shock formation. Including the UV background pre-ionises the IGM, suppressing NEI, and restoring the CIE threshold. The IGM temperatures deviate from thermal equilibrium due to adiabatic expansion and collapse, while ionisation remains close to equilibrium in the presence of a UV background, except in transient rapidly cooling regions where NEI occurs. We compute absorption columns of OVI, CIV, and HI, showing that a photoionised IGM may produce substantial warm-ion columns extending beyond Rvir, including OVI column densities comparable to observed values. Our models indicate weak halo-mass dependence and extended distributions. We also find that z>~3 haloes can produce CIV (NCIV~1e13-15cm^-2) and HI (NHI~1e15-17cm^-2) columns out to ~10Rvir. Our results highlight the role of the UV background in regulating the thermal state and observable signatures of the gas surrounding galaxies, and emphasize the importance of accounting for IGM contributions when interpreting CGM absorption-line observations.

We demonstrate that in highly magnetized pair plasmas, nonlinear Alfven waves with wave-number $k \leq k_0 = ω_p^2 /(δω_B)$ ($δ=( δB)/B_0$ are relative fluctuations of the magnetic field) experience powerful modulational instability. In the two-fluid approximation, we develop an analytic set-up for circularly polarized (CP) Alfven mode in its frame (where the initial configuration is stationary; it is moving with relativistic, amplitude-dependent Alfven velocity $v_A (σ, δ) $, while both charges experience different, amplitude-dependent, synchrotron gyration). PIC simulations using EPOCH code demonstrate that for Alfven waves with $k$ near $k_0$, large, parametrically-driven density fluctuations develop, and lead to fast modulational instability. Charge separation effects, for a CP wave in magnetized pair plasma, might be temporarily important; on longer time-scales the density fluctuations are charge neutral and in symmetric pair plasma quickly grow to large amplitudes. In highly magnetized plasma, $σ\gg 1$, high frequency modes $k / k_0 \sim (2-3 ) \times σ\gg 1 $ are quickly generated; for smaller plasma magnetization, the dominant mode is at the Bragg's condition $k = 2 k_0$. Long term behavior of CP and LP modes is similar. We discuss application of the results to the physics of Fast Radio Bursts generated/propagating in the magnetospheres of magnetars.

We constrain Einstein-aether gravity -- a Lorentz-violating extension of General Relativity in which a dynamical, unit timelike vector field selects a preferred frame -- using updated high-precision pulsar timing observations of PSR J1738+0333 from EPTA second Data Release and the NANOGrav 9-year release, in combination with ToAs from Arecibo, Green Bank, Nancay, Parkes, and Westerbork. Our method accounts for both conservative and dissipative first post-Newtonian corrections arising from Lorentz violation; here we apply it to PSR J1738+0333 using the Bayesian timing pipeline Vela to process the full ToA dataset. We sample the joint posterior over binary component masses, post-Keplerian parameters and center-of-mass velocity components, and then apply a resampling scheme to propagate posteriors into robust constraints on the fundamental theory parameters, obtaining the most stringent strong-field bounds on the Einstein-aether coupling constants from a single binary pulsar system to date.

Black hole supernovae (BHSNe), the term we use for core-collapse events in which black hole (BH) formation occurs after shock revival but before the explosion is complete, have emerged as a natural outcome of multidimensional simulations as these calculations have been extended to seconds after bounce. Yet they remain one of the least studied outcomes of core collapse. Here, we assess whether they are confined to the most compact and massive progenitors, whose birth rates are low, or whether they arise systematically across a wider range of progenitor structures. We perform 23 long-term axisymmetric core-collapse simulations of progenitors spanning 19.51-60$\,M_\odot$ and compactnesses $0.31 \lesssim ξ_{2.5} \lesssim 0.63$. We find 18 BHSN outcomes across nearly the full ZAMS mass range considered, corresponding to progenitors with $0.40 \lesssim ξ_{2.5} \lesssim 0.63$. BH formation occurs between $\sim0.7$ s and $\sim4.4$ s after bounce. After BH formation, we continue the evolution with an excision treatment to at least 5000 s. The final explosion energies span $\sim2\times10^{49}$-$3\times10^{51}$ erg, while the final BH gravitational masses span $\sim3$-$26\,M_\odot$. We find a clear remnant-mass trend with CO-core mass, but show that the CO core alone is not an adequate proxy for the final BH mass, especially for progenitors at the low- and high-mass ends of the CO-core distribution. Except for the highest CO-core mass models, no single spherical mass coordinate cleanly separates ejecta from remnant material. Finally, a 2D axisymmetric and a 3D model are compared as we discuss differences between the two geometries.

Over the past decade, reionization studies have yielded persistent factor-of-two-to-five disagreements in the inferred ionizing escape fraction $f_{\mathrm{esc}}$ and peak star formation efficiency $f_{*,0}$, compounded by JWST's discovery of unexpectedly bright $z>10$ galaxies. We show that this discrepancy arises from an algebraically exact structural degeneracy: the ionizing photon rate $\dot{n}_{\mathrm{ion}} \propto f_{\mathrm{esc}} \times f_{*,0}$ renders all reionization-history probes, including Thomson optical depth, neutral hydrogen fraction, UV luminosity function, and quasar proximity zones, sensitive only to their product, leading to an intrinsically non-invertible mapping between model parameters and observations. We demonstrate the robustness of this degeneracy using a large suite of N-body simulations of self-interacting dark matter haloes spanning $10^9$-$10^{11} M_\odot$. Despite substantial changes to galaxy-scale structure, observables remain indistinguishable once the effective ionizing emissivity is matched, severely limiting reionization-based dark matter probes. We identify that only observables sensitive to the spatial topology of ionized regions can break this degeneracy. Our results provide a unified explanation for the scatter among published constraints and establish a framework for interpreting reionization observations and their implications for early galaxy formation and dark matter.

We present a forward-modeling framework to forecast the galaxies detected in the Chinese Space Station Survey Telescope (CSST) spectroscopic survey and the Square Kilometre Array (SKA) HI survey. Starting from the L-Galaxies 2020 semi-analytic model run on the Millennium-II N-body simulation (MS-II), the cold gas in galaxies is partitioned into atomic and molecular components self-consistently within the model. We further model the emission-lines (H $α$, H $β$, O III) relevant for the slitless spectrograph of the CSST in a post-processing step. We construct mock lightcones using the Mock Map Facility (MoMaF) approach, simulating the neutral hydrogen (HI) data cubes representing a 2000 hour SKA-Mid spectral line observation from redshifts 0.25--0.5, and employ the Source Finding Application 2(SOFIA-2) source-finding package to generate an HI galaxy catalog. In parallel, we apply the CSST selection function and noise model to obtain a realistic catalog of emission-line galaxies; the emission-line signal is proportional to the star formation rate. These products allow us to cross compare the galaxy samples and assess the synergy between CSST and SKA. We study the correlations of the HI and the emission-line signal with the halo mass, HI mass, and the stellar mass, and the baryonic Tully-Fisher relation (BTFR). We also perform stacking analysis of the HI signal from the CSST-selected sample, which probes the HI content in galaxies with low HI mass. Finally, we derive the optical-HI cross-correlation power spectrum of the galaxies, and measure the bias of these galaxies. These results can provide useful insight on the cold gas and stellar content of the galaxies.

Type Ib and Ic supernovae (SNe Ib/Ic) are the bright finale of massive stars that have lost their hydrogen envelopes, making them powerful probes of mass stripping in massive star evolution. The advent of modern large photometric and spectroscopic surveys presents the unique opportunity to investigate systematic differences between these two kinds of SNe. In this study, we analyze a large, homogeneous sample of SNe Ib/Ic light curves from the Zwicky Transient Facility. We find a systematic difference in their optical colors: SNe Ib are, on average, bluer than SNe Ic at a statistically significant level. This difference appears intrinsic, likely reflecting progenitors with different degrees of stripping -- helium-rich for SNe Ib and helium-poor for SNe Ic. In addition, we find that SNe Ib/Ic with narrow lines (SNe Ibn/Icn) are bluer than those without, which might originate from circumstellar matter interaction, with potential connection to fast blue optical transients. We demonstrate that SN colors offer a promising probe of mass stripping in massive stars, potentially providing a useful tool for analyzing large photometric data and improving predictions for the final outcomes of stripped massive stars.

The study of X-ray pulsar accretion columns helps us characterize accretion physics in this extreme regime of strong gravity and strong magnetic fields. Previous observations of the X-ray pulsar Hercules X-1 revealed a highly broadened Fe K emission line, associated with Doppler motions exceeding 0.1c, suggesting its origin in the accretion column. We obtained a high-spectral resolution view of the Fe K energy band of Hercules X-1 thanks to a 200 ks observation with the XRISM observatory. The XRISM/Resolve microcalorimeter spectra allow us to separate the different spectral components and accurately model them with phenomenological models. We confirm the presence of a broad line near 6.5 keV with a typical $1σ$ width of 1 keV. Performing a pulse-phase-resolved analysis, we find that the feature is strongly variable with Her X-1 pulse phase. This is consistent with the proposed origin due to collisional recombination or by reprocessing of the primary X-ray emission in the accretion column, where strong variability with pulse phase is expected due to the rotation of the columns alongside with the neutron star. Additionally, the Fe K line pulsation pattern evolves with the 35-day cycle of Hercules X-1, supporting the scenario that the neutron star and its accretion columns undergo precession, in agreement with recent polarimetric results from the IXPE observatory. We discuss the future applications of modeling of this broad line in X-ray pulsars with physical spectral models. This could be used to detect and track neutron star precession, advancing our understanding of neutron star interiors.

We investigate the detectability of polarized gamma-ray emission from blazar flares with the Compton Spectrometer and Imager (COSI). Using 17 years of Fermi Large Area Telescope observations, we analyze light curves for 1413 blazars and identify a maximum of 787 sources with flaring episodes through Bayesian block analysis. For each flare, we estimate the minimum detectable polarization MDP99 in the COSI energy band (0.2-5 MeV) using instrument response functions under a range of spectral assumptions and background conditions. Under baseline background levels (1 counts/s), and assuming that blazar flare statistics in the MeV band are comparable to those observed at GeV energies, we find that COSI can realistically detect polarization in up to ~6 flares with MDP99<50% over its two-year prime mission depending on different spectral and flare identification assumptions, with only a few most powerful ones reaching MDP99<20%. These expectations are shown to improve when shorter intervals around bright peaks within long flares are considered. We provide a ranked list of the most promising targets, finding that flat-spectrum radio quasars dominate the population of polarization-detectable events. Through its continuous all-sky monitoring in the largely unexplored MeV band, COSI will open a new observational window on blazar variability and deliver the first direct measurements of MeV polarization, offering unique insights into jet geometry and high-energy emission processes.

Galactic winds shape galaxy evolution; however, the outflowing gas is complex: it consists of multiple ionization phases, and its properties vary spatially. Therefore, methods that combine high-fidelity observations with state-of-the-art galactic-wind models are limited. Here we investigate methods for fitting the column density profiles derived from high-quality outflow observations with the multiphase, multiscale wind model from Fielding & Bryan 2022. We identify three key outflow parameters: the initial hot-phase mass-loading factor ($η_\text{ M,hot,0}$), the initial cool-phase mass-loading factor ($η_\text{ M,cool,0}$), and the initial cool-cloud mass. We obtain good fits for most galaxies, with tight constraints on $η_\text{ M,cool,0}$ and moderate constraints on the other two parameters. We find the inferred $η_\text{ M,cool,0}$ and $η_\text{ M,hot,0}$ are mostly of order unity, with significant scatter. The constraints on $η_\text{ M,hot,0}$ suggest that the interaction between the cool and hot phases allows us to indirectly constrain the properties of the hot wind from cool-outflow observations. The model also predicts various radial trends. First, for all galaxies, the cool-phase outflow velocity increases between $1-2$ times of the half-light radius, then reaches a plateau. Second, most galaxies exhibit increasing $η_\text{ M,cool}$ and decreasing $η_\text{ M,hot}$ with radius, with a few showing the reverse trends. These results are effective, model-conditional constraints, and are consistent with other recent multiphase simulations and observations. This highlights that the velocity-radius mapping encoded in UV absorption profiles enables recovery of outflow spatial structures from spatially integrated spectra. Our method paves the way for future broad parameter studies and guides updates of outflow simulations in future work.

We simulate a self-interacting three-flavor neutrino system within a core-collapse supernova using a hybrid classical-quantum algorithm on a qutrit computer. Based on the Dirac-Frenkel evolution equations, we employ a variation of the quantum-assisted simulator (QAS) to calculate the system's time evolution operator by performing qutrit Hadamard tests to find expectation values of unitary operators in the Hamiltonian. The time evolution simulation is then done classically. We find that the hybrid algorithm produces results comparable to an exact numerical integration out to times of $t \approx 30 \,ω_0^{-1}$ with time step $δt = 0.005 \,ω_0^{-1}$, where $ω_0$ is the energy scale of the single neutrino vacuum oscillations. We discuss the lessons learned in simulating neutrino systems using this hybrid quantum-classical algorithm, along with the advantages it offers over quantum Trotterization.

We analyze spectroscopy from one NIRISS pointing in the JWST-PASSAGE program for seven candidate $z \gtrsim 7.5$ photometrically-selected COSMOS-Web sources. We spectroscopically confirm one out of seven sources as a Lyman break galaxy (LBG) at $z=7.962^{+0.003}_{-0.006}$, with $m_{F150W} = 25.9$ (AB). The remaining sources are too faint in the continuum (i.e., $m_{F150W} \gtrsim 26$ AB) to provide a redshift measurement from the Lyman break, and do not show emission lines in their spectra. Although this study contains only one spectroscopically confirmed source, the confirmation of a luminous $z \sim 8$ galaxy within this $\sim4.8$ arcmin$^2$ field implies a surface density of $\sim 0.21^{+0.59}_{-0.17}$ arcmin$^{-2}$, $\approx 10\times$ higher than inferred from wide-area photometric surveys, suggesting a potential overdensity at $z\sim8$ in the COSMOS field.

An analysis of the kinematics of open star clusters (OSCs) using their characteristics from the new Hunt and Reffert catalog was conducted. Based on 4003 OSCs younger than 200 million years, the following values for the angular velocity of the Galaxy's rotation were found: $Ω_0 = 28.99\pm0.11$ km/s/kpc, $Ω^{'}_0 = -3.909\pm0.026$ km/s/kpc$^{2}$ and $Ω^{''}_0 = 0.5662\pm0.018$ km/s/kpc$^{3}$, where $V_0=234.8\pm3.0$ km/s for $R_0=8.1\pm0.1$ kpc. It was found that periodicity in the radial velocities of OSCs is manifested in clusters younger than 600 Myr, while a wave in residual tangential velocities is observed only in the youngest ones, younger than 40 Myr. A spectral Fourier analysis of the radial velocities of three OSC samples with average ages of 18, 72, and 143 Myr was used to obtain the following values of the wavelength $λ$ and the velocity perturbation amplitude $f_R$: $λ=2.0$ kpc and $f_R=4.3$ km/s, $λ=2.2$ kpc and $f_R=8.2$ km/s, $λ=2.1$ kpc and $f_R=9.6$ km/s, respectively. A systematic change in the positions of the maxima and minima of the waves in the radial velocities of OSCs was found depending on the age of the sample. From the analysis of these shifts, the value of the absolute value of the difference $|ΔΩ|$ between the angular velocity of rotation of the spiral pattern $Ω_p$ and the rotation velocity of the Galaxy was found, $|ΔΩ|=2.0\pm0.5_{stat}\pm2.3_{syst}$ km/s/kpc. Based on this, an estimate of two possible values of the corotation radius was obtained: $8.6\pm0.2$ kpc and $7.6\pm0.2$ kpc, which indicates that the Sun is very close to the corotation.

As the sensitivity of CMB telescopes increases, the need for precise calibration becomes critical. Started in 2022, the COSMOCal project aims to place an artificial polarized source in geostationary orbit, which will serve as a reference for CMB telescopes. This source will emit at 90, 150 and 270 GHz and will be linearly polarized with a highly precise orientation smaller than 0.1 deg. This proceeding presents the scientific motivations for the project, the current status of the development of the instrument and the results of a calibration campaign performed in March 2026 at the Institut d'Astrophysique Spatiale.

Dark matter density spikes forming around adiabatically growing black holes can dramatically enhance indirect and direct detection signals. Canonical predictions, however, assume a zero-mass seed in a purely dark matter environment and do not track the long-term dynamical impact of surrounding stars. We present a semi-analytic framework that first generalizes adiabatic spike formation to include finite seed masses, stellar cusps, and non-circular orbits, and then studies the subsequent cosmic evolution by solving coupled Fokker-Planck equations for the dark matter and stellar phase-space distributions, with a heating rate modulated by the cosmic star formation rate. Starting conservatively from canonical Gondolo-Silk spikes and marginalizing over astrophysical uncertainties, we find that stellar gravitational heating drives the inner slope towards $γ_χ\simeq 1.5$ within a few Gyrs (e.g by $z \lesssim 2$ for spikes formed at $z\simeq 10$), yielding overdensities two to four orders of magnitude below canonical expectations but still well above an NFW-like cusp. We provide redshift-dependent benchmarks for the column density and $J$-factor relevant to scattering, decay and annihilation signatures. Any robust interpretation of indirect dark matter signals from galactic nuclei must account for this evolution.

We present the results of a joint targeted search of candidate symbiotic stars at millimeter wavelengths using the South Pole Telescope (SPT) and the Atacama Cosmology Telescope (ACT). Candidates are selected from the New Online Database of Symbiotic Variables, restricting to objects that are within either the SPT-3G or ACT~DR6 footprint, covering most of the southern hemisphere and up to a declination of $+20^\circ$. Forced photometry on the 828 candidate symbiotic star locations in SPT and ACT data results in 31 unique objects detected with more than a $3σ$ significance using two frequency bands: 18 confirmed and 13 suspected symbiotic stars. We provide the SPT and ACT 95/98, 150, and 220~GHz light curves, along with optical and infrared light curves from 2016--2026, as well as spectral energy distributions, physical parameters from the literature, and brief summaries regarding the nature of each individual object. Using Herschel SPIRE data from 2013, we place upper limits on millimeter flux for CN Cha near the beginning of the optical rise in its 2012/2013 nova, which suggests a strong variability and lag at millimeter wavelengths and results in a rare observance of a Galactic millimeter slow transient. In addition, we provide coadded thumbnails and light curves for the remaining 797 candidate symbiotic stars that did not pass our detection thresholds. Millimeter-wave emission from symbiotic stars is primarily a combination of free-free emission of the ionization region and optically thick blackbody emission of the cooler dust components of the system. When combined with contemporaneous multi-wavelength observations, millimeter-wave observations can be used to test binary models of symbiotic stars and provide insight on the geometry and physical properties of these systems.

The identification of ultra-high energy cosmic ray sources is one of the open challenges of high-energy astrophysics. As charged particles travel through the Universe, they are deflected by extragalactic magnetic fields and lose energy through interactions with background radiation, making source inference highly non-trivial. Existing approaches either rely on simplified propagation models or on computationally prohibitive Monte Carlo methods. Here we present a simulation-based inference framework trained on three-dimensional \texttt{CRPropa~3} propagation simulations that produces calibrated posterior distributions over source energy, distance, direction, and primary composition for individual UHECR events. The model combines a Deep Set encoder, handling the variable number of detected secondary particles, with a normalizing flow, and is trained on approximately 5 million simulated events covering a broad range of extragalactic magnetic field configurations. Validated on held-out simulations, all source parameters are recovered without systematic bias, with directional parameters best constrained and source distance most uncertain, consistent with the underlying propagation physics. Primary composition classification achieves $\geq$~98.2\% accuracy across all mass groups. This framework provides a scalable and physically interpretable interface between detailed propagation simulations and Bayesian source inference relevant for current UHECR data.

Active galactic nuclei often exhibit highly collimated relativistic plasma outflows launched from the vicinity of their central black holes. One of the key theoretical challenges in understanding black hole jet formation is the origin of the plasma that feeds the jet, which remains poorly understood, particularly in explaining the observed jet emission. In this study, we focus on electron positron pair production generated by high energy photons from non axisymmetric magnetic reconnection near the black hole, as suggested by recent three dimensional general relativistic magnetohydrodynamics simulations. By employing general relativistic ray tracing, we calculate the spatial distribution of the pair production rate in the jet, taking into account photon propagation and collision angles in curved spacetime. We find that our scenario can naturally supply a sufficient amount of plasma to explain the observed radio emission from the M87 jet, even when photon anisotropy is considered. Furthermore, we show that a spinning black hole plays a crucial role in shaping the spartial dsitribution of the pairs, which in turn affects jet acceleration and very high energy emission from the jet base.

The superb sensitivity and angular resolution of the next-generation radio telescopes with combined frequency coverage of approximately over three orders of magnitude (100 MHz--100 GHz) will sample the radio and far-infrared (FIR) spectral energy distribution (SED) of galaxies and revolutionize the galaxy formation study at the epoch of re-ionization and beyond. We present a prospect of observing the radio--FIR continuum SEDs of galaxies in the redshift of up to $z\approx 20$ based on an ensemble of the simulated `energy balanced' panchromatic SED (from UV to FIR) extended to the radio. For `realistic' populations of UV star-forming galaxies and dusty star-forming galaxies, we simulate their SEDs by accounting for the CMB effect and the radio--IR correlation. The flux density evolution of the UV-bright star-forming galaxies and the dusty star-forming galaxies at the selected observing frequencies covered by the current (ALMA) and next generation (SKA and ngVLA) radio-millimeter telescopes, suggest that massive galaxies (M$_* \gtrsim 10^{10}$M$_{\odot}$) are detectable at any redshift ($0<z<20$) in high frequency ($ν>90$GHz). In particular, when operating, the ngVLA high-frequency ($\approx 100$ GHz) band is capable of detecting galaxies with M$_* \gtrsim 10^{9}$M$_{\odot}$ almost independently from redshift and the SKA low-frequency observing window ($\lesssim1$ GHz) has sufficient sensitivity to detect M$_* \gtrsim 10^{10}$M$_{\odot}$ dusty star-forming galaxies up to the epoch of reionization ($z=5\sim7$). We also show that the brightness of anomalous microwave emission (AME) in the galaxy SED is insignificant if the galaxies are beyond the local Universe (e.g., $z\gtrsim 0.1$).

Star-forming low-mass galaxies in the dense environments of galaxy clusters provide opportunities to study how environmental effects such as ram-pressure stripping, tidal interactions, or galaxy mergers shape a galaxy's star formation history. We combined integral-field spectroscopic observations with the Multi Unit Spectroscopic Explorer (MUSE) and available multi-band imaging of the star-forming galaxy NGC 1427A, located near the centre of the Fornax galaxy cluster, at a distance of 20 Mpc. Our aim was to trace the evolutionary history of NGC 1427A using the star formation history reconstructed from the integrated spectra and employing star clusters as surviving tracers of past star formation episodes. We fitted the spectral energy distribution of 222 star cluster candidates using archival $u,g,r$, and $i$ photometry to derive the ages and masses. For 58 clusters, we additionally incorporated their MUSE spectra in the fits and found an encouraging agreement between the photometric and spectroscopic results. The comparison of the age distribution of star clusters with star formation histories from a full spectrum fitting of the MUSE data found a reasonable agreement, with evidence for multiple episodes of star formation throughout the history of NGC 1427A. In particular, we found a population of young clusters ($\sim$ 10 Myr) that is located along the star formation edge and within the northern object, and a population of intermediate-age clusters ($\sim$ 100 - 300 Myr) with corresponding peaks in the star formation history of NGC 1427A. We interpret these populations in the context of the orbital evolution of NGC 1427A in the Fornax cluster and conclude that this galaxy has experienced not only ram-pressure stripping, but also tidal interactions or even a minor galaxy merger. The northern object is likely a regular component of the galaxy.

The 21-cm signal from cosmic dawn is a potentially sensitive probe of interactions between dark matter (DM) and baryons. We investigate the implications of the SARAS3 non-detection in the 55.5-84.4 MHz band for Coulomb-like interacting DM (IDM). In contrast to earlier constraint analyses that focused primarily on baryon cooling, we model the interaction self-consistently by including both excess cooling of the gas and the suppression of structure formation, which delays the onset of star formation and hence suppresses the Ly$α$, X-ray, and ionizing backgrounds at early times. We perform a joint Bayesian fit of a global 21-cm signal model and a flexible foreground model to the SARAS3 antenna temperature, and find that the signal parameters remain weakly constrained after marginalizing over the foregrounds. The null result is nonetheless informative: the data disfavour deep absorption features within the observed band, with the strongest bound at $z = 23.6$ ($ν\approx 57.7$ MHz), where $T_{21} \gtrsim -277.6$ mK at $3σ$. Comparing the IDM and standard cold dark matter scenarios, we find no statistically significant preference for IDM (Bayes factor $B \approx 1.7$). While we do not constrain the strength of baryon-DM interactions, the SARAS3 non-detection places a meaningful upper bound on the amplitude of the global 21-cm signal in this class of models.

We use a new implementation of the NEFERTITI galaxy formation model, coupled to $\sim 30$ high-resolution Caterpillar dark-matter simulations of Milky Way (MW) analogues, to connect early galaxy formation with the MW's assembly down to $z=0$. Our locally-constrained model resolves minihaloes hosting the first PopIII stars and self-consistently tracks inhomogeneous ionization and chemical enrichment. PopIII star formation begins at $z\simeq27$, peaks at $z\simeq10-15$, and persists down to $z\lesssim5$, producing PopIII systems with $M_*\sim10-5\times10^5\:{\rm M_\odot}$. The present-day descendants of PopIII stars span ${\rm [Fe/H]<-9}$ to ${\rm [Fe/H]\approx-1}$, with the most metal-poor stars typically enriched by a few (1-4) low-energy supernova progenitors. Pair-instability supernova descendants more commonly form in massive haloes ($M_{\rm vir}>10^8\:{\rm M_\odot}$), often externally enriched, reflecting the strong feedback and delayed recovery following energetic explosions. These early systems serve as building blocks for the present-day Galaxy's metal-poor component: although 90$\%$ of the total stellar mass formed in situ, the accreted component dominates at $[{\rm Fe/H}]<-1$ and accounts for nearly all stars with $[{\rm Fe/H}]<-3$. This accreted population is largely built by a few ($\sim5$) massive ($M_*>10^8\:{\rm M_\odot}$) destroyed dwarfs, but lower-mass systems become increasingly important at low metallicities, with ultra-faint and classical dSph analogues contributing $\sim25\%$ at $[{\rm Fe/H}]<-3$. Our model simultaneously reproduces the properties of metal-poor MW stars and the JWST "Hebe" galaxy at $z\sim11$, supporting its identification as a pure PopIII system. Ultimately, NEFERTITI is a key tool to interpret upcoming local and high-$z$ observations linking the near- and far-field cosmology.

Accretion discs are fundamental to many astrophysical systems, providing the conversion of gravitational potential energy into radiation that we can observe. In many systems there is evidence that discs are warped; from spatially-resolved observations of protoplanetary discs, to the features of lightcurves and line profiles from discs around supermassive black holes in galaxy centres. The dynamics of warped discs is largely controlled by the physical nature of the internal disc viscosity. While typically disc viscosity is hydromagnetic in origin, simulations of magnetized discs cannot match observed rates of angular momentum transport in planar discs and thus cannot be used to determine the ratio of the torques responsible for driving accretion to those responsible for evolving the disc warp. The analytic work of Ogilvie is the most comprehensive model for warped disc evolution, but makes assumptions that need to be tested. In particular, it assumes that the disc viscosity is Navier-Stokes, and therefore small-scale and isotropic. Here we attempt to test this model using the long periods of X-ray binaries that are due to precession of the disc. These systems have well-constrained estimates of the component of viscosity responsible for driving accretion, and by looking at systems with and without evidence for disc misalignment and precession we can constrain the component of viscosity responsible for flattening the disc. We conclude that the observational constraints suggest that the Ogilvie model provides an adequate description of the disc evolution, but that there are indications that the internal disc viscosity might be marginally non-isotropic.

Motivated by the discovery of peculiar dwarf galaxies inside galaxy clusters such as blue candidates (BCs), dark galaxies and ultra-diffuse galaxies (UDGs), we present hydrodynamic simulations of galaxy mergers in cluster environments. We vary the viscosity and stellar feedback prescriptions, realistically modelling possible conditions for hydrodynamic drag and fluid instabilities, as well as internal destabilization through stellar feedback-driven heating and gas loss. We find that long-lived tidal dwarf galaxies (TDGs) can form throughout all viscosity values applicable to galaxy clusters if stellar feedback is moderate. Our results expand on studies of cloud crushing simulations, investigating the entrainment problem in intracluster medium ambience. The smallest clouds have gas masses on the order of $M_\text{gas} \sim 10^7 \text{ M}_\odot$ and reach relatively low final drift velocities of $\sim 100 \text{ km/s}$. The lowest possible Reynolds number acting on this class of clouds is $Re \sim 1$ for full Spitzer viscosity. Almost all TDGs display elevated star formation rates of $0.01-0.1 \text{ M}_\odot / \text{yr}$, which are stable across several Gyr. Based on their matching properties, we support that BCs observed in the Virgo cluster are likely stripped TDGs. Similar features are also found in comparison with dark galaxies and baryon-dominated UDGs, implying that a subsample of these objects are also long-lived TDGs. This work provides robust evidence that stripping from galaxy mergers is a viable channel for the formation of stable cold gas clouds and dark matter-deficient galaxies observed in galaxy clusters.

The early environmental transformation of low-mass cluster galaxies from gas-rich to gas-poor remains poorly constrained, partly because clear, phase-resolved observations are rare. NGC 1427A, a disturbed star-forming dwarf in the Fornax cluster, offers a favorable case for studying this process. We aim to build a spatially resolved, multi-phase picture of NGC 1427A to constrain the roles of ram-pressure stripping and tidal perturbations. We combine a deep, spatially contiguous VLT/MUSE mosaic with ancillary data from the FUV to the radio. Full-spectrum fitting of the MUSE cube yields maps of stellar kinematics, ages, metallicities, and continuum attenuation, while emission-line modeling provides ionized-gas kinematics, Balmer-decrement reddening, and star-formation-rate surface densities. Ancillary multi-wavelength data trace neutral and molecular gas, dust, and recent star formation, placing the MUSE-based results in a broader multi-phase context. We find a pronounced decoupling between stars and gas: the H I and ionized gas rotate about an axis tilted with respect to the stellar field and are globally blueshifted. Stellar and nebular attenuation, infrared dust tracers, and H I morphology indicate stripping with a strong line-of-sight component that has reached the ISM. At the same time, the asymmetric gas and dust distribution, together with structured and time-dependent star formation, points to an additional gravitational perturbation, with a recent mild fly-by by a nearby dwarf being the favored interpretation. We propose that dwarf-dwarf tidal effects have torqued and pre-conditioned the gas, while the Fornax intracluster medium is driving ram-pressure stripping that now reaches the ISM and coincides with a declining global star formation rate. This places NGC 1427A at the onset of environmental quenching, making it a useful benchmark for early cluster dwarf transformation.

We present general series solutions to the Tolman-Oppenheimer-Volkoff equations for compact stellar objects. We develop an algorithm to compute the coefficients of the power series in terms of the equation of state and its derivatives with respect to the thermodynamic variables. Using these results, we establish general properties of analytic solutions and their relation to the regularity of the equation of state. Applying the theory of Padé approximants, we derive series representations for meromorphic functions whose domains of convergence may include isolated poles. These analytic solutions are then used to obtain closed-form expressions to approximate the radius and mass of stellar objects. We apply the formalism to specific models, namely fluids with affine equations of state and polytropic fluids, and compare the results with those obtained from numerical integration. Lastly, we extend the formalism to piecewise equations of state, deriving series solutions that can be matched across transition hypersurfaces.

In this work, we investigate whether the location of galaxies within the large-scale structures (LSS) of the Universe affects their stellar mass ($M_\star$) and morphology. To this end, we attempt to disentangle the effects of local and large-scale environments in their distributions. We classify 25309 galaxies in the redshift range ${0.02 < z \leq 0.04}$ with $\log M_\star/\rm{M}_\odot \geq 9.5$ in terms of the main LSS (voids, clusters, and not clusters nor voids, referred to as NCNV) and local environment (singlets and multiplets; galaxies with and without companions). We present the stellar mass and morphology distributions in these environments, and for a subsample of galaxy pairs. Even in voids, we find that $\sim22\%$ of galaxies have companions. Stellar mass distributions show that galaxies are less massive in voids, regardless of their local environment. Satellites in voids are, too, less massive relative to their centrals than in NCNV pairs. In terms of morphology, the denser the LSS, the greater is the proportion of early-type galaxies, even among singlets. In voids and NCNV, late-type multiplets tend to be later-type spirals than singlets. In pairs, centrals tend to be more early-type than satellites. The sample, curated to avoid morphology incompleteness, yields slightly higher fractions of early-type galaxies and multiplets than previous studies. We conclude that the local environment alone is insufficient to explain the distribution of stellar mass and the morphology of galaxies in the local Universe. The observed mass distributions support a scenario in which galaxy assembly depends critically on the host halos, and the properties of these halos are related to their large-scale environment. This would explain the finding of lower-mass galaxies in voids than in denser environments, and provide a basis for considering a common evolutionary origin for multiplets.

(Abridged) Simulation-based inference (SBI) has emerged as a powerful framework for extracting cosmological information from complex, non-linear data where analytical likelihoods are unavailable. Its reliability is commonly assessed using coverage-based diagnostics under the prior predictive distribution, which probe calibration only in an averaged sense and do not constrain posterior behavior at fixed parameter value, the regime relevant for practical inference. We investigate these limitations in the context of primordial non-Gaussianity, parameterized by $f_\mathrm{NL}$, using simulations of the dark matter halo field. We compare SBI based on contrastive neural ratio estimation (CNRE) with likelihood-based inference (LBI) using the power spectrum, bispectrum, and wavelet scattering transform (WST) coefficients across 1000 realizations. SBI and LBI agree well on posterior means and skewness, while the variance agrees on average but shows weaker realization-by-realization consistency. Larger differences arise in the kurtosis, indicating discrepancies in the posterior tails. These effects are already present for the power spectrum - where the Gaussian likelihood assumed in LBI is best justified - and are most pronounced for the combined power spectrum and bispectrum, where SBI posteriors are often underconfident and can yield weaker constraints than either statistic individually, despite passing coverage tests. WST coefficients further tighten constraints on $f_\mathrm{NL}$, even when restricted to large scales. Our results highlight both the potential of higher-order statistics and the need for validation strategies that probe the posterior shape beyond standard coverage diagnostics.

In the nearby universe, quasars are well known to exhibit variability in their brightness over time, offering a powerful tool to probe the physics of accretion onto the SMBH and directly measure the mass of the SMBH. However, detecting variability in early quasars remains challenging. Here, we report the detection of multi-wavelength infrared and X-ray variability in a quasar observed just 850 million years after the Big Bang. The infrared variability spans five filters, tracing rest-frame ultraviolet and optical emission from the accretion disk, while the X-ray variability probes the corona. The variable spectrum reveals that the accretion disk has a geometrically thin, optically thick structure. This provides observational constraints on the accretion disk structure at early times, when quasars are accreting at high Eddington ratios and reside in extreme environments. Our findings demonstrate the feasibility of characterizing accretion physics using variability in the early universe, laying the groundwork for studies exploiting upcoming facilities such as the Rubin Observatory and Roman Space Telescope. These facilities will discover large samples of variable high-redshift quasars, enabling population-level variability studies of accretion physics and black hole masses, filling key missing ingredients in understanding early SMBH growth.

A Keplerian binary immersed in a bath of lighter particles hardens by ejecting them through gravitational slingshots. This process drives, for example, the evolution of supermassive black hole binaries following galaxy mergers, and has long been described with just two parameters: the hardening rate and the eccentricity growth rate. Here we show that the secular dynamics is substantially richer. Combining symmetry arguments with extensive three-body scattering experiments, we demonstrate that the medium exerts a net force on the binary's center of mass (CoM), induces apsidal precession, and rotates the orbital plane when the CoM velocity has an out-of-plane component. Remarkably, these deterministic effects persist even in a perfectly uniform and isotropic medium, as the binary's own asymmetry provides the propulsion. The interplay of self-acceleration, precession, and dynamical friction drives the CoM along an outward spiral. For supermassive black hole binaries, this displacement dominates over Brownian motion and approaches the radius of influence, suggesting they may be significantly offset from their host galaxies' centers. The displacement also enlarges the stellar loss cone, with direct implications for the final-parsec problem. We further show that the previously reported circularization of small-mass-ratio binaries is a numerical artifact of truncating long-lived encounters: all binaries undergo eccentricity growth. Our results enrich the standard picture of binary hardening and have implications in a variety of astrophysical contexts, including gravitational-wave source populations.