arXiv Daily Digest - 2026-04-13

Reinforcement learning (RL) for large language models (LLMs) increasingly relies on sparse, outcome-level rewards -- yet determining which actions within a long trajectory caused the outcome remains difficult. This credit assignment (CA) problem manifests in two regimes: reasoning RL, where credit must be distributed across tokens and steps within a single chain-of-thought generation (500--30K+ tokens); and agentic RL, where multi-turn environment interaction introduces stochastic transitions, partial observability, and horizons of 100+ turns (100K--1M tokens), making episode-level credit increasingly uninformative. We survey 47 CA methods (41 core, 6 adjacent enablers) published between 2024 and early 2026, organizing them in a two-dimensional taxonomy by assignment granularity (token, segment, step, turn, multi-agent) and methodology (Monte Carlo, temporal difference, model-based, game-theoretic, information-theoretic). Beyond the survey itself, we contribute three reusable resources: (1) a structured, machine-readable paper inventory with taxonomy labels, baseline families, and evidence levels; (2) a reporting checklist for future CA papers, validated against the reviewed literature to identify systematic methodological gaps; and (3) a benchmark protocol specification with task families, metadata requirements, and controlled bifurcation tasks, accompanied by a method selection decision tree. Our synthesis suggests that the shift from reasoning to agentic RL complicates and reshapes the credit assignment landscape: reasoning CA is maturing around process reward models and critic-free group comparison, while agentic CA is driving genuinely new approaches -- hindsight counterfactual analysis, privileged asymmetric critics, and turn-level MDP reformulations -- that have no direct precedent in reasoning RL.

While Large Language Models (LLMs) have demonstrated significant potential in Tool-Integrated Reasoning (TIR), existing training paradigms face significant limitations: Zero-RL suffers from inefficient exploration and mode degradation due to a lack of prior guidance, while SFT-then-RL is limited by high data costs and capability plateaus caused by low-entropy collapse. To address these challenges, we propose E3-TIR (Enhanced Experience Exploitation), a warm-up paradigm for the early stages of agent training. Specifically, we formulate training as the dynamic integration of three experience types: Expert Prefixes, Expert Guided, and Self-Exploration. By executing diverse branching exploration around expert "anchors" and employing a mix policy optimization mechanism, we effectively mitigate distribution shifts and resolve optimization conflicts arising from shared prefixes. Our method dynamically adapts the model's knowledge boundaries, effectively balancing exploration diversity with training efficiency.Experimental results demonstrate that E3-TIR achieves a 6 performance improvement over traditional paradigms on tool-use tasks, while requiring less than 10 of the synthetic data. Furthermore, in terms of ROI, a comprehensive metric integrating performance, data cost, and training efficiency we achieve a 1.46x gain compared to baselines. Code is available at https://github.com/yuki-younai/E3-TIR.

Safety guarantees are a prerequisite to the deployment of reinforcement learning (RL) agents in safety-critical tasks. Often, deployment environments exhibit non-stationary dynamics or are subject to changing performance goals, requiring updates to the learned policy. This leads to a fundamental challenge: how to update an RL policy while preserving its safety properties on previously encountered tasks? The majority of current approaches either do not provide formal guarantees or verify policy safety only a posteriori. We propose a novel a priori approach to safe policy updates in continual RL by introducing the Rashomon set: a region in policy parameter space certified to meet safety constraints within the demonstration data distribution. We then show that one can provide formal, provable guarantees for arbitrary RL algorithms used to update a policy by projecting their updates onto the Rashomon set. Empirically, we validate this approach across grid-world navigation environments (Frozen Lake and Poisoned Apple) where we guarantee an a priori provably deterministic safety on the source task during downstream adaptation. In contrast, we observe that regularisation-based baselines experience catastrophic forgetting of safety constraints while our approach enables strong adaptation with provable guarantees that safety is preserved.

The accelerometer has become an almost ubiquitous device, providing enormous opportunities in healthcare monitoring beyond step counting or other average energy estimates in 15-60 second epochs. Objective: To develop an open data set with associated open-source code for processing 50 Hz tri-axial accelerometry-based to classify patient activity levels and natural types of movement. Approach: Data were collected from 23 healthy subjects (16 males and seven females) aged between 23 and 62 years using an ambulatory device, which included a triaxial accelerometer and synchronous lead II equivalent ECG for an average of 26 minutes each. Participants followed a standardized activity routine involving five distinct activities: lying, sitting, standing, walking, and jogging. Two classifiers were constructed: a signal processing technique to distinguish between high and low activity levels and a convolutional neural network (CNN)-based approach to classify each of the five activities. Main results: The binary (high/low) activity classifier exhibited an F1 score of 0.79. The multi-class CNN-based classifier provided an F1 score of 0.83. The code for this analysis has been made available under an open-source license together with the data on which the classifiers were trained and tested. Significance: The classification of behavioral activity, as demonstrated in this study, offers valuable context for interpreting traditional health metrics and may provide contextual information to support the future development of clinical decision-making tools for patient monitoring, predictive analytics, and personalized health interventions.

Chest X-ray report generation (CXR-RG) has the potential to substantially alleviate radiologists' workload. However, conventional autoregressive vision--language models (VLMs) suffer from high inference latency due to sequential token decoding. Diffusion-based models offer a promising alternative through parallel generation, but they still require multiple denoising iterations. Compressing multi-step denoising to a single step could further reduce latency, but often degrades textual coherence due to the mean-field bias introduced by token-factorized denoisers. To address this challenge, we propose \textbf{ECHO}, an efficient diffusion-based VLM (dVLM) for chest X-ray report generation. ECHO enables stable one-step-per-block inference via a novel Direct Conditional Distillation (DCD) framework, which mitigates the mean-field limitation by constructing unfactorized supervision from on-policy diffusion trajectories to encode joint token dependencies. In addition, we introduce a Response-Asymmetric Diffusion (RAD) training strategy that further improves training efficiency while maintaining model effectiveness. Extensive experiments demonstrate that ECHO surpasses state-of-the-art autoregressive methods, improving RaTE and SemScore by \textbf{64.33\%} and \textbf{60.58\%} respectively, while achieving an \textbf{$8\times$} inference speedup without compromising clinical accuracy.

The Tactile Internet demands sub-millisecond latency and ultra-high reliability, as high latency or packet loss could lead to haptic control instability. To address this, we propose the Mode-Domain Architecture (MDA), a bilateral predictive neural network architecture designed to restore missing signals on both the human and robot sides. Unlike conventional models that extract features implicitly from raw data, MDA utilizes a novel Continuous-Orthogonal Mode Decomposition framework. By integrating an orthogonality constraint, we overcome the pervasive issue of "mode overlapping" found in state-of-the-art decomposition methods. Experimental results demonstrate that this structured feature extraction achieves high prediction accuracies of 98.6% (human) and 97.3% (robot). Furthermore, the model achieves ultra-low inference latency of 0.065 ms, significantly outperforming existing benchmarks and meeting the stringent real-time requirements of haptic teleoperation.

Large language model agents receive instructions from many sources-system messages, user prompts, tool outputs, and more-each carrying different levels of trust and authority. When these instructions conflict, models must reliably follow the highest-privilege instruction to remain safe and effective. The dominant paradigm, instruction hierarchy (IH), assumes a fixed, small set of privilege levels (typically fewer than five) defined by rigid role labels (e.g., system > user). This is inadequate for real-world agentic settings, where conflicts can arise across far more sources and contexts. In this work, we propose Many-Tier Instruction Hierarchy (ManyIH), a paradigm for resolving instruction conflicts among instructions with arbitrarily many privilege levels. We introduce ManyIH-Bench, the first benchmark for ManyIH. ManyIH-Bench requires models to navigate up to 12 levels of conflicting instructions with varying privileges, comprising 853 agentic tasks (427 coding and 426 instruction-following). ManyIH-Bench composes constraints developed by LLMs and verified by humans to create realistic and difficult test cases spanning 46 real-world agents. Our experiments show that even the current frontier models perform poorly (~40% accuracy) when instruction conflict scales. This work underscores the urgent need for methods that explicitly target fine-grained, scalable instruction conflict resolution in agentic settings.

UI-to-Code generation requires vision-language models (VLMs) to produce thousands of tokens of structured HTML/CSS from a single screenshot, making visual token efficiency critical. Existing compression methods either select tokens at inference time using task-agnostic heuristics, or zero out low-attention features without actually shortening the sequence -- neither truly reduces prefill latency or adapts to the non-uniform information density of UI screenshots. Meanwhile, optical (encoder-side learned) compression has shown strong results for document OCR, yet no prior work has adapted this paradigm to UI-to-Code generation. We propose UIPress, a lightweight learned compression module inserted between the frozen ViT encoder and the LLM decoder of Qwen3-VL-8B. UIPress combines depthwise-separable convolutions, element-guided spatial reweighting, and Transformer refinement to compress ${\sim}$6{,}700 visual tokens to a fixed budget of 256. Together with Low-Rank Adaptation (LoRA) on the decoder to bridge the representation gap, the entire system adds only ${\sim}$21.7M trainable parameters (0.26\% of the 8B base model). Under a fair comparison on the same base model against four baselines on Design2Code, UIPress at 256 tokens achieves a CLIP score of 0.8127, outperforming the uncompressed baseline by +7.5\% and the strongest inference-time method by +4.6\%, while delivering 9.1$\times$ time-to-first-token speedup. To the best of our knowledge, UIPress is the first encoder-side learned compression method for the UI-to-Code task.

In this paper, we propose a sequential recommendation model that integrates Time-aware personalization, Multi-interest personalization, and Explanation personalization for Personalized Sequential Recommendation (TME-PSR). That is, we consider the differences across different users in temporal rhythm preference, multiple fine-grained latent interests, and the personalized semantic alignment between recommendations and explanations. Specifically, the proposed TME-PSR model employs a dual-view gated time encoder to capture personalized temporal rhythms, a lightweight multihead Linear Recurrent Unit architecture that enables fine-grained sub-interest modeling with improved efficiency, and a dynamic dual-branch mutual information weighting mechanism to achieve personalized alignment between recommendations and explanations. Extensive experiments on real-world datasets demonstrate that our method consistently improves recommendation accuracy and explanation quality, at a lower computational cost.

A novel regularization technique, AdaCubic, is proposed that adapts the weight of the cubic term. The heart of AdaCubic is an auxiliary optimization problem with cubic constraints that dynamically adjusts the weight of the cubic term in Newton's cubic regularized method. We use Hutchinson's method to approximate the Hessian matrix, thereby reducing computational cost. We demonstrate that AdaCubic inherits the cubically regularized Newton method's local convergence guarantees. Our experiments in Computer Vision, Natural Language Processing, and Signal Processing tasks demonstrate that AdaCubic outperforms or competes with several widely used optimizers. Unlike other adaptive algorithms that require hyperparameter fine-tuning, AdaCubic is evaluated with a fixed set of hyperparameters, rendering it a highly attractive optimizer in settings where fine-tuning is infeasible. This makes AdaCubic an attractive option for researchers and practitioners alike. To our knowledge, AdaCubic is the first optimizer to leverage cubic regularization in scalable deep learning applications.

Turbulent boundary layers over aerodynamic surfaces are a major source of aircraft drag, yet their control remains challenging due to multiscale dynamics and spatial variability, particularly under adverse pressure gradients. Reinforcement learning has outperformed state-of-the-art strategies in canonical flows, but its application to realistic geometries is limited by computational cost and transferability. Here we show that these limitations can be overcome by exploiting local structures of wall-bounded turbulence. Policies are trained in turbulent channel flows matched to wing boundary-layer statistics and deployed directly onto a NACA4412 wing at $Re_c=2\times10^5$ without further training, being the so-called zero-shot control. This achieves a 28.7\% reduction in skin-friction drag and a 10.7\% reduction in total drag, outperforming the state-of-the-art opposition control by 40\% in friction drag reduction and 5\% in total drag. Training cost is reduced by four orders of magnitude relative to on-wing training, enabling scalable flow control.

Text embeddings are central to modern information retrieval and Retrieval-Augmented Generation (RAG). While dense models derived from Large Language Models (LLMs) dominate current practice, recent work has explored quantum-inspired alternatives motivated by the geometric properties of Hilbert-like spaces and their potential to encode richer semantic structure. This paper presents an experimental framework for constructing quantum-inspired 1024-dimensional document embeddings based on overlapping windows and multi-scale aggregation. The pipeline combines semantic projections (e.g., EigAngle), circuit-inspired feature mappings, and optional teacher-student distillation, together with a fingerprinting mechanism for reproducibility and controlled evaluation. We introduce a set of diagnostic tools for hybrid retrieval, including static and dynamic interpolation between BM25 and embedding-based scores, candidate union strategies, and a conceptual alpha-oracle that provides an upper bound for score-level fusion. Experiments on controlled corpora of Italian and English documents across technical, narrative, and legal domains, using synthetic queries, show that BM25 remains a strong baseline, teacher embeddings provide stable semantic structure, and standalone quantum-inspired embeddings exhibit weak and unstable ranking signals. Distillation yields mixed effects, improving alignment in some cases but not consistently enhancing retrieval performance, while hybrid retrieval can recover competitive results when lexical and embedding-based signals are combined. Overall, the results highlight structural limitations in the geometry of quantum-inspired embeddings, including distance compression and ranking instability, and clarify their role as auxiliary components rather than standalone retrieval representations.

Recovering camera parameters from images and rendering scenes from novel viewpoints have long been treated as separate tasks in computer vision and graphics. This separation breaks down when image coverage is sparse or poses are ambiguous, since each task needs what the other produces. We propose Rays as Pixels, a Video Diffusion Model (VDM) that learns a joint distribution over videos and camera trajectories. We represent each camera as dense ray pixels (raxels) and denoise them jointly with video frames through Decoupled Self-Cross Attention mechanism. A single trained model handles three tasks: predicting camera trajectories from video, jointly generating video and camera trajectory from input images, and generating video from input images along a target camera trajectory. Because the model can both predict trajectories from a video and generate views conditioned on its own predictions, we evaluate it through a closed-loop self-consistency test, demonstrating that its forward and inverse predictions agree. Notably, trajectory prediction requires far fewer denoising steps than video generation, even a few denoising steps suffice for self-consistency. We report results on pose estimation and camera-controlled video generation.

Three-dimensional (3D) data visualizations, such as surface plots, are vital in STEM fields from biomedical imaging to spectroscopy, yet remain largely inaccessible to blind and low-vision (BLV) people. To address this gap, we conducted an Experience-Based Co-Design with BLV co-designers with expertise in non-visual data representations to create an accessible, multi-modal, web-native visualization tool. Using a multi-phase methodology, our team of five BLV and one non-BLV researcher(s) participated in two iterative sessions, comparing a low-fidelity tactile probe with a high-fidelity digital prototype. This process produced a prototype with empirically grounded features, including reference sonification, stereo and volumetric audio, and configurable buffer aggregation, which our co-designers validated as improving analytic accuracy and learnability. In this study, we target core analytic tasks essential for non-visual 3D data exploration: orientation, landmark and peak finding, comparing local maxima versus global trends, gradient tracing, and identifying occluded or partially hidden features. Our work offers accessibility researchers and developers a co-design protocol for translating tactile knowledge to digital interfaces, concrete design guidance for future systems, and opportunities to extend accessible 3D visualization into embodied data environments.

Combinatorial multi-armed bandits provide a fundamental online decision-making environment where a decision-maker interacts with an environment across $T$ time steps, each time selecting an action and learning the cost of that action. The goal is to minimize regret, defined as the loss compared to the optimal fixed action in hindsight under full-information. There has been substantial interest in leveraging what is known about offline algorithm design in this online setting. Offline greedy and linear optimization algorithms (both exact and approximate) have been shown to provide useful guarantees when deployed online. We investigate local search methods, a broad class of algorithms used widely in both theory and practice, which have thus far been under-explored in this context. We focus on problems where offline local search terminates in an approximately optimal solution and give a generic method for converting such an offline algorithm into an online stochastic combinatorial bandit algorithm with $O(\log^3 T)$ (approximate) regret. In contrast, existing offline-to-online frameworks yield regret (and approximate regret) which depend sub-linearly, but polynomially on $T$. We demonstrate the flexibility of our framework by applying it to three online stochastic combinatorial optimization problems: scheduling to minimize total completion time, finding a minimum cost base of a matroid and uncertain clustering.

Artificial spin ice (ASI) are metamaterials composed of interacting nanomagnets. Although ASI hold promise for low-power computing, the ability to transmit information through these two-dimensional systems has been limited. Inspired by non-reciprocal transport in nature, we develop a framework for non-reciprocal influence between nanomagnets. Using the framework we discover a family of ASI geometries with inherent directionality. Directional ASI have the property that, when driven by an external field protocol, domains grow and reverse in the same direction, illustrating an emergent non-reciprocity of the system. Combining growth and reversal results in unidirectional domain movement through the metamaterial. We focus on one member of the directional ASI family, and demonstrate unidirectional domain growth experimentally. Furthermore, we show that the direction of growth is reconfigurable by tuning the external field strengths. Finally, we demonstrate how the directionality of the system significantly improves memory capabilities in a reservoir computing framework. Our work is the first demonstration of an ASI with inherent directionality, offering a magnetic computing platform that combines memory and computation within a single neuromorphic substrate.

Successful machine learning on graphs or networks requires embeddings that not only represent nodes and edges as low-dimensional vectors but also preserve the graph structure. Established methods for generating embeddings require flexible exploration of the entire graph through repeated use of random walks that capture graph structure with samples of nodes and edges. These methods create scalability challenges for massive graphs with millions-to-billions of edges because single-node solutions have inadequate memory and processing capabilities. We present NOMAD, a distributed-memory graph embedding framework using the Message Passing Interface (MPI) for distributed graphs. NOMAD implements proximity-based models proposed in the widely popular LINE (Large-scale Information Network Embedding) algorithm. We propose several practical trade-offs to improve the scalability and communication overheads confronted by irregular and distributed graph embedding methods, catering to massive-scale graphs arising in web and science domains. NOMAD demonstrates median speedups of 10/100x on CPU-based NERSC Perlmutter cluster relative to the popular reference implementations of multi-threaded LINE and node2vec, 35-76x over distributed PBG, and competitive embedding quality relative to LINE, node2vec, and GraphVite, while yielding 12-370x end-to-end speedups on real-world graphs.

This paper studies Automated Instruction Revision (AIR), a rule-induction-based method for adapting large language models (LLMs) to downstream tasks using limited task-specific examples. We position AIR within the broader landscape of adaptation strategies, including prompt optimization, retrieval-based methods, and fine-tuning. We then compare these approaches across a diverse benchmark suite designed to stress different task requirements, such as knowledge injection, structured extraction, label remapping, and logical reasoning. The paper argues that adaptation performance is strongly task-dependent: no single method dominates across all settings. Across five benchmarks, AIR was strongest or near-best on label-remapping classification, while KNN retrieval performed best on closed-book QA, and fine-tuning dominated structured extraction and event-order reasoning. AIR is most promising when task behavior can be captured by compact, interpretable instruction rules, while retrieval and fine-tuning remain stronger in tasks dominated by source-specific knowledge or dataset-specific annotation regularities.

Many-objective optimisation, a subset of multi-objective optimisation, involves optimisation problems with more than three objectives. As the number of objectives increases, the number of solutions needed to adequately represent the entire Pareto front typically grows substantially. This makes it challenging, if not infeasible, to design a search algorithm capable of effectively exploring the entire Pareto front. This difficulty is particularly acute in the Bayesian optimisation paradigm, where sample efficiency is critical and only a limited number of solutions (often a few hundred) are evaluated. Moreover, after the optimisation process, the decision-maker eventually selects just one solution for deployment, regardless of how many high-quality, diverse solutions are available. In light of this, we argue an idea that under a very limited evaluation budget, it may be more useful to focus on finding a single solution of the highest possible quality for the decision-maker, rather than aiming to approximate the entire Pareto front as existing many-/multi-objective Bayesian optimisation methods typically do. Bearing this idea in mind, this paper proposes a \underline{s}ingle \underline{p}oint-based \underline{m}ulti-\underline{o}bjective search framework (SPMO) that aims to improve the quality of solutions along a direction that leads to a good tradeoff between objectives. Within SPMO, we present a simple acquisition function, called expected single-point improvement (ESPI), working under both noiseless and noisy scenarios. We show that ESPI can be optimised effectively with gradient-based methods via the sample average approximation (SAA) approach and theoretically prove its convergence guarantees under the SAA. We also empirically demonstrate that the proposed SPMO is computationally tractable and outperforms state-of-the-arts on a wide range of benchmark and real-world problems.

We present PhysInOne, a large-scale synthetic dataset addressing the critical scarcity of physically-grounded training data for AI systems. Unlike existing datasets limited to merely hundreds or thousands of examples, PhysInOne provides 2 million videos across 153,810 dynamic 3D scenes, covering 71 basic physical phenomena in mechanics, optics, fluid dynamics, and magnetism. Distinct from previous works, our scenes feature multiobject interactions against complex backgrounds, with comprehensive ground-truth annotations including 3D geometry, semantics, dynamic motion, physical properties, and text descriptions. We demonstrate PhysInOne's efficacy across four emerging applications: physics-aware video generation, long-/short-term future frame prediction, physical property estimation, and motion transfer. Experiments show that fine-tuning foundation models on PhysInOne significantly enhances physical plausibility, while also exposing critical gaps in modeling complex physical dynamics and estimating intrinsic properties. As the largest dataset of its kind, orders of magnitude beyond prior works, PhysInOne establishes a new benchmark for advancing physics-grounded world models in generation, simulation, and embodied AI.

Learning-to-Defer routes each input to the expert that minimizes expected cost, but it assumes that the information available to every expert is fixed at decision time. Many modern systems violate this assumption: after selecting an expert, one may also choose what additional information that expert should receive, such as retrieved documents, tool outputs, or escalation context. We study this problem and call it Learning-to-Defer with advice. We show that a broad family of natural separated surrogates, which learn routing and advice with distinct heads, is inconsistent even in the smallest non-trivial setting. We then introduce an augmented surrogate that operates on the composite expert--advice action space and prove an $\mathcal{H}$-consistency guarantee together with an excess-risk transfer bound, yielding recovery of the Bayes-optimal policy in the limit. Experiments on tabular, language, and multi-modal tasks show that the resulting method improves over standard Learning-to-Defer while adapting its advice-acquisition behavior to the cost regime; a synthetic benchmark confirms the failure mode predicted for separated surrogates.

This paper argues that a one-size-fits-all approach to specifying consent for the use of creative works in generative AI is insufficient. Real-world ownership and rights holder structures, the imitation of artistic styles and likeness, and the limitless contexts of use of AI outputs make the status quo of binary consent with opt-in by default untenable. To move beyond the current impasse, we consider levers of control in generative AI workflows at training, inference, and dissemination. Based on these insights, we position inference-time opt-in as an overlooked opportunity for nuanced consent verification. We conceptualize nuanced consent conditions for opt-in and propose an agent-based inference-time opt-in architecture to verify if user intent requests meet conditional consent granted by rights holders. In a case study for music, we demonstrate that nuanced opt-in at inference can account for established rights and re-establish a balance of power between rights holders and AI developers.

We study the population loss landscape of two-layer ReLU networks of the form $\sum_{k=1}^K \mathrm{ReLU}(w_k^\top x)$ in a realisable teacher-student setting with Gaussian covariates. We show that local minima admit an exact low-dimensional representation in terms of summary statistics, yielding a sharp and interpretable characterisation of the landscape. We further establish a direct link with one-pass SGD: local minima correspond to attractive fixed points of the dynamics in summary statistics space. This perspective reveals a hierarchical structure of minima: they are typically isolated in the well-specified regime, but become connected by flat directions as network width increases. In this overparameterised regime, global minima become increasingly accessible, attracting the dynamics and reducing convergence to spurious solutions. Overall, our results reveal intrinsic limitations of common simplifying assumptions, which may miss essential features of the loss landscape even in minimal neural network models.

Software logging is essential for maintaining and debugging complex systems, yet it remains unclear how AI coding agents handle this non-functional requirement. While prior work characterizes human logging practices, the behaviors of AI coding agents and the efficacy of natural language instructions in governing them are unexplored. To address this gap, we conduct an empirical study of 4,550 agentic pull requests across 81 open-source repositories. We compare agent logging patterns against human baselines and analyze the impact of explicit logging instructions. We find that agents change logging less often than humans in 58.4% of repositories, though they exhibit higher log density when they do. Furthermore, explicit logging instructions are rare (4.7%) and ineffective, as agents fail to comply with constructive requests 67% of the time. Finally, we observe that humans perform 72.5% of post-generation log repairs, acting as "silent janitors" who fix logging and observability issues without explicit review feedback. These findings indicate a dual failure in natural language instruction (i.e., scarcity of logging instructions and low agent compliance), suggesting that deterministic guardrails might be necessary to ensure consistent logging practices.

Frontier coding agents solve complex tasks when given complete context but collapse when specifications are incomplete or ambiguous. The bottleneck is not raw capability, but judgment: knowing when to act autonomously and when to ask for help. Current benchmarks are blind to this failure mode. They supply unambiguous detailed instructions and solely reward execution correctness, so an agent that makes a lucky guess for a missing requirement will score identically to one that would have asked to be certain. We present HiL-Bench (Human-in-the-Loop Benchmark) to measure this selective escalation skill. Each task contains human-validated blockers (missing information, ambiguous requests, contradictory information) that surface only through progressive exploration, not upfront inspection. Our core metric, Ask-F1, the harmonic mean of question precision and blocker recall, captures the tension between over-asking and silent guessing; its structure architecturally prevents gaming through question spam. Evaluation across SWE and text-to-SQL domains reveals a large universal judgment gap: no frontier model recovers more than a fraction of its full-information performance when deciding whether to ask. Failure analysis identifies three key help-seeking patterns: overconfident wrong beliefs with no gap detection; high uncertainty detection yet persistent errors; broad, imprecise escalation without self-correction. These consistent patterns confirm poor help-seeking is a model-level flaw, not task-specific. RL training on shaped Ask-F1 reward shows judgment is trainable: a 32B model improves both help-seeking quality and task pass rate, with gains that transfer across domains. The model does not learn domain-specific heuristics for when to ask; it learns to detect unresolvable uncertainty and act on it.

Training large language models (LLMs) is constrained by memory requirements, with activations accounting for a substantial fraction of the total footprint. Existing approaches reduce memory using low-rank weight parameterizations or low-rank gradient subspaces for optimizer states, while activation memory is addressed through architectural modifications or compression schemes based on periodically updated projections. We propose OASIS, an online activation subspace learning algorithm for memory-efficient training that tracks and continuously updates a low-dimensional activation subspace during training. Intermediate activations are projected onto this evolving subspace, reducing memory without modifying forward-pass computations. The evolving activation subspace induces low-rank gradient representations, enabling both gradients and optimizer states to be maintained directly in this subspace, while a projection-aware optimizer consistently transports optimizer states across subspace updates for stable training. Across various finetuning and pretraining tasks, OASIS achieves up to $2\times$ lower peak memory than full fine-tuning while matching its performance and outperforming prior low-rank methods.

Machine unlearning poses challenges in removing mislabeled, contaminated, or problematic data from a pretrained model. Current unlearning approaches and evaluation metrics are solely focused on model predictions, which limits insight into the model's true underlying data characteristics. To address this issue, we introduce a new metric called relearning convergence delay, which captures both changes in weight space and prediction space, providing a more comprehensive assessment of the model's understanding of the forgotten dataset. This metric can be used to assess the risk of forgotten data being recovered from the unlearned model. Based on this, we propose the Influence Eliminating Unlearning framework, which removes the influence of the forgetting set by degrading its performance and incorporates weight decay and injecting noise into the model's weights, while maintaining accuracy on the retaining set. Extensive experiments show that our method outperforms existing metrics and our proposed relearning convergence delay metric, approaching ideal unlearning performance. We provide theoretical guarantees, including exponential convergence and upper bounds, as well as empirical evidence of strong retention and resistance to relearning in both classification and generative unlearning tasks.

Training Transformer language models is expensive, as performance typically improves with increasing dataset size and computational budget. Although scaling laws describe this trend at large scale, their implications in controlled, smaller-scale settings remain less explored. In this work, we isolate dataset-size effects using a strongly reduced attention-only decoder architecture. By training on progressively larger power-of-two subsets, we observe smooth performance improvements accompanied by clear diminishing returns, consistent with scaling-law behavior. Using only about 30% of the training data is sufficient to reach approximately 90% of the full-data validation token-level accuracy. These results provide actionable insights into dataset scaling in a controlled, component-isolated setting and offer practical guidance for balancing dataset size and computational cost in compute- and data-restricted environments, such as small research labs and exploratory model development.

AI coding tools are widely adopted, but most teams plateau at prompt-and-review without a framework for systematic progression. This paper presents the AI Codebase Maturity Model (ACMM), a 5-level framework describing how codebases evolve from basic AI-assisted coding to self-sustaining systems. Inspired by CMMI, each level is defined by its feedback loop topology the specific mechanisms that must exist before the next level becomes possible. I validate the model through a 4-month experience report maintaining KubeStellar Console, a CNCF Kubernetes dashboard built from scratch with Claude Code (Opus) and GitHub Copilot. The system currently operates with 63 CI/CD workflows, 32 nightly test suites, 91% code coverage, and achieves bug-to-fix times under 30 minutes 24 hours a day. The central finding: the intelligence of an AI-driven development system resides not in the AI model itself, but in the infrastructure of instructions, tests, metrics, and feedback loops that surround it. You cannot skip levels, and at each level, the thing that unlocks the next one is another feedback mechanism. Testing the volume of test cases, the coverage thresholds, and the reliability of test execution proved to be the single most important investment in the entire journey.

Agent ecosystems increasingly rely on installable skills to extend functionality, and some skills bundle learned model artifacts as part of their execution logic. This creates a supply-chain risk that is not captured by prompt injection or ordinary plugin misuse: a third-party skill may appear benign while concealing malicious behavior inside its bundled model. We present BadSkill, a backdoor attack formulation that targets this model-in-skill threat surface. In BadSkill, an adversary publishes a seemingly benign skill whose embedded model is backdoor-fine-tuned to activate a hidden payload only when routine skill parameters satisfy attacker-chosen semantic trigger combinations. To realize this attack, we train the embedded classifier with a composite objective that combines classification loss, margin-based separation, and poison-focused optimization, and evaluate it in an OpenClaw-inspired simulation environment that preserves third-party skill installation and execution while enabling controlled multi-model study. Our benchmark spans 13 skills, including 8 triggered tasks and 5 non-trigger control skills, with a combined main evaluation set of 571 negative-class queries and 396 trigger-aligned queries. Across eight architectures (494M--7.1B parameters) from five model families, BadSkill achieves up to 99.5\% average attack success rate (ASR) across the eight triggered skills while maintaining strong benign-side accuracy on negative-class queries. In poison-rate sweeps on the standard test split, a 3\% poison rate already yields 91.7\% ASR. The attack remains effective across the evaluated model scales and under five text perturbation types. These findings identify model-bearing skills as a distinct model supply-chain risk in agent ecosystems and motivate stronger provenance verification and behavioral vetting for third-party skill artifacts.

Large language models (LLMs) exhibit substantial variability in performance and computational cost across tasks and queries, motivating routing systems that select models to meet user-specific cost-performance trade-offs. However, existing routers generalize poorly in cold-start scenarios where in-domain training data is unavailable. We address this limitation with a multi-level task-profile-guided data synthesis framework that constructs a hierarchical task taxonomy and produces diverse question-answer pairs to approximate the test-time query distribution. Building on this, we introduce TRouter, a task-type-aware router approach that models query-conditioned cost and performance via latent task-type variables, with prior regularization derived from the synthesized task taxonomy. This design enhances TRouter's routing utility under both cold-start and in-domain settings. Across multiple benchmarks, we show that our synthesis framework alleviates cold-start issues and that TRouter delivers effective LLM routing.

We propose a Hybrid Quantum-Classical Physics-Informed Neural Network (HQC-PINN) that integrates parameterized variational quantum circuits into the PINN framework for hydrological PDE-constrained learning. Our architecture encodes multi-source remote sensing features into quantum states via trainable angle encoding, processes them through a hardware-efficient variational ansatz with entangling layers, and constrains the output using the Saint-Venant shallow water equations and Manning's flow equation as differentiable physics loss terms. The inherent stochasticity of quantum measurement provides a natural mechanism for uncertainty quantification without requiring explicit Bayesian inference machinery. We further introduce a quantum transfer learning protocol that pre-trains on multi-hazard disaster data before fine-tuning on flood-specific events. Numerical simulations on multi-modal satellite and meteorological data from the Kalu River basin, Sri Lanka, show that the HQC-PINN achieves convergence in ~3x fewer training epochs and uses ~44% fewer trainable parameters compared to an equivalent classical PINN, while maintaining competitive classification accuracy. Theoretical analysis indicates that hydrological physics constraints narrow the effective optimization landscape, providing a natural mitigation against barren plateaus in variational quantum circuits. This work establishes the first application of quantum-enhanced physics-informed learning to hydrological prediction and demonstrates a viable path toward quantum advantage in environmental science.

Generative models can now propose thousands of \emph{de novo} antibody sequences, yet translating these designs into viable therapeutics remains constrained by the cost of biophysical characterization. Here we present CrossAbSense, a framework of property-specific neural oracles that combine frozen protein language model encoders with configurable attention decoders, identified through a systematic hyperparameter campaign totaling over 200 runs per property. On the GDPa1 benchmark of 242 therapeutic IgGs, our oracles achieve notable improvements of 12--20\% over established baselines on three of five developability assays and competitive performance on the remaining two. The central finding is that optimal decoder architectures \emph{invert} our initial biological hypotheses: self-attention alone suffices for aggregation-related properties (hydrophobic interaction chromatography, polyreactivity), where the relevant sequence signatures -- such as CDR-H3 hydrophobic patches -- are already fully resolved within single-chain embeddings by the high-capacity 6B encoder. Bidirectional cross-attention, by contrast, is required for expression yield and thermal stability -- properties that inherently depend on the compatibility between heavy and light chains. Learned chain fusion weights independently confirm heavy-chain dominance in aggregation ($w_H = 0.62$) versus balanced contributions for stability ($w_H = 0.51$). We demonstrate practical utility by deploying CrossAbSense on 100 IgLM-generated antibody designs, illustrating a path toward substantial reduction in experimental screening costs.

When a Vision-Language Model (VLM) sees a blue banana and answers "yellow", is the problem of perception or arbitration? We explore the question in ten VLMs with various sizes and reveal an Encoding--Grounding Dissociation: models that fail to report what they see (and thus provide a wrong answer) still encode the visual evidence as strongly as models that provide the correct answer. Using Multimodal Arbitration Crossover (MAC) analysis with layer-by-layer Logit Lens probing, we track the competition between visual and prior signals across every layer of each model. We show that visual attributes can be linearly decodable from early layers (AUC > 0.86). The accuracy remains nearly identical for both successful and failed samples. However, the gap in the final-layer logit -- not the strength of encoding -- better predicts grounding outcomes with a correlation of . After having studied when VLMs base their answers on image clues rather than prior knowledge, we want to understand the causal relationships. We establish causality through full-sequence activation patching. The standard last-token interventions in LLM interpretability do not affect VLMs. In contrast, replacing the full token sequence at layers identified by MAC alters 60 to 84% of outputs. Partial-token decomposition shows that image tokens carry almost all of the causal impact, while text tokens have none. Scaling addresses the remaining architectural differences to achieve perfect retention. Moving from diagnosis to intervention, we show that training-free activation steering -- both linear and sparse autoencoder-guided -- in early layers can improve visual grounding by up to +3.8% with degrading performance in some setups. Overall, these findings lead to a clear conclusion: VLMs already see well, but the challenge is acting on what they see. Targeted interventions can help to bridge this gap.

In this paper, we propose a stochastic-dimension frozen sampled neural network (SD-FSNN) for solving a class of high-dimensional Gross-Pitaevskii equations (GPEs) on unbounded domains. SD-FSNN is unbiased across all dimensions, and its computational cost is independent of the dimension, avoiding the exponential growth in computational and memory costs associated with Hermite-basis discretizations. Additionally, we randomly sample the hidden weights and biases of the neural network, significantly outperforming iterative, gradient-based optimization methods in terms of training time and accuracy. Furthermore, we employ a space-time separation strategy, using adaptive ordinary differential equation (ODE) solvers to update the evolution coefficients and incorporate temporal causality. To preserve the structure of the GPEs, we integrate a Gaussian-weighted ansatz into the neural network to enforce exponential decay at infinity, embed a normalization projection layer for mass normalization, and add an energy conservation constraint to mitigate long-time numerical dissipation. Comparative experiments with existing methods demonstrate the superior performance of SD-FSNN across a range of spatial dimensions and interaction parameters. Compared to existing random-feature methods, SD-FSNN reduces the complexity from linear to dimension-independent. Additionally, SD-FSNN achieves better accuracy and faster training compared to general high-dimensional solvers, while focusing specifically on high-dimensional GPEs on unbounded domains.

The rapid proliferation of Large Language Model (LLM) providers--each exposing proprietary API formats--has created a fragmented ecosystem where applications become tightly coupled to individual vendors. Switching or bridging providers requires $O(N^2)$ bilateral adapters, impeding portability and multi-provider architectures. We observe that despite substantial syntactic divergence, the major LLM APIs share a common semantic core: the practical challenge is the combinatorial surface of syntactic variations, not deep semantic incompatibility. Based on this finding, we present LLM-Rosetta, an open-source translation framework built on a hub-and-spoke Intermediate Representation (IR) that captures the shared semantic core--messages, content parts, tool calls, reasoning traces, and generation controls--in a 9-type content model and 10-type stream event schema. A modular Ops-composition converter architecture enables each API standard to be added independently. LLM-Rosetta supports bidirectional conversion (provider-to-IR-to-provider) for both request and response payloads, including chunk-level streaming with stateful context management. We implement converters for four API standards (OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, and Google GenAI), covering the vast majority of commercial providers. Empirical evaluation demonstrates lossless round-trip fidelity, correct streaming behavior, and sub-100 microsecond conversion overhead--competitive with LiteLLM's single-pass approach while providing bidirectionality and provider neutrality. LLM-Rosetta passes the Open Responses compliance suite and is deployed in production at Argonne National Laboratory. Code is available at https://github.com/Oaklight/llm-rosetta.

Label noise in multi-label learning (MLL) poses significant challenges for model training, particularly in partial multi-label learning (PML) where candidate labels contain both relevant and irrelevant labels. While clustering offers a natural approach to exploit data structure for noise identification, traditional clustering methods cannot be directly applied to multi-label scenarios due to a fundamental incompatibility: clustering produces membership values that sum to one per instance, whereas multi-label assignments require binary values that can sum to any number. We propose a novel weakly-supervised clustering approach for PML (WSC-PML) that bridges clustering and multi-label learning through membership matrix decomposition. Our key innovation decomposes the clustering membership matrix $\mathbf{A}$ into two components: $\mathbf{A} = \mathbfΠ \odot \mathbf{F}$, where $\mathbfΠ$ maintains clustering constraints while $\mathbf{F}$ preserves multi-label characteristics. This decomposition enables seamless integration of unsupervised clustering with multi-label supervision for effective label noise handling. WSC-PML employs a three-stage process: initial prototype learning from noisy labels, adaptive confidence-based weak supervision construction, and joint optimization via iterative clustering refinement. Extensive experiments on 24 datasets demonstrate that our approach outperforms six state-of-the-art methods across all evaluation metrics.

Accurate prediction of nonstationary multivariate time series remains a critical challenge in complex industrial systems such as iron ore sintering. In practice, pronounced concept drift compounded by significant label verification latency rapidly degrades the performance of offline-trained models. Existing methods based on static architectures or passive update strategies struggle to simultaneously extract multi-scale spatiotemporal features and overcome the stability-plasticity dilemma without immediate supervision. To address these limitations, a Drift-Aware Multi-Scale Dynamic Learning (DA-MSDL) framework is proposed to maintain robust multi-output predictive performance via online adaptive mechanisms on nonstationary data streams. The framework employs a multi-scale bi-branch convolutional network as its backbone to disentangle local fluctuations from long-term trends, thereby enhancing representational capacity for complex dynamic patterns. To circumvent the label latency bottleneck, DA-MSDL leverages Maximum Mean Discrepancy (MMD) for unsupervised drift detection. By quantifying online statistical deviations in feature distributions, DA-MSDL proactively triggers model adaptation prior to inference. Furthermore, a drift-severity-guided hierarchical fine-tuning strategy is developed. Supported by prioritized experience replay from a dynamic memory queue, this approach achieves rapid distribution alignment while effectively mitigating catastrophic forgetting. Long-horizon experiments on real-world industrial sintering data and a public benchmark dataset demonstrate that DA-MSDL consistently outperforms representative baselines under severe concept drift. Exhibiting strong cross-domain generalization and predictive stability, the proposed framework provides an effective online dynamic learning paradigm for quality monitoring in nonstationary environments.

In this letter, we consider the problem of decentralized decision making among connected autonomous vehicles at unsignalized intersections, where existing centralized approaches do not scale gracefully under mixed maneuver intentions and coordinator failure. We propose a closed-loop opinion-dynamic decision model for intersection coordination, where vehicles exchange intent through dual signed networks: a conflict topology based communication network and a commitment-driven belief network that enable cooperation without a centralized coordinator. Continuous opinion states modulate velocity optimizer weights prior to commitment; a closed-form predictive feasibility gate then freezes each vehicle's decision into a GO or YIELD commitment, which propagates back through the belief network to pre-condition neighbor behavior ahead of physical conflicts. Crossing order emerges from geometric feasibility and arrival priority without the use of joint optimization or a solver. The approach is validated across three scenarios spanning fully competitive, merge, and mixed conflict topologies. The results demonstrate collision-free coordination and lower last-vehicle exit times compared to first come first served (FCFS) in all conflict non-trivial configurations.

Reinforcement learning with verifiable rewards (RLVR) has significantly advanced the reasoning ability of vision-language models (VLMs). However, the inherent text-dominated nature of VLMs often leads to insufficient visual faithfulness, characterized by sparse attention activation to visual tokens. More importantly, our empirical analysis reveals that temporal visual forgetting along reasoning steps exacerbates this deficiency. To bridge this gap, we propose Visually-Guided Policy Optimization (VGPO), a novel framework to reinforce visual focus during policy optimization. Specifically, VGPO initially introduces a Visual Attention Compensation mechanism that leverages visual similarity to localize and amplify visual cues, while progressively elevating visual expectations in later steps to counteract visual forgetting. Building on this mechanism, we implement a dual-grained advantage re-weighting strategy: the intra-trajectory level highlights tokens exhibiting relatively high visual activation, while the inter-trajectory level prioritizes trajectories demonstrating superior visual accumulation. Extensive experiments demonstrate that VGPO achieves better visual activation and superior performance in mathematical multimodal reasoning and visual-dependent tasks.

Spatial reasoning is central to navigation and robotics, yet measuring model capabilities on these tasks remains difficult. Existing benchmarks evaluate models in a one-shot setting, requiring full solution generation in a single response, unlike humans, who work in interactive environments step-by-step. We introduce Spatial-Gym, a Gymnasium environment that isolates spatial constraint reasoning by testing pathfinding in 2D-grid puzzles as a sequential decision task with optional backtracking. We evaluate eight models in three settings (one-shot, step-by-step, step-by-step with backtracking) against human, random, and A* baselines on 500 episodes. The best model, GPT-OSS 120B, achieves a solve rate of 16.0%, 82 points below the human baseline (98.0%). Step-by-step format helps weaker models (up to +5.4%) by removing formatting errors, but hurts stronger models (up to 5.6%) by constraining global planning. Backtracking improves episode completion, but increases solve rate only for weaker models; stronger models rarely backtrack and do not benefit from it. Our experiments have three key findings: (1) models fail to scale reasoning effort with difficulty, (2) vision models receiving images of the spatial environment reduce solve rate by 73%, and (3) extended chain-of-thought reasoning retains a 3-5x accuracy advantage over standard inference even in the step-by-step setting. Spatial-Gym enables diagnosis of model limitations and provides a framework for improving spatial reasoning through reinforcement learning.

Accurate prediction of intersection turning movements is essential for adaptive signal control but remains difficult due to the high volatility of directional flows. This study proposes HFD-TM (Hierarchical Flow-Decomposition for Turning Movement Prediction), a hierarchical deep learning framework that predicts turning movements by first forecasting corridor through-movements and then expanding these predictions to individual turning streams. This design is motivated by empirical traffic structure, where corridor flows account for 65.1% of total volume, exhibit lower volatility than turning movements, and explain 35.5% of turning-movement variance. A physics-informed loss function enforces flow conservation to maintain structural consistency. Evaluated on six months of 15-minute interval LiDAR (Light Detection and Ranging) data from a six-intersection corridor in Nashville, Tennessee, HFD-TM achieves a mean absolute error of 2.49 vehicles per interval, reducing MAE by 5.7% compared to a Transformer and by 27.0% compared to a GRU (Gated Recurrent Unit). Ablation results show that hierarchical decomposition provides the largest performance gain, while training time is 12.8 times lower than DCRNN (Diffusion Convolutional Recurrent Neural Network), demonstrating suitability for real-time traffic applications.

A novel stability-enhanced Gaussian process variational autoencoder (SEGP-VAE) is proposed for indirectly training a low-dimensional linear time invariant (LTI) system, using high-dimensional video data. The mean and covariance function of the novel SEGP prior are derived from the definition of an LTI system, enabling the SEGP to capture the indirectly observed latent process using a combined probabilistic and interpretable physical model. The search space of LTI parameters is restricted to the set of semi-contracting systems via a complete and unconstrained parametrisation. As a result, the SEGP-VAE can be trained using unconstrained optimisation algorithms. Furthermore, this parametrisation prevents numerical issues caused by the presence of a non-Hurwitz state matrix. A case study applies SEGP-VAE to a dataset containing videos of spiralling particles. This highlights the benefits of the approach and the application-specific design choices that enabled accurate latent state predictions.

Mechanistic understanding and rational design of complex chemical systems depend on fast and accurate predictions of electronic structures beyond individual building blocks. However, if the system exceeds hundreds of atoms, first-principles quantum mechanical (QM) modeling becomes impractical. In this study, we developed FB-GNN-MBE by integrating a fragment-based graph neural network (FB-GNN) into the many-body expansion (MBE) theory and demonstrated its capacity to reproduce first-principles potential energy surfaces (PES) for hierarchically structured systems with manageable accuracy, complexity, and interpretability. Specifically, we divided the entire system into basic building blocks (fragments), evaluated their one-fragment energies using a QM model, and addressed many-fragment interactions using the structure-property relationships trained by FB-GNNs. Our investigation shows that FB-GNN-MBE achieves chemical accuracy in predicting two-body (2B) and three-body (3B) energies across water, phenol, and mixture benchmarks, as well as the one-dimensional dissociation curves of water and phenol dimers. To transfer the success of FB-GNN-MBE across various systems with minimal computational costs and data demands, we developed and validated a teacher-student learning protocol. A heavy-weight FB-GNN trained on a mixed-density water cluster ensemble (teacher) distills its learned knowledge and passes it to a light-weight GNN (student), which is later fine-tuned on a uniform-density (H2O)21 cluster ensemble. This transfer learning strategy resulted in efficient and accurate prediction of 2B and 3B energies for variously sized water clusters without retraining. Our transferable FB-GNN-MBE framework outperformed conventional non-FB-GNN-based models and showed high practicality for large-scale molecular simulations.

This work introduces an ultralow-power voltage-to-spike encoder that achieves near-linear voltage-to-firing-rate conversion by pairing a linearized bulk-driven transconductor with a DPI-based LIF neuron. A tail-less bulk-driven differential pair improves large-signal linearity, while a translinear linearization network suppresses the dominant sinh nonlinearity and stabilizes the bias-tunable V-to-I gain. The resulting current feeds a DPI front-end that linearizes current-to-spike conversion. Fabricated in TSMC 0.18-um CMOS and operating at VDD = 0.5 V with 2-27 nA reference current, the encoder achieves a deviation of less than 5.6 percent from linearity over 0.1-0.4 V input, consumes 22-180 nW, and occupies 0.0074 mm^2.

The Half-Trek Criterion (HTC) is the primary graphical tool for determining generic identifiability of causal effect coefficients in linear structural equation models (SEMs) with latent confounders. However, HTC is inherently node-wise: it simultaneously resolves all incoming edges of a node, leaving a gap of "inconclusive" causal effects (15-23% in moderate graphs). We introduce Iterative Identification Closure (IIC), a general framework that decouples causal identification into two phases: (1) a seed function S_0 that identifies an initial set of edges from any external source of information (instrumental variables, interventions, non-Gaussianity, prior knowledge, etc.); and (2) Reduced HTC propagation that iteratively substitutes known coefficients to reduce system dimension, enabling identification of edges that standard HTC cannot resolve. The core novelty is iterative identification propagation: newly identified edges feed back to unlock further identification -- a mechanism absent from all existing graphical criteria, which treat each edge (or node) in isolation. This propagation is non-trivial: coefficient substitution alters the covariance structure, and soundness requires proving that the modified Jacobian retains generic full rank -- a new theoretical result (Reduced HTC Theorem). We prove that IIC is sound, monotone, converges in O(|E|) iterations (empirically <=2), and strictly subsumes both HTC and ancestor decomposition. Exhaustive verification on all graphs with n<=5 (134,144 edges) confirms 100% precision (zero false positives); with combined seeds, IIC reduces the HTC gap by over 80%. The propagation gain is gamma~4x (2 seeds identifying ~3% of edges to 97.5% total identification), far exceeding gamma<=1.2x of prior methods that incorporate side information without iterative feedback.

Large language models are making autonomous drug discovery agents increasingly feasible, but reliable success in this setting is not determined by any single action or molecule. It is determined by whether the final returned set jointly satisfies protocol-level requirements such as set size, diversity, binding quality, and developability. This creates a fundamental control problem: the agent plans step by step, while task validity is decided at the level of the whole candidate set. Existing language-based drug discovery systems therefore tend to rely on long raw history and under-specified self-reflection, making failure localization imprecise and planner-facing agent states increasingly noisy. We present CACM (Constraint-Aware Corrective Memory), a language-based drug discovery framework built around precise set-level diagnosis and a concise memory write-back mechanism. CACM introduces protocol auditing and a grounded diagnostician, which jointly analyze multimodal evidence spanning task requirements, pocket context, and candidate-set evidence to localize protocol violations, generate actionable remediation hints, and bias the next action toward the most relevant correction. To keep planning context compact, CACM organizes memory into static, dynamic, and corrective channels and compresses them before write-back, thereby preserving persistent task information while exposing only the most decision-relevant failures. Our experimental results show that CACM improves the target-level success rate by 36.4% over the state-of-the-art baseline. The results show that reliable language-based drug discovery benefits not only from more powerful molecular tools, but also from more precise diagnosis and more economical agent states.

The sustainability impacts of ICT systems are difficult to assess and govern due to structural complexity, fragmented measurement practices, and unclear responsibilities across system layers. We argue that these challenges cannot be addressed solely by metrics and motivate the need for a shared Green ICT reference framework that integrates sustainability across multiple perspectives and domains, lifecycle phases, and governance contexts. We present an initial framework developed within the Informatics Europe Green ICT Working Group as a first step towards a comprehensive reference framework.

Quantum networks are expected to become a key enabler for interconnecting quantum devices. In contrast to classical communication networks, however, information transfer in quantum networks is usually restricted to short distances due to physical constraints of entanglement distribution. Satellites can extend entanglement distribution over long distances, but routing in such networks is challenging because satellite motion and stochastic link generation create a highly dynamic quantum topology. Existing routing methods often rely on global topology information that quickly becomes outdated due to delays in the classical control plane, while decentralized methods typically act on incomplete local information. We propose SatQNet, a reinforcement learning approach for entanglement routing in satellite-assisted quantum networks that can be decentralized at runtime. Its key innovation is an edge-centric directed line graph neural network that performs local message passing on directed edge embeddings, enabling it to better capture link properties in high-degree and time-varying topologies. By exchanging messages with neighboring repeaters, SatQNet learns a local graph representation at runtime that supports agents in establishing high-fidelity end-to-end entanglements. Trained on random graphs, SatQNet outperforms heuristic and learning-based approaches across diverse settings, including a real-world European backbone topology, and generalizes to unseen topologies without retraining.

This paper presents an online intention prediction framework for estimating the goal state of autonomous systems in real time, even when intention is time-varying, and system dynamics or objectives include unknown parameters. The problem is formulated as an inverse optimal control / inverse reinforcement learning task, with the intention treated as a parameter in the objective. A shifting horizon strategy discounts outdated information, while online control-informed learning enables efficient gradient computation and online parameter updates. Simulations under varying noise levels and hardware experiments on a quadrotor drone demonstrate that the proposed approach achieves accurate, adaptive intention prediction in complex environments.

Agent skills provide modular, task-specific guidance for LLM- based coding agents, but manually tuning skill bundles to balance success rate, cost, and runtime is expensive and fragile. We present SkillMOO, a multi-objective optimization framework that automatically evolves skill bundles using LLM-proposed edits and NSGA-II survivor selection: a solver agent evaluates candidate skill bundles on coding tasks and an optimizer agent proposes bundle edits based on failure analysis. On three SkillsBench software engineering tasks, SkillMOO improves pass rate by up to 131% while reducing cost up to 32% relative to the best baseline per task at low optimization overhead. Pattern analysis reveals pruning and substitution as primary drivers of improvement, suggesting effective bundles favor minimal, focused content over accumulated instructions.

We propose a hybrid physics-informed framework for solving families of parametric linear partial differential equations (PDEs) by combining a meta-learned predictor with a least-squares corrector. The predictor, termed \textbf{KAPI} (Kernel-Adaptive Physics-Informed meta-learner), is a shallow task-conditioned model that maps query coordinates and PDE parameters to solution values while internally generating an interpretable, task-adaptive Gaussian basis geometry. A lightweight meta-network maps PDE parameters to basis centers, widths, and activity patterns, thereby learning how the approximation space should adapt across the parametric family. This predictor-generated geometry is transferred to a second-stage corrector, which augments it with a background basis and computes the final solution through a one-shot physics-informed Extreme Learning Machine (PIELM)-style least-squares solve. We evaluate the method on four linear PDE families spanning diffusion, transport, mixed advection--diffusion, and variable-speed transport. Across these cases, the predictor captures meaningful physics through localized and transport-aligned basis placement, while the corrector further improves accuracy, often by one or more orders of magnitude. Comparisons with parametric PINNs, physics-informed DeepONet, and uniform-grid PIELM correctors highlight the value of predictor-guided basis adaptation as an interpretable and efficient strategy for parametric PDE solving.

Trusted multi-view classification typically relies on a view-wise evidential fusion process: each view independently produces class evidence and uncertainty, and the final prediction is obtained by aggregating these independent opinions. While this design is modular and uncertainty-aware, it implicitly assumes that evidence from different views is numerically comparable. In practice, however, this assumption is fragile. Different views often differ in feature space, noise level, and semantic granularity, while independently trained branches are optimized only for prediction correctness, without any constraint enforcing cross-view consistency in evidence strength. As a result, the uncertainty used for fusion can be dominated by branch-specific scale bias rather than true sample-level reliability. To address this issue, we propose Trusted Multi-view learning with Unified Routing (TMUR), which decouples view-specific evidence extraction from fusion arbitration. TMUR uses view-private experts and one collaborative expert, and employs a unified router that observes the global multi-view context to generate sample-level expert weights. Soft load-balancing and diversity regularization further encourage balanced expert utilization and more discriminative expert specialization. We also provide theoretical analysis showing why independent evidential supervision does not identify a common cross-view evidence scale, and why unified global routing is preferable to branch-local arbitration when reliability is sample-dependent.

The development of Large Language Models (LLMs) has catalyzed automation in customer service, yet benchmarking their performance remains challenging. Existing benchmarks predominantly rely on static paradigms and single-dimensional metrics, failing to account for diverse user behaviors or the strict adherence to structured Standard Operating Procedures (SOPs) required in real-world deployments. To bridge this gap, we propose SAGE (Service Agent Graph-guided Evaluation), a universal multi-agent benchmark for automated, dual-axis assessment. SAGE formalizes unstructured SOPs into Dynamic Dialogue Graphs, enabling precise verification of logical compliance and comprehensive path coverage. We introduce an Adversarial Intent Taxonomy and a modular Extension Mechanism, enabling low-cost deployment across domains and facilitating automated dialogue data synthesis. Evaluation is conducted via a framework where Judge Agents and a Rule Engine analyze interactions between User and Service Agents to generate deterministic ground truth. Extensive experiments on 27 LLMs across 6 industrial scenarios reveal a significant ``Execution Gap'' where models accurately classify intents but fail to derive correct subsequent actions. We also observe ``Empathy Resilience'', a phenomenon where models maintain polite conversational facades despite underlying logical failures under high adversarial intensity. Code and resources are available at https://anonymous.4open.science/r/SAGE-Bench-4CD3/.

Enabling observability in software systems brings many benefits. It can, for example, ease the identification of issues or the implementation of improvements. It is especially critical to be able to observe sustainability-related dimensions of systems to know and mitigate their impact. However, adding observability to a system, especially related to software sustainability, requires technical knowledge that may not be available in every project that would benefit from it. In this work, we propose an architectural blueprint along with its deployment code that can be used to facilitate the addition of observability in software systems. As a special case, it includes measuring the energy consumption of software. This toolkit provides support in defining which components are necessary for a given use case and for structuring their deployment. Moreover, we exemplify the addition of observability in two different use cases.

Distributed online convex optimization (D-OCO) is a powerful paradigm for modeling distributed scenarios with streaming data. However, the communication cost between local learners and the central server is substantial in large-scale applications. To alleviate this bottleneck, we initiate the study of D-OCO with compressed communication. Firstly, to quantify the compression impact, we establish the $Ω(δ^{-1/2}\sqrt{T})$ and $Ω(δ^{-1}\log{T})$ lower bounds for convex and strongly convex loss functions, respectively, where $δ\in (0,1]$ is the compression ratio. Secondly, we propose an optimal algorithm, which enjoys regret bounds of $O(δ^{-1/2}\sqrt{T})$ and $O(δ^{-1} \log T)$ for convex and strongly convex loss functions, respectively. Our method incorporates the error feedback mechanism into the Follow-the-Regularized-Leader framework to address the coupling between the compression error and the projection error. Furthermore, we employ the online compression strategy to mitigate the accumulated error arising from the bidirectional compression. Our online method has great generality, and can be extended to the offline stochastic setting via online-to-batch conversion. We establish convergence rates of $O(δ^{-1/2}T^{-1/2})$ and $O(δ^{-1} T^{-1})$ for convex and strongly convex loss functions, respectively, providing the first guarantees for distributed non-smooth optimization with compressed communication and domain constraints.

The transition to electric mobility hinges on maximising aggregate adoption while also facilitating equitable access. This study examines whether the 'charging divide' between households with and without off-street parking reflects a genuine infrastructure constraint or a by-product of socio-economic disparity. Moving beyond conventional predictive models, we apply a probabilistic causal framework to a nationally representative dataset of Scottish households, enabling estimation of policy interventions while explicitly neutralising the confounding effect of other causal factors. The results reveal a structural hierarchy in the EV adoption process. Private off-street parking functions as a conversion catalyst: enabling access to home-charging increases the probability of EV ownership from 3.3% to 5.6% (a 70% relative, 2.3 percentage point absolute increase). However, this effect primarily accelerates households already economically positioned to purchase an EV rather than recruiting new entrants. By contrast, household income operates as the fundamental affordability ceiling. A causal contrast between lower- and higher-income strata, shows a reduction in market non-participation by 23.1 percentage points, identifying financial capacity as the principal gatekeeper to entering the EV transition funnel. Crucially, the analysis demonstrates that standard observational models overstate the isolated effect of off-street parking infrastructure. The apparent effect emerges from selection bias: higher-income households are disproportionately likely to possess both private parking and the means to purchase EVs. These findings support a dual-track policy strategy: lowering the affordability ceiling for non-participants through financial instruments, while addressing EV home-charging access for the 'latent intent' cohort in high-density urban contexts.

Intelligent dialogue systems are increasingly deployed in emotionally and ethically sensitive settings, where failures in either emotional attunement or ethical judgment can cause significant harm. Existing dialogue models typically address empathy and ethical safety in isolation, and often fail to adapt their behavior as ethical risk and user emotion evolve across multi-turn interactions. We formulate ethical-emotional alignment in dialogue as an explicit turn-level decision problem, and propose \textsc{EthicMind}, a risk-aware framework that implements this formulation in multi-turn dialogue at inference time. At each turn, \textsc{EthicMind} jointly analyzes ethical risk signals and user emotion, plans a high-level response strategy, and generates context-sensitive replies that balance ethical guidance with emotional engagement, without requiring additional model training. To evaluate alignment behavior under ethically complex interactions, we introduce a risk-stratified, multi-turn evaluation protocol with a context-aware user simulation procedure. Experimental results show that \textsc{EthicMind} achieves more consistent ethical guidance and emotional engagement than competitive baselines, particularly in high-risk and morally ambiguous scenarios.

We consider machine learning tasks with low-rank functional tree tensor networks (TTN) as the learning model. While in the case of least-squares regression, low-rank functional TTNs can be efficiently optimized using alternating optimization, this is not directly possible in other problems, such as multinomial logistic regression. We propose a natural Riemannian gradient descent type approach applicable to arbitrary losses which is based on the natural gradient by Amari. In particular, the search direction obtained by the natural gradient is independent of the choice of basis of the underlying functional tensor product space. Our framework applies to both the factorized and manifold-based approach for representing the functional TTN. For practical application, we propose a hierarchy of efficient approximations to the true natural Riemannian gradient for computing the updates in the parameter space. Numerical experiments confirm our theoretical findings on common classification datasets and show that using natural Riemannian gradient descent for learning considerably improves convergence behavior when compared to standard Riemannian gradient methods.

Standard object detectors typically treat architectural elements independently, often resulting in facade parsings that lack the structural coherence required for downstream procedural reconstruction. We address this limitation by augmenting the YOLOv8 training objective with a custom lightweight alignment loss. This regularization encourages grid-consistent arrangements of bounding boxes during training, effectively injecting geometric priors without altering the standard inference pipeline. Experiments on the CMP dataset demonstrate that our method successfully improves structural regularity, correcting alignment errors caused by perspective and occlusion while maintaining a controllable trade-off with standard detection accuracy.

Pretraining is the cornerstone of Large Language Models (LLMs), dominating the vast majority of computational budget and data to serve as the primary engine for their capabilities. During pretraining, LLMs acquire foundational knowledge from an unprecedentedly massive and diverse data sources, encompassing a vast array of domains such as general language, mathematics, code, and complex reasoning. In this work, we investigate an interesting geometric question regarding the converged state of pretraining: Does the model converge to a common minimizer across all data sources (e.g., \cref{fig:cwa_illustration:close}), or merely a minimizer of the summed loss (e.g., \cref{fig:cwa_illustration:distant})? We hypothesize that the geometric "closeness" of task-specific minima is intrinsically linked to downstream generalization. We reveal that standard optimizers (e.g., AdamW) often converge to points where task-specific minima are distant from each other. To address this, we propose the Nexus optimizer, which encourages the closeness of these minima by maximizing gradient similarity during optimization. Experiments across models ranging from 130M to 3B parameters, various data mixtures and hyperparameter schedules, show that Nexus \textit{significantly boosts downstream performance}, despite \textit{achieving the same pretraining loss} (see \cref{fig:demo:benchmark}). Notably, on the 3B model, Nexus reduces the out-of-distribution loss by 0.012 and yields up to a 15.0\% accuracy improvement on complex reasoning tasks (e.g., GSM8k). This finding challenges the reliance on pretraining loss as the sole proxy for model evaluation and demonstrates the importance of implicit biases in unlocking downstream generalization.

Vision-Language Models (VLMs) are powerful but remain vulnerable to multimodal jailbreak attacks. Existing attacks mainly rely on either explicit visual prompt attacks or gradient-based adversarial optimization. While the former is easier to detect, the latter produces subtle perturbations that are less perceptible, but is usually optimized and evaluated under homogeneous open-source surrogate-target settings, leaving its effectiveness on commercial closed-source VLMs under heterogeneous settings unclear. To examine this issue, we study different surrogate-target settings and observe a consistent gap between homogeneous and heterogeneous settings, a phenomenon we term surrogate dependency. Motivated by this finding, we propose Mosaic, a Multi-view ensemble optimization framework for multimodal jailbreak against closed-source VLMs, which alleviates surrogate dependency under heterogeneous surrogate-target settings by reducing over-reliance on any single surrogate model and visual view. Specifically, Mosaic incorporates three core components: a Text-Side Transformation module, which perturbs refusal-sensitive lexical patterns; a Multi-View Image Optimization module, which updates perturbations under diverse cropped views to avoid overfitting to a single visual view; and a Surrogate Ensemble Guidance module, which aggregates optimization signals from multiple surrogate VLMs to reduce surrogate-specific bias. Extensive experiments on safety benchmarks demonstrate that Mosaic achieves state-of-the-art Attack Success Rate and Average Toxicity against commercial closed-source VLMs.

Deep research agents increasingly interleave web browsing with multi-step computation, yet existing benchmarks evaluate these capabilities in isolation, creating a blind spot in assessing real-world performance. We introduce DRBENCHER, a synthetic benchmark generator for questions that require both browsing and computation. It enforces four criteria: verifiability (gold answers are computed by executing parameterized code over knowledge-graph values), complexity (multi-hop entity identification, property retrieval, and domain-specific computation), difficulty (a two-stage verification cascade filters out questions solvable by the generating model), and diversity (a greedy max-min embedding filter maximizes coverage). These criteria are realized via a unified answer-first pipeline spanning five domains: biochemistry, financial, geophysical, security, and history. Human evaluation shows 76% validity (84% excluding stale data), with 35% of errors due to outdated knowledge-graph entries, highlighting an inherent limitation of systems that reason over evolving data. Automatic evaluation shows that the strongest frontier model achieves only 20% answer accuracy. Compared to manually constructed benchmarks (BrowseComp+, MATH-500, GPQA), DRBENCHER achieves the highest semantic diversity.

Differentiable Digital Signal Processing (DDSP) pipelines for voice conversion rely on subtractive synthesis, where a periodic excitation signal is shaped by a learned spectral envelope to reconstruct the target voice. In DDSP-QbE, the excitation is generated via phase accumulation, producing a sawtooth-like waveform whose abrupt discontinuities introduce aliasing artefacts that manifest perceptually as buzziness and spectral distortion, particularly at higher fundamental frequencies. We propose two targeted improvements to the excitation stage of the DDSP-QbE subtractive synthesizer. First, we incorporate explicit voicing detection to gate the harmonic excitation, suppressing the periodic component in unvoiced regions and replacing it with filtered noise, thereby avoiding aliased harmonic content where it is most perceptually disruptive. Second, we apply Polynomial Band-Limited Step (PolyBLEP) correction to the phase-accumulated oscillator, substituting the hard waveform discontinuity at each phase wrap with a smooth polynomial residual that cancels alias-generating components without oversampling or spectral truncation. Together, these modifications yield a cleaner harmonic roll-off, reduced high-frequency artefacts, and improved perceptual naturalness, as measured by MOS. The proposed approach is lightweight, differentiable, and integrates seamlessly into the existing DDSP-QbE training pipeline with no additional learnable parameters.

High-Level Synthesis (HLS) compiles C/C++ into RTL, but exploring pragma-driven optimization choices remains expensive because each design point requires time-consuming synthesis. We propose \textbf{\DiffHLS}, a differential learning framework for HLS Quality-of-Result (QoR) prediction that learns from kernel--design pairs: a kernel baseline and a pragma-inserted design variant. \DiffHLS~encodes kernel and design intermediate-representation graphs with dedicated graph neural network (GNN) branches, and augments the delta pathway with code embeddings from a pretrained code large language model (LLM). Instead of regressing absolute targets directly, we jointly predict the kernel baseline and the design-induced delta, and compose them to obtain the design prediction. On PolyBench, \DiffHLS~attains lower average MAPE than GNN baselines under four GNN backbones, and LLM code embeddings consistently improve over a GNN-only ablation. We further validate scalability on the ForgeHLS dataset.

Many disciplines pose natural-language research questions over large document collections whose answers typically require structured evidence, traditionally obtained by manually designing an annotation schema and exhaustively labeling the corpus, a slow and error-prone process. We introduce ScheMatiQ, which leverages calls to a backbone LLM to take a question and a corpus to produce a schema and a grounded database, with a web interface that lets steer and revise the extraction. In collaboration with domain experts, we show that ScheMatiQ yields outputs that support real-world analysis in law and computational biology. We release ScheMatiQ as open source with a public web interface, and invite experts across disciplines to use it with their own data. All resources, including the website, source code, and demonstration video, are available at: www.ScheMatiQ-ai.com

The King Wen sequence of the I-Ching (c. 1000 BC) orders 64 hexagrams -- states of a six-dimensional binary space -- in a pattern that has puzzled scholars for three millennia. We present a rigorous statistical characterization of this ordering using Monte Carlo permutation analysis against 100,000 random baselines. We find that the sequence has four statistically significant properties: higher-than-random transition distance (98.2nd percentile), negative lag-1 autocorrelation (p=0.037), yang-balanced groups of four (p=0.002), and asymmetric within-pair vs. between-pair distances (99.2nd percentile). These properties superficially resemble principles from curriculum learning and curiosity-driven exploration, motivating the hypothesis that they might benefit neural network training. We test this hypothesis through three experiments: learning rate schedule modulation, curriculum ordering, and seed sensitivity analysis, conducted across two hardware platforms (NVIDIA RTX 2060 with PyTorch and Apple Silicon with MLX). The results are uniformly negative. King Wen LR modulation degrades performance at all tested amplitudes. As curriculum ordering, King Wen is the worst non-sequential ordering on one platform and within noise on the other. A 30-seed sweep confirms that only King Wen's degradation exceeds natural seed variance. We explain why: the sequence's high variance -- the very property that makes it statistically distinctive -- destabilizes gradient-based optimization. Anti-habituation in a fixed combinatorial sequence is not the same as effective training dynamics.

LiDAR-based perception is critical for autonomous driving due to its robustness to poor lighting and visibility conditions. Yet, current models operate under the closed-set assumption and often fail to recognize unexpected out-of-distribution (OOD) objects in the open world. Existing OOD scoring functions exhibit limited performance because they ignore the pronounced class imbalance inherent in LiDAR OOD detection and assume a uniform class distribution. To address this limitation, we propose the Neural Distribution Prior (NDP), a framework that models the distributional structure of network predictions and adaptively reweights OOD scores based on alignment with a learned distribution prior. NDP dynamically captures the logit distribution patterns of training data and corrects class-dependent confidence bias through an attention-based module. We further introduce a Perlin noise-based OOD synthesis strategy that generates diverse auxiliary OOD samples from input scans, enabling robust OOD training without external datasets. Extensive experiments on the SemanticKITTI and STU benchmarks demonstrate that NDP substantially improves OOD detection performance, achieving a point-level AP of 61.31\% on the STU test set, which is more than 10$\times$ higher than the previous best result. Our framework is compatible with various existing OOD scoring formulations, providing an effective solution for open-world LiDAR perception.

Von Economo neurons (VENs) are large bipolar projection neurons found exclusively in the anterior cingulate cortex (ACC) and frontal insula of species with complex social cognition, including humans, great apes, and cetaceans. Their selective depletion in frontotemporal dementia (FTD) and altered development in autism implicate them in rapid social decision-making, yet no computational model of VEN function has previously existed. We introduce the Fast Lane Hypothesis: VENs implement a biological speed-accuracy tradeoff (SAT) by providing a sparse, fast projection pathway that enables rapid social decisions at the cost of deliberate processing accuracy. We model VENs as fast leaky integrate-and-fire (LIF) neurons with membrane time constant 5 ms and sparse dendritic fan-in of eight afferents, compared to 20 ms and eighty afferents for standard pyramidal neurons, within a spiking cortical circuit of 2,000 neurons trained on a social discrimination task. Networks are evaluated under three clinically motivated conditions across 10 independent random seeds: typical (2% VENs), autism-like (0.4% VENs), and FTD-like (post-training VEN ablation). All configurations achieve equivalent asymptotic classification accuracy (99.4%), consistent with the prediction that VENs modulate decision speed rather than representational capacity. Temporal analysis confirms that VENs produce median first-spike latencies 4 ms earlier than pyramidal neurons. At a fixed decision threshold, the typical condition is significantly faster than FTD-like (t=-23.31, p<0.0001), while autism-like is intermediate (mean RT=26.91+/-9.01 ms vs. typical 20.70+/-2.02 ms; p=0.078). A preliminary evolutionary analysis shows qualitative correspondence between model-optimal VEN fraction and the primate phylogenetic gradient. To our knowledge, this is the first computational model that asks what a Von Economo neuron actually computes.

Audio large language models (ALLMs) enable rich speech-text interaction, but they also introduce jailbreak vulnerabilities in the audio modality. Existing audio jailbreak methods mainly optimize jailbreak success while overlooking utility preservation, as reflected in transcription quality and question answering performance. In practice, stronger attacks often come at the cost of degraded utility. To study this trade-off, we revisit existing attacks by varying their perturbation coverage in the frequency domain, from partial-band to full-band, and find that broader frequency coverage does not necessarily improve jailbreak performance, while utility consistently deteriorates. This suggests that concentrating perturbation on a subset of bands can yield a better attack-utility trade-off than indiscriminate full-band coverage. Based on this insight, we propose GRM, a utility-aware frequency-selective jailbreak framework. It ranks Mel bands by their attack contribution relative to utility sensitivity, perturbs only a selected subset of bands, and learns a reusable universal perturbation under a semantic-preservation objective. Experiments on four representative ALLMs show that GRM achieves an average Jailbreak Success Rate (JSR) of 88.46% while providing a better attack-utility trade-off than representative baselines. These results highlight the potential of frequency-selective perturbation for better balancing attack effectiveness and utility preservation in audio jailbreak. Content Warning: This paper includes harmful query examples and unsafe model responses.

Large language models are increasingly deployed in multi-turn settings such as tutoring, support, and counseling, where reliability depends on preserving consistent roles, personas, and goals across long horizons. This requirement becomes critical when LLMs are used to generate synthetic dialogues for training and evaluation, since LLM--LLM conversations can accumulate identity-related failures such as persona drift, role confusion, and "echoing", where one agent gradually mirrors its partner. We introduce SPASM (Stable Persona-driven Agent Simulation for Multi-turn dialogue generation), a modular, stability-first framework that decomposes simulation into (i) persona creation via schema sampling, plausibility validation, and natural-language persona crafting, (ii) Client--Responder dialogue generation, and (iii) termination detection for coherent stopping. To improve long-horizon stability without changing model weights, we propose Egocentric Context Projection (ECP): dialogue history is stored in a perspective-agnostic representation and deterministically projected into each agent's egocentric view before generation. Across three LLM backbones (GPT-4o-mini, DeepSeek-V3.2, Qwen-Plus) and nine Client--Responder pairings, we construct a dataset of 4,500 personas and 45,000 conversations (500 personas X 10 conversations per pairing). Ablations show ECP substantially reduces persona drift and, under human validation, eliminates echoing; embedding analyses recover persona structure and reveal strong responder-driven interaction geometry. Our code is available at https://github.com/lhannnn/SPASM.

We develop a predictive-first optimisation framework for streaming hidden Markov models. Unlike classical approaches that prioritise full posterior recovery under a fully specified generative model, we assume access to regime-specific predictive models whose parameters are learned online while maintaining a fixed transition prior over regimes. Our objective is to sequentially identify latent regimes while maintaining accurate step-ahead predictive distributions. Because the number of possible regime paths grows exponentially, exact filtering is infeasible. We therefore formulate streaming inference as a constrained projection problem in predictive-distribution space: under a fixed hypothesis budget, we approximate the full posterior predictive by the forward-KL optimal mixture supported on $S$ paths. The solution is the renormalised top-$S$ posterior-weighted mixture, providing a principled derivation of beam search for HMMs. The resulting algorithm is fully recursive and deterministic, performing beam-style truncation with closed-form predictive updates and requiring neither EM nor sampling. Empirical comparisons against Online EM and Sequential Monte Carlo under matched computational budgets demonstrate competitive prequential performance.

Cloud providers must assign heterogeneous compute resources to workflow DAGs while balancing competing objectives such as completion time, cost, and energy consumption. In this work, we study a single-workflow, queue-free scheduling setting and consider a graph neural network (GNN)-based deep reinforcement learning scheduler designed to minimize workflow completion time and energy usage. We identify specific out-of-distribution (OOD) conditions under which GNN-based deep reinforcement learning schedulers fail and provide a principled explanation of why these failures occur. Through controlled OOD evaluations, we demonstrate that performance degradation stems from structural mismatches between training and deployment environments, which disrupt message passing and undermine policy generalization. Our analysis exposes fundamental limitations of current GNN-based schedulers and highlights the need for more robust representations to ensure reliable scheduling performance under distribution shifts.

There is substantial concern about the ability of advanced artificial intelligence to influence people's behaviour. A rapidly growing body of research has found that AI can produce large persuasive effects on people's attitudes, but whether AI can persuade people to take consequential real-world actions has remained unclear. In two large preregistered experiments N=17,950 responses from 14,779 people), we used conversational AI models to persuade participants on a range of attitudinal and behavioural outcomes, including signing real petitions and donating money to charity. We found sizable AI persuasion effects on these behavioural outcomes (e.g. +19.7 percentage points on petition signing). However, we observed no evidence of a correlation between AI persuasion effects on attitudes and behaviour. Moreover, we replicated prior findings that information provision drove effects on attitudes, but found no such evidence for our behavioural outcomes. In a test of eight behavioural persuasion strategies, all outperformed the most effective attitudinal persuasion strategy, but differences among the eight were small. Taken together, these results suggest that previous findings relying on attitudinal outcomes may generalize poorly to behaviour, and therefore risk substantially mischaracterizing the real-world behavioural impact of AI persuasion.

Purpose. High-grade serous ovarian carcinoma (HGSOC) is characterized by pronounced biological and spatial heterogeneity and is frequently diagnosed at an advanced stage. Neoadjuvant chemotherapy (NACT) followed by delayed primary surgery is commonly employed in patients unsuitable for primary cytoreduction. The Chemotherapy Response Score (CRS) is a validated histopathological biomarker of response to NACT, but it is only available postoperatively. In this study, we investigate whether pre-treatment computed tomography (CT) imaging and clinical data can be used to predict CRS as an investigational decision-support adjunct to inform multidisciplinary team (MDT) discussions regarding expected treatment response. Methods. We proposed a 2.5D multimodal deep learning framework that processes lesion-dense omental slices using a pre-trained Vision Transformer encoder and integrates the resulting visual representations with clinical variables through an intermediate fusion module to predict CRS. Results. Our multimodal model, integrating imaging and clinical data, achieved a ROC-AUC of 0.95 alongside 95% accuracy and 80% precision on the internal test cohort (IEO, n=41 patients). On the external test set (OV04, n=70 patients), it achieved a ROC-AUC of 0.68, alongside 67% accuracy and 75% precision. Conclusion. These preliminary results demonstrate the feasibility of transformer-based deep learning for preoperative prediction of CRS in HGSOC using routine clinical data and CT imaging. As an investigational, pre-treatment decision-support tool, this approach may assist MDT discussions by providing early, non-invasive estimates of treatment response.

We propose Camera Artist, a multi-agent framework that models a real-world filmmaking workflow to generate narrative videos with explicit cinematic language. While recent multi-agent systems have made substantial progress in automating filmmaking workflows from scripts to videos, they often lack explicit mechanisms to structure narrative progression across adjacent shots and deliberate use of cinematic language, resulting in fragmented storytelling and limited filmic quality. To address this, Camera Artist builds upon established agentic pipelines and introduces a dedicated Cinematography Shot Agent, which integrates recursive storyboard generation to strengthen shot-to-shot narrative continuity and cinematic language injection to produce more expressive, film-oriented shot designs. Extensive quantitative and qualitative results demonstrate that our approach consistently outperforms existing baselines in narrative consistency, dynamic expressiveness, and perceived film quality.

LLMs internalize safety policies through RLHF, yet these policies are never formally specified and remain difficult to inspect. Existing benchmarks evaluate models against external standards but do not measure whether models understand and enforce their own stated boundaries. We introduce the Symbolic-Neural Consistency Audit (SNCA), a framework that (1) extracts a model's self-stated safety rules via structured prompts, (2) formalizes them as typed predicates (Absolute, Conditional, Adaptive), and (3) measures behavioral compliance via deterministic comparison against harm benchmarks. Evaluating four frontier models across 45 harm categories and 47,496 observations reveals systematic gaps between stated policy and observed behavior: models claiming absolute refusal frequently comply with harmful prompts, reasoning models achieve the highest self-consistency but fail to articulate policies for 29% of categories, and cross-model agreement on rule types is remarkably low (11%). These results demonstrate that the gap between what LLMs say and what they do is measurable and architecture-dependent, motivating reflexive consistency audits as a complement to behavioral benchmarks.

Diffusion models and their variations, such as rectified flows, generate diverse and high-quality images, but they are still hindered by slow iterative sampling caused by the highly curved generative paths they learn. An important cause of high curvature, as shown by previous work, is independence between the source distribution (standard Gaussian) and the data distribution. In this work, we tackle this limitation by two complementary contributions. First, we attempt to break away from the standard Gaussian assumption by introducing $κ\texttt{-FC}$, a general formulation that conditions the source distribution on an arbitrary signal $κ$ that aligns it better with the data distribution. Then, we present MixFlow, a simple but effective training strategy that reduces the generative path curvatures and considerably improves sampling efficiency. MixFlow trains a flow model on linear mixtures of a fixed unconditional distribution and a $κ\texttt{-FC}$-based distribution. This simple mixture improves the alignment between the source and data, provides better generation quality with less required sampling steps, and accelerates the training convergence considerably. On average, our training procedure improves the generation quality by 12\% in FID compared to standard rectified flow and 7\% compared to previous baselines under a fixed sampling budget. Code available at: $\href{https://github.com/NazirNayal8/MixFlow}{https://github.com/NazirNayal8/MixFlow}$

We develop a theory of generalization and scaling for Mixture-of-Experts (MoE) Transformers that cleanly separates \emph{active} per-input capacity from routing combinatorics. By conditioning on fixed routing patterns and union-bounding across them, we derive a sup-norm covering-number bound whose metric entropy scales with the active parameter budget and incurs a MoE-specific routing overhead. Combined with a standard ERM analysis for squared loss, this yields a generalization bound under a $d$-dimensional manifold data model and $C^β$ targets, showing that approximation and estimation trade off as in dense networks once active parameters are accounted for appropriately. We further prove a constructive approximation theorem for MoE architectures, showing that, under the approximation construction, error can decrease either by scaling active capacity or by increasing the number of experts, depending on the dominant bottleneck. From these results we derive neural scaling laws for model size, data size, and compute-optimal tradeoffs. Overall, our results provide a transparent statistical reference point for reasoning about MoE scaling, clarifying which behaviors are certified by worst-case theory and which must arise from data-dependent routing structure or optimization dynamics.

Retrieval-Augmented Generation (RAG) aims to reduce hallucination by grounding answers in retrieved evidence, yet hallucinated answers remain common even when relevant documents are available. Existing evaluations focus on answer-level or passage-level accuracy, offering limited insight into how evidence is used during generation. In this work, we introduce a facet-level diagnostics framework for QA that decomposes each input question into atomic reasoning facets. For each facet, we assess evidence sufficiency and grounding using a structured Facet x Chunk matrix that combines retrieval relevance with natural language inference-based faithfulness scores. To diagnose evidence usage, we analyze three controlled inference modes: Strict RAG, which enforces exclusive reliance on retrieved evidence; Soft RAG, which allows integration of retrieved evidence and parametric knowledge; and LLM-only generation without retrieval. Comparing these modes enables thorough analysis of retrieval-generation misalignment, defined as cases where relevant evidence is retrieved but not correctly integrated during generation. Across medical QA and HotpotQA, we evaluate three open-source and closed-source LLMs (GPT, Gemini, and LLaMA), providing interpretable diagnostics that reveal recurring facet-level failure modes, including evidence absence, evidence misalignment, and prior-driven overrides. Our results demonstrate that hallucinations in RAG systems are driven less by retrieval accuracy and more by how retrieved evidence is integrated during generation, with facet-level analysis exposing systematic evidence override and misalignment patterns that remain hidden under answer-level evaluation.

Search-Based Software Testing (SBST) automates test input generation but is frequently hindered by challenging fitness landscapes characterized by numerous deceptive local optima that impede search progress, as well as extended plateaus where informative fitness signals are scarce. To address this bottleneck, we propose SHIFT (Sigmoid-Based Heuristic Invertible Fitness-Landscape Transformation for Accelerating SBST), a method designed to compress local landscapes and facilitate escape from stagnant regions without altering global semantics. By systematically contracting dense regions where search points cluster, the approach preserves mapping invertibility while enabling optimization algorithms to traverse more effectively toward global coverage with the same step size. When evaluated against established baselines, including pure hill climbing and genetic algorithms, under a normalized experimental protocol, the proposed technique yields consistent improvements in convergence speed and search efficiency. These results demonstrate that sigmoid compression constitutes a lightweight yet effective mechanism for achieving more reliable coverage discovery in complex testing environments.

Vision-language models (VLMs) have achieved strong performance in multimodal understanding and reasoning, yet grounded reasoning in 3D scenes remains underexplored. Effective 3D reasoning hinges on accurate grounding: to answer open-ended queries, a model must first identify query-relevant objects and regions in a complex scene, and then reason about their spatial and geometric relationships. Recent approaches have demonstrated strong potential for grounded 3D reasoning. However, they often rely on in-domain tuning or hand-crafted reasoning pipelines, which limit their flexibility and zero-shot generalization to novel environments. In this work, we present MAG-3D, a training-free multi-agent framework for grounded 3D reasoning with off-the-shelf VLMs. Instead of relying on task-specific training or fixed reasoning procedures, MAG-3D dynamically coordinates expert agents to address the key challenges of 3D reasoning. Specifically, we propose a planning agent that decomposes the task and orchestrates the overall reasoning process, a grounding agent that performs free-form 3D grounding and relevant frame retrieval from extensive 3D scene observations, and a coding agent that conducts flexible geometric reasoning and explicit verification through executable programs. This multi-agent collaborative design enables flexible training-free 3D grounded reasoning across diverse scenes and achieves state-of-the-art performance on challenging benchmarks.

Anomaly detection (AD) in chemical processes based on deep learning offers significant opportunities but requires large, diverse, and well-annotated training datasets that are rarely available from industrial operations. In a recent work, we introduced a large, fully annotated experimental dataset for batch distillation under normal and anomalous operating conditions. In the present study, we augment this dataset with a corresponding simulation dataset, creating a novel hybrid dataset. The simulation data is generated in an automated workflow with a novel Python-based process simulator that employs a tailored index-reduction strategy for the underlying differential-algebraic equations. Leveraging the rich metadata and structured anomaly annotations of the experimental database, experimental records are automatically translated into simulation scenarios. After calibration to a single reference experiment, the dynamics of the other experiments are well predicted. This enabled the fully automated, consistent generation of time-series data for a large number of experimental runs, covering both normal operation and a wide range of actuator- and control-related anomalies. The resulting hybrid dataset is released openly. From a process simulation perspective, this work demonstrates the automated, consistent simulation of large-scale experimental campaigns, using batch distillation as an example. From a data-driven AD perspective, the hybrid dataset provides a unique basis for simulation-to-experiment style transfer, the generation of pseudo-experimental data, and future research on deep AD methods in chemical process monitoring.

Most affective computing research treats emotion as a static property of text, focusing on the writer's sentiment while overlooking the reader's perspective. This approach ignores how individual personalities lead to diverse emotional appraisals of the same event. Although role-playing Large Language Models (LLMs) attempt to simulate such nuanced reactions, they often suffer from "personality illusion'' -- relying on surface-level stereotypes rather than authentic cognitive logic. A critical bottleneck is the absence of ground-truth human data to link personality traits to emotional shifts. To bridge the gap, we introduce Persona-E$^2$ (Persona-Event2Emotion), a large-scale dataset grounded in annotated MBTI and Big Five traits to capture reader-based emotional variations across news, social media, and life narratives. Extensive experiments reveal that state-of-the-art LLMs struggle to capture precise appraisal shifts, particularly in social media domains. Crucially, we find that personality information significantly improves comprehension, with the Big Five traits alleviating "personality illusion.'

Maximum entropy reinforcement learning (MaxEnt RL) has become a standard framework for sequential decision making, yet its standard Gaussian policy parameterization is inherently unimodal, limiting its ability to model complex multimodal action distributions. This limitation has motivated increasing interest in generative policies based on diffusion and flow matching as more expressive alternatives. However, incorporating such policies into MaxEnt RL is challenging for two main reasons: the likelihood and entropy of continuous-time generative policies are generally intractable, and multi-step sampling introduces both long-horizon backpropagation instability and substantial inference latency. To address these challenges, we propose Truncated Rectified Flow Policy (TRFP), a framework built on a hybrid deterministic-stochastic architecture. This design makes entropy-regularized optimization tractable while supporting stable training and effective one-step sampling through gradient truncation and flow straightening. Empirical results on a toy multigoal environment and 10 MuJoCo benchmarks show that TRFP captures multimodal behavior effectively, outperforms strong baselines on most benchmarks under standard sampling, and remains highly competitive under one-step sampling.

Supporting students in developing diagnostic reasoning is a key challenge across educational domains. Novices often face cognitive biases such as premature closure and over-reliance on heuristics, and they struggle to transfer diagnostic strategies to new cases. Scenario-based learning (SBL) enhanced by Learning Analytics (LA) and large language models (LLM) offers a promising approach by combining realistic case experiences with personalized scaffolding. Yet, how different scaffolding approaches shape reasoning processes remains insufficiently explored. This study introduces PharmaSim Switch, an SBL environment for pharmacy technician training, extended with an LA- and LLM-powered pharmacist agent that implements pedagogical conversations rooted in two theory-driven scaffolding approaches: \emph{structuring} and \emph{problematizing}, as well as a student learning trajectory. In a between-groups experiment, 63 vocational students completed a learning scenario, a near-transfer scenario, and a far-transfer scenario under one of the two scaffolding conditions. Results indicate that both scaffolding approaches were effective in supporting the use of diagnostic strategies. Performance outcomes were primarily influenced by scenario complexity rather than students' prior knowledge or the scaffolding approach used. The structuring approach was associated with more accurate Active and Interactive participation, whereas problematizing elicited more Constructive engagement. These findings underscore the value of combining scaffolding approaches when designing LA- and LLM-based systems to effectively foster diagnostic reasoning.

We introduce a novel learning framework for accelerated Monte Carlo (MC) dose calculation termed Energy-Shifting. This approach leverages deep learning to synthesize 6 MV TrueBeam Linear Accelerator (LINAC) dose distributions directly from monoenergetic inputs under identical beam configurations. Unlike conventional denoising techniques, which rely on noisy low-count dose maps that compromise beam profile integrity, our method achieves superior cross-domain generalization on unseen datasets by integrating high-fidelity anatomical textures and source-specific beam similarity into the model's input space. Furthermore, we propose a novel 3D architecture termed TransUNetSE3D, featuring Transformer blocks for global context and Residual Squeeze-and-Excitation (SE) modules for adaptive channel-wise feature recalibration. Hierarchical representations of these blocks are fused into the network's latent space alongside the primary dose-map parameters, allowing physics-aware reconstruction. This hybrid design outperforms existing UNet and Transformer-based benchmarks in both spatial precision and structural preservation, while maintaining the execution speed necessary for real-time use. Our proposed pipeline achieves a Gamma Passing Rate exceeding 98% (3%/3mm) compared to the MC reference, evaluated within the framework of a treatment planning system (TPS) for prostate radiotherapy. These results offer a robust solution for fast volumetric dosimetry in adaptive radiotherapy.

Graphical user interface (GUI) agents powered by vision language models (VLMs) are rapidly moving from passive assistance to autonomous operation. However, this unrestricted action space exposes users to severe and irreversible financial, privacy or social harm. Existing safeguards rely on prompt engineering, brittle heuristics and VLM-as-critic lack formal verification and user-tunable guarantees. We propose CORA (COnformal Risk-controlled GUI Agent), a post-policy, pre-action safeguarding framework that provides statistical guarantees on harmful executed actions. CORA reformulates safety as selective action execution: we train a Guardian model to estimate action-conditional risk for each proposed step. Rather than thresholding raw scores, we leverage Conformal Risk Control to calibrate an execute/abstain boundary that satisfies a user-specified risk budget and route rejected actions to a trainable Diagnostician model, which performs multimodal reasoning over rejected actions to recommend interventions (e.g., confirm, reflect, or abort) to minimize user burden. A Goal-Lock mechanism anchors assessment to a clarified, frozen user intent to resist visual injection attacks. To rigorously evaluate this paradigm, we introduce Phone-Harm, a new benchmark of mobile safety violations with step-level harm labels under real-world settings. Experiments on Phone-Harm and public benchmarks against diverse baselines validate that CORA improves the safety--helpfulness--interruption Pareto frontier, offering a practical, statistically grounded safety paradigm for autonomous GUI execution. Code and benchmark are available at cora-agent.github.io.

Large Reasoning Models (LRMs) achieve strong performance on complex tasks by leveraging long Chain-of-Thought (CoT), but often suffer from overthinking, leading to excessive reasoning steps and high inference latency. Existing CoT compression methods struggle to balance accuracy and efficiency, and lack fine-grained, step-level adaptation to redundancy and reasoning bias. Therefore, we propose State-Aware Reasoning Compression with Knowledge Guidance (STACK), a framework that performs step-wise CoT compression by explicitly modeling stage-specific redundancy sources and integrating with a retrieval-augmented guidance. STACK constructs online long-short contrastive samples and dynamically switches between knowledge-guided compression for uncertain or biased reasoning state and self-prompted compression for overly long but confident state, complemented by an answer-convergence-based early stopping mechanism to suppress redundant verification. We further propose a reward-difference-driven training strategy by combining Proximal Policy Optimization (PPO) and Direct Preference Optimization (DPO), enabling models to learn state-conditioned compression strategies. Experiments on three mathematical reasoning benchmarks show that STACK achieves a superior accuracy-efficiency balance, reducing average response length by 59.9% while improving accuracy by 4.8 points over existing methods.

This paper introduces a score-driven rating system, a generalization of the classical Elo rating system that employs the score, i.e. the gradient of the log-likelihood, as the updating mechanism for player and team ratings. The proposed framework extends beyond simple win/loss game outcomes and accommodates a wide range of game results, such as point differences, win/draw/loss outcomes, or complete rankings. Theoretical properties of the score are derived, showing that it has zero expected value, sums to zero across all players, and decreases with increasing value of a player's rating, thereby ensuring internal consistency and fairness. Furthermore, the score-driven rating system exhibits a reversion property, meaning that ratings tend to follow the underlying unobserved true skills over time. The proposed framework provides a theoretical rationale for existing dynamic models of sports performance and offers a systematic approach for constructing new ones.

Unobserved confounding is a key challenge when estimating causal effects from a treatment on an outcome in scientific applications. In this work, we assume that we observe a single, potentially multi-dimensional proxy variable of the unobserved confounder and that we know the mechanism that generates the proxy from the confounder. Under a completeness assumption on this mechanism, which we call Single Proxy Identifiability of Causal Effects or simply SPICE, we prove that causal effects are identifiable. We extend the proxy-based causal identifiability results by Kuroki and Pearl (2014); Pearl (2010) to higher dimensions, more flexible functional relationships and a broader class of distributions. Further, we develop a neural network based estimation framework, SPICE-Net, to estimate causal effects, which is applicable to both discrete and continuous treatments.

As $SE(3)$-equivariant graph neural networks mature as a core tool for 3D atomistic modeling, improving their efficiency, expressivity, and physical consistency has become a central challenge for large-scale applications. In this work, we introduce EquiformerV3, the third generation of the $SE(3)$-equivariant graph attention Transformer, designed to advance all three dimensions: efficiency, expressivity, and generality. Building on EquiformerV2, we have the following three key advances. First, we optimize the software implementation, achieving $1.75\times$ speedup. Second, we introduce simple and effective modifications to EquiformerV2, including equivariant merged layer normalization, improved feedforward network hyper-parameters, and attention with smooth radius cutoff. Third, we propose SwiGLU-$S^2$ activations to incorporate many-body interactions for better theoretical expressivity and to preserve strict equivariance while reducing the complexity of sampling $S^2$ grids. Together, SwiGLU-$S^2$ activations and smooth-cutoff attention enable accurate modeling of smoothly varying potential energy surfaces (PES), generalizing EquiformerV3 to tasks requiring energy-conserving simulations and higher-order derivatives of PES. With these improvements, EquiformerV3 trained with the auxiliary task of denoising non-equilibrium structures (DeNS) achieves state-of-the-art results on OC20, OMat24, and Matbench Discovery.

Deploying DNNs on System-on-Chips (SoC) with multiple heterogeneous acceleration engines is challenging, and the majority of deployment frameworks cannot fully exploit heterogeneity. We present MATCHA, a unified DNN deployment framework that generates highly concurrent schedules for parallel, heterogeneous accelerators and uses constraint programming to optimize L3/L2 memory allocation and scheduling. Using pattern matching, tiling, and mapping across individual HW units enables parallel execution and high accelerator utilization. On the MLPerf Tiny benchmark, using a SoC with two heterogeneous accelerators, MATCHA improves accelerator utilization and reduces inference latency by up to 35% with respect to the the state-of-the-art MATCH compiler.

Aspect Sentiment Triplet Extraction (ASTE) aims to extract all sentiment triplets of aspect terms, opinion terms, and sentiment polarities from a sentence. Existing methods are typically trained on individual datasets in isolation, failing to jointly capture the common feature representations shared across domains. Moreover, data privacy constraints prevent centralized data aggregation. To address these challenges, we propose Prototype-based Cross-Domain Span Prototype extraction (PCD-SpanProto), a prototype-regularized federated learning framework to enable distributed clients to exchange class-level prototypes instead of full model parameters. Specifically, we design a weighted performance-aware aggregation strategy and a contrastive regularization module to improve the global prototype under domain heterogeneity and the promotion between intra-class compactness and inter-class separability across clients. Extensive experiments on four ASTE datasets demonstrate that our method outperforms baselines and reduces communication costs, validating the effectiveness of prototype-based cross-domain knowledge transfer.

Recent years have witnessed remarkable progress in automatic speech recognition (ASR), driven by advances in model architectures and large-scale training data. However, two important aspects remain underexplored. First, Word Error Rate (WER), the dominant evaluation metric for decades, treats all words equally and often fails to reflect the semantic correctness of an utterance at the sentence level. Second, interactive correction-an essential component of human communication-has rarely been systematically studied in ASR research. In this paper, we integrate these two perspectives under an agentic framework for interactive ASR. We propose leveraging LLM-as-a-Judge as a semantic-aware evaluation metric to assess recognition quality beyond token-level accuracy. Furthermore, we design an LLM-driven agent framework to simulate human-like multi-turn interaction, enabling iterative refinement of recognition outputs through semantic feedback. Extensive experiments are conducted on standard benchmarks, including GigaSpeech (English), WenetSpeech (Chinese), the ASRU 2019 code-switching test set. Both objective and subjective evaluations demonstrate the effectiveness of the proposed framework in improving semantic fidelity and interactive correction capability. We will release the code to facilitate future research in interactive and agentic ASR.

While much prior work examines Large Language Models (LLMs) for solo development tasks (e.g., coding), far less is known about how LLMs shape collaborative group work in software engineering. This study focuses on one such collaborative task, namely software design. It presents the results of an exploratory laboratory study of 18 pairs of software professionals who could use an LLM however they saw fit, to design a University campus bicycle parking application. Our findings reveal that introducing an LLM leads to distinct patterns of joint use: shared-instance use facilitated shared understanding, whereas parallel use across separate instances sometimes led to ''context drift''. We also observe wide variation in reliance, from non-use to treating the LLM as an information source or producer. Across these modes, professionals scrutinized and reflected on LLM responses, often yielding design insights; however, early anchoring sometimes curtailed exploration. We provide implications for tools to aid designers while retaining the human-centricity important to design.

Composed image retrieval (CIR) aims to retrieve a target image that depicts a reference image modified by a textual description. While recent vision-language models (VLMs) achieve promising CIR performance by embedding images and text into a shared space for retrieval, they often fail to reason about what to preserve and what to change. This limitation hinders interpretability and yields suboptimal results, particularly in fine-grained domains like fashion. In this paper, we introduce FIRE-CIR, a model that brings compositional reasoning and interpretability to fashion CIR. Instead of relying solely on embedding similarity, FIRE-CIR performs question-driven visual reasoning: it automatically generates attribute-focused visual questions derived from the modification text, and verifies the corresponding visual evidence in both reference and candidate images. To train such a reasoning system, we automatically construct a large-scale fashion-specific visual question answering dataset, containing questions requiring either single- or dual-image analysis. During retrieval, our model leverages this explicit reasoning to re-rank candidate results, filtering out images inconsistent with the intended modifications. Experimental results on the Fashion IQ benchmark show that FIRE-CIR outperforms state-of-the-art methods in retrieval accuracy. It also provides interpretable, attribute-level insights into retrieval decisions.

Recently, artificial intelligence-based dubbing technology has advanced, enabling automated dubbing (AD) to convert the source speech of a video into target speech in different languages. However, natural AD still faces synchronization challenges such as duration and lip-synchronization (lip-sync), which are crucial for preserving the viewer experience. Therefore, this paper proposes a synchronization method for AD processes that paraphrases translated text, comprising two steps: isochrony for timing constraints and phonetic synchronization (PS) to preserve lip-sync. First, we achieve isochrony by paraphrasing the translated text with a language model, ensuring the target speech duration matches that of the source speech. Second, we introduce PS, which employs dynamic time warping (DTW) with local costs of vowel distances measured from training data so that the target text composes vowels with pronunciations similar to source vowels. Third, we extend this approach to PSComet, which jointly considers semantic and phonetic similarity to preserve meaning better. The proposed methods are incorporated into text-to-speech systems, PS-TTS and PS-Comet TTS. The performance evaluation using Korean and English lip-reading datasets and a voice-actor dubbing dataset demonstrates that both systems outperform TTS without PS on several objective metrics and outperform voice actors in Korean-to-English and English-to-Korean dubbing. We extend the experiments to French, testing all pairs among these languages to evaluate cross-linguistic applicability. Across all language pairs, PS-Comet performed best, balancing lip-sync accuracy with semantic preservation, confirming that PS-Comet achieves more accurate lip-sync with semantic preservation than PS alone.

Modern LLM reinforcement learning (RL) workloads require a highly efficient weight transfer system to scale training across heterogeneous computational resources. However, existing weight transfer approaches either fail to provide flexibility for dynamically scaling clusters or incur fundamental data movement overhead, resulting in poor performance. We introduce Reference-Oriented Storage (ROS), a new storage abstraction for RL weight transfer that exploits the highly replicated model weights in place. ROS presents the illusion that certain versions of the model weights are stored and can be fetched on demand. Underneath, ROS does not physically store any copies of the weights; instead, it tracks the workers that hold these weights on GPUs for inference. Upon request, ROS directly uses them to serve reads. We build TensorHub, a production-quality system that extends the ROS idea with topology-optimized transfer, strong consistency, and fault tolerance. Evaluation shows that TensorHub fully saturates RDMA bandwidth and adapts to three distinct rollout workloads with minimal engineering effort. Specifically, TensorHub reduces total GPU stall time by up to 6.7x for standalone rollouts, accelerates weight update for elastic rollout by 4.8x, and cuts cross-datacenter rollout stall time by 19x. TensorHub has been deployed in production to support cutting-edge RL training.

Diffusion models are known for generating high-quality images, causing serious security concerns. To combat this, most efforts rely on deep neural networks (e.g., CNNs and Transformers), while largely overlooking the potential of traditional machine learning models. In this paper, we freshly investigate such alternatives and proposes a novel Dynamic Assembly Forest model (DAF) to detect diffusion-generated images. Built upon the deep forest paradigm, DAF addresses the inherent limitations in feature learning and scalable training, making it an effective diffusion-generated image detector. Compared to existing DNN-based methods, DAF has significantly fewer parameters, much lower computational cost, and can be deployed without GPUs, while achieving competitive performance under standard evaluation protocols. These results highlight the strong potential of the proposed method as a practical substitute for heavyweight DNN models in resource-constrained scenarios. Our code and models are available at https://github.com/OUC-VAS/DAF.

Scheming, the covert pursuit of misaligned goals by AI systems, represents a potentially catastrophic risk, yet scheming research suffers from significant limitations. In particular, scheming evaluations demonstrate behaviours that may not occur in real-world settings, limiting scientific understanding, hindering policy development, and not enabling real-time detection of loss of control incidents. Real-world evidence is needed, but current monitoring techniques are not effective for this purpose. This paper introduces a novel open-source intelligence (OSINT) methodology for detecting real-world scheming incidents: collecting and analysing transcripts from chatbot conversations or command-line interactions shared online. Analysing over 183,420 transcripts from X (formerly Twitter), we identify 698 real-world scheming-related incidents between October 2025 and March 2026. We observe a statistically significant 4.9x increase in monthly incidents from the first to last month, compared to a 1.7x increase in posts discussing scheming. We find evidence of multiple scheming-related behaviours in real-world deployments previously reported only in experiments, many resulting in real-world harms. While we did not detect catastrophic scheming incidents, the behaviours observed demonstrate concerning precursors, such as willingness to disregard instructions, circumvent safeguards, lie to users, and single-mindedly pursue goals in harmful ways. As AI systems become more capable, these could evolve into more strategic scheming with potentially catastrophic consequences. Our findings demonstrate the viability of transcript-based OSINT as a scalable approach to real-world scheming detection supporting scientific research, policy development, and emergency response. We recommend further investment towards OSINT techniques for monitoring scheming and loss of control.

Organisations with limited data and computational resources increasingly outsource model training to Machine Learning as a Service (MLaaS) providers, who adapt vision-language models (VLMs) such as CLIP to downstream tasks via prompt tuning rather than training from scratch. This semi-honest setting creates a security risk where a malicious provider can follow the prompt-tuning protocol yet implant a backdoor, forcing triggered inputs to be classified into an attacker-chosen class, even for out-of-distribution (OOD) data. Such backdoors leave encoders untouched, making them undetectable to existing methods that focus on encoder corruption. Other data-level methods that sanitize data before training or during inference, also fail to answer the critical question, "Is the delivered model backdoored or not?" To address this model-level verification problem, we introduce CLIP-Inspector (CI), a backdoor detection method designed for prompt-tuned CLIP models. Assuming white-box access to the delivered model and a pool of unlabeled OOD images, CI reconstructs possible triggers for each class to determine if the model exhibits backdoor behaviour or not. Additionally, we demonstrate that using CI's reconstructed trigger for fine-tuning on correctly labeled triggered inputs enables us to re-align the model and reduce backdoor effectiveness. Through extensive experiments across ten datasets and four backdoor attacks, we demonstrate that CI can reconstruct effective triggers in a single epoch using only 1,000 OOD images, achieving a 94% detection accuracy (47/50 models). Compared to adapted trigger-inversion baselines, CI yields a markedly higher AUROC score (0.973 vs 0.495/0.687), thus enabling the vetting and post-hoc repair of prompt-tuned CLIP models to ensure safe deployment.

Automated algorithm selection in continuous black-box optimisation typically relies on fixed landscape descriptors computed under a limited probing budget, yet such descriptors can degrade under problem-split or cross-benchmark evaluation. We propose GeoPAS, a geometric probing approach that represents a problem instance by multiple coarse two-dimensional slices sampled across locations, orientations, and logarithmic scales. A shared validity-aware convolutional encoder maps each slice to an embedding, conditions it on slice-scale and amplitude statistics, and aggregates the resulting features permutation-invariantly for risk-aware solver selection via log-scale performance prediction with an explicit penalty on tail failures. On COCO/BBOB with a 12-solver portfolio in dimensions 2--10, GeoPAS improves over the single best solver under leave-instance-out, grouped random, and leave-problem-out evaluation. These results suggest that multi-scale geometric slices provide a useful transferable static signal for algorithm selection, although a small number of heavy-tail regimes remain and continue to dominate the mean. Our code is available at $\href{https://github.com/BradWangW/GeoPAS}{GitHub}$.

Abusive speech detection is becoming increasingly important as social media shifts towards voice-based interaction, particularly in multilingual and low-resource settings. Most current systems rely on automatic speech recognition (ASR) followed by text-based hate speech classification, but this pipeline is vulnerable to transcription errors and discards prosodic information carried in speech. We investigate whether Contrastive Language-Audio Pre-training (CLAP) can support abusive speech detection directly from audio. Using the ADIMA dataset, we evaluate CLAP-based representations under few-shot supervised contrastive adaptation in cross-lingual and leave-one-language-out settings, with zero-shot prompting included as an auxiliary analysis. Our results show that CLAP yields strong cross-lingual audio representations across ten Indic languages, and that lightweight projection-only adaptation achieves competitive performance with respect to fully supervised systems trained on complete training data. However, the benefits of few-shot adaptation are language-dependent and not monotonic with shot size. These findings suggest that contrastive audio-text models provide a promising basis for cross-lingual audio abuse detection in low-resource settings, while also indicating that transfer remains incomplete and language-specific in important ways.

The use of synthetic data in machine learning applications and research offers many benefits, including performance improvements through data augmentation, privacy preservation of original samples, and reliable method assessment with fully synthetic data. This work proposes a time-efficient synthetic data generation method based on a fully connected neural network and a randomized loss function that transforms a random Gaussian distribution to approximate a target real-world dataset. The experiments conducted on 25 diverse tabular real-world datasets confirm that the proposed solution surpasses the state-of-the-art generative methods and achieves reference MMD scores orders of magnitude faster than modern deep learning solutions. The experiments involved analyzing distributional similarity, assessing the impact on classification quality, and using PCA for dimensionality reduction, which further enhances data privacy and can boost classification quality while reducing time and memory complexity.

Large Language Models (LLMs) for code generation can replicate insecure patterns from their training data. To mitigate this, a common strategy for security hardening is to fine-tune models using supervision derived from the final transformer layer. However, this design may suffer from a final-layer bottleneck: vulnerability-discriminative cues can be distributed across layers and become less detectable near the output representations optimized for next-token prediction. To diagnose this issue, we perform layer-wise linear probing. We observe that vulnerability-related signals are most detectable in a band of intermediate-to-upper layers yet attenuate toward the final layers. Motivated by this observation, we introduce DeepGuard, a framework that leverages distributed security-relevant cues by aggregating representations from multiple upper layers via an attention-based module. The aggregated signal powers a dedicated security analyzer within a multi-objective training objective that balances security enhancement and functional correctness, and further supports a lightweight inference-time steering strategy. Extensive experiments across five code LLMs demonstrate that DeepGuard improves the secure-and-correct generation rate by an average of 11.9% over strong baselines such as SVEN. It also preserves functional correctness while exhibiting generalization to held-out vulnerability types. Our code is public at https://github.com/unknownhl/DeepGuard.

Large-scale digital platforms generate billions of timestamped user-item interactions (events) that are crucial for predicting user attributes in, e.g., fraud prevention and recommendations. While self-supervised learning (SSL) effectively models the temporal order of events, it typically overlooks the global structure of the user-item interaction graph. To bridge this gap, we propose three model-agnostic strategies for integrating this structural information into contrastive SSL: enriching event embeddings, aligning client representations with graph embeddings, and adding a structural pretext task. Experiments on four financial and e-commerce datasets demonstrate that our approach consistently improves the accuracy (up to a 2.3% AUC) and reveals that graph density is a key factor in selecting the optimal integration strategy.

Deploying large language models (LLMs) on mobile devices is an emerging trend to enable data privacy and offline accessibility of LLM applications. Modern mobile neural processing units (NPUs) make such deployment increasingly feasible. However, existing mobile LLM inference frameworks suffer from high start-up latency due to their inevitable cold starts, i.e., launching LLM inferences when the model is not hosted in device memory. In this paper, we identify the key bottleneck of mobile LLM cold starts as the waste of flash bandwidth on unimportant model parameters. We design EdgeFlow, a mobile LLM inference framework that mitigates the cold start issue by adaptively adjusting the precisions of LLM parameters. Specifically, EdgeFlow leverages 1) an NPU-aware adaptive quantization algorithm that assigns different precisions to weights in a finer granularity according to their importance and NPU constraints, 2) an SIMD-friendly packing format that accelerates the transformation of various-precision weights into fixed-sized NPU-native data types, and 3) a synergistic granular pipeline that coordinates CPU and NPU computation in a fine-grained and dynamic manner. Experimental results show that EdgeFlow reduces cold-start latency by up to 4.07x compared with three state-of-the-art mobile LLM inference frameworks, i.e., llama.cpp, MNN, and llm.npu, under comparable model accuracy.

Large language models increasingly operate under multiple instructions from heterogeneous sources with different authority levels, including system policies, user requests, tool outputs, and retrieved context. While prior work on instruction hierarchy highlights the importance of respecting instruction priorities, it mainly focuses on adversarial attacks and overlooks the benign but common instruction conflicts that arise in real-world applications. In such settings, models must not only avoid security violations but also preserve task utility and behavioral consistency when instructions partially or implicitly conflict. We propose Neuro-Symbolic Hierarchical Alignment (NSHA) for hierarchical instruction-following by explicitly modeling and enforcing instruction priorities. At inference time, we introduce solver-guided reasoning that formulates instruction resolution as a constraint satisfaction problem, enabling the model to derive a maximally consistent set of applicable instructions under hierarchical constraints. At training time, NSHA distills solver-based decisions into model parameters using automatically constructed supervision. We evaluate our approach on rule following, task execution, tool use, and safety, covering both single-turn and multi-turn interactions, and show that NSHA significantly improves performance under such conflicts while maintaining competitive utility in reference settings.

Diffusion model deployment has been suffering from high energy consumption and inference latency despite its superior performance in visual generation tasks. Dynamic voltage and frequency scaling (DVFS) offers a promising solution to exploit the potential of the underlying accelerators. However, existing approaches often lead to either limited efficiency gains or degraded output quality because they overlook the inherent fault tolerance of the diffusion model. Therefore, in this paper, we propose DRIFT, a novel algorithmarchitecture co-optimization framework that harnesses the fault tolerance for efficient and reliable diffusion model inference. We first perform a comprehensive resilience analysis on representative diffusion models. Building on these observations, we introduce a fine-grained, resilience-aware DVFS strategy that selectively protects error-sensitive network blocks and timesteps, and a rollback algorithm-based fault tolerance (ABFT) mechanism that adaptively corrects only critical errors by reverting to previous timesteps. We further optimize offloading intervals and reorganize data layouts to reduce memory overhead. Experiments across diverse models and datasets show that DRIFT can achieve on average 36% energy savings through voltage underscaling or 1.7x speedup via overclocking while maintaining generation quality.

Humans effortlessly navigate the physical world by predicting how objects behave under gravity and contact forces, yet how such judgments support sequential physical planning under resource constraints remains poorly understood. Research on intuitive physics debates whether prediction relies on the Intuitive Physics Engine (IPE) or fast, cue-based heuristics; separately, decision-making research debates deliberative lookahead versus myopic strategies. These debates have proceeded in isolation, leaving the cognitive architecture of sequential physical planning underspecified. How physical prediction mechanisms and planning strategies jointly adapt under limited cognitive resources remains an open question. Here we show that humans exhibit a dual transition under resource pressure, simultaneously shifting both physical prediction mechanism and planning strategy to match cognitive budget. Using Overhang Tower, a construction task requiring participants to maximize horizontal overhang while maintaining stability, we find that IPE-based simulation dominates early stages while CNN-based visual heuristics prevail as complexity grows; concurrently, time pressure truncates deliberative lookahead, shifting planning toward shallower horizons: a dual transition unpredicted by prior single-mechanism accounts. These findings reveal a hierarchical, resource-rational architecture that flexibly trades computational cost against predictive fidelity. Our results unify two long-standing debates (simulation vs. heuristics and myopic vs. deliberative planning) as a dynamic repertoire reconfigured by cognitive budget.

Court Judgment Prediction and Explanation (CJPE) aims to predict a judicial decision and provide a legally grounded explanation for a given case based on the facts, legal issues, arguments, cited statutes, and relevant precedents. For such systems to be practically useful in judicial or legal research settings, they must not only achieve high predictive performance but also generate transparent and structured legal reasoning that aligns with established judicial practices. In this work, we present NyayaMind, an open-source framework designed to enable transparent and scalable legal reasoning for the Indian judiciary. The proposed framework integrates retrieval, reasoning, and verification mechanisms to emulate the structured decision-making process typically followed in courts. Specifically, NyayaMind consists of two main components: a Retrieval Module and a Prediction Module. The Retrieval Module employs a RAG pipeline to identify legally relevant statutes and precedent cases from large-scale legal corpora, while the Prediction Module utilizes reasoning-oriented LLMs fine-tuned for the Indian legal domain to generate structured outputs including issues, arguments, rationale, and the final decision. Our extensive results and expert evaluation demonstrate that NyayaMind significantly improves the quality of explanation and evidence alignment compared to existing CJPE approaches, providing a promising step toward trustworthy AI-assisted legal decision support systems.

Data augmentation is a crucial technique for improving model generalization and robustness, particularly in deep learning models where training data is limited. Although many augmentation methods have been developed for time series classification, most are not directly applicable to time series forecasting due to the need to preserve temporal coherence. In this work, we propose Temporal Patch Shuffle (TPS), a simple and model-agnostic data augmentation method for forecasting that extracts overlapping temporal patches, selectively shuffles a subset of patches using variance-based ordering as a conservative heuristic, and reconstructs the sequence by averaging overlapping regions. This design increases sample diversity while preserving forecast-consistent local temporal structure. We extensively evaluate TPS across nine long-term forecasting datasets using five recent model families (TSMixer, DLinear, PatchTST, TiDE, and LightTS), and across four short-term forecasting datasets using PatchTST, observing consistent performance improvements. Comprehensive ablation studies further demonstrate the effectiveness, robustness, and design rationale of the proposed method.

Linguistic steganography based on language models typically assumes that steganographic texts are transmitted without alteration, making them fragile to even minor modifications. While previous work mitigates this fragility by limiting the context window, it significantly compromises text quality. In this paper, we propose the anchored sliding window (ASW) framework to improve imperceptibility and robustness. In addition to the latest tokens, the prompt and a bridge context are anchored within the context window, encouraging the model to compensate for the excluded tokens. We formulate the optimization of the bridge context as a variant of prompt distillation, which we further extend using self-distillation strategies. Experiments show that our ASW significantly and consistently outperforms the baseline method in text quality, imperceptibility, and robustness across diverse settings. The code is available at github.com/ryehr/ASW_steganography.

In partial multi-label learning (PML), each instance is associated with a set of candidate labels containing both ground-truth and noisy labels. The presence of noisy labels disrupts the correspondence between features and labels, degrading classification performance. To address this challenge, we propose a novel PML method based on feature-label modal alignment (PML-MA), which treats features and labels as two complementary modalities and restores their consistency through systematic alignment. Specifically, PML-MA first employs low-rank orthogonal decomposition to generate pseudo-labels that approximate the true label distribution by filtering noisy labels. It then aligns features and pseudo-labels through both global projection into a common subspace and local preservation of neighborhood structures. Finally, a multi-peak class prototype learning mechanism leverages the multi-label nature where instances simultaneously belong to multiple categories, using pseudo-labels as soft membership weights to enhance discriminability. By integrating modal alignment with prototype-guided refinement, PML-MA ensures pseudo-labels better reflect the true distribution while maintaining robustness against label noise. Extensive experiments on both real-world and synthetic datasets demonstrate that PML-MA significantly outperforms state-of-the-art methods, achieving superior classification accuracy and noise robustness.

Human action recognition is pivotal in computer vision, with applications ranging from surveillance to human-robot interaction. Despite the effectiveness of supervised skeleton-based methods, their reliance on exhaustive annotation limits generalization to novel actions. Zero-Shot Skeleton Action Recognition (ZSAR) emerges as a promising paradigm, yet it faces challenges due to the spectral bias of diffusion models, which oversmooth high-frequency dynamics. Here, we propose Frequency-Aware Diffusion for Skeleton-Text Matching (FDSM), integrating a Semantic-Guided Spectral Residual Module, a Timestep-Adaptive Spectral Loss, and Curriculum-based Semantic Abstraction to address these challenges. Our approach effectively recovers fine-grained motion details, achieving state-of-the-art performance on NTU RGB+D, PKU-MMD, and Kinetics-skeleton datasets. Code has been made available at https://github.com/yuzhi535/FDSM. Project homepage: https://yuzhi535.github.io/FDSM.github.io/

Vision-Language-Action (VLA) models have recently achieved notable progress in end-to-end autonomous driving by integrating perception, reasoning, and control within a unified multimodal framework. However, they often lack explicit modeling of temporal dynamics and global world consistency, which limits their foresight and safety. In contrast, world models can simulate plausible future scenes but generally struggle to reason about or evaluate the imagined future they generate. In this work, we present VLA-World, a simple yet effective VLA world model that unifies predictive imagination with reflective reasoning to improve driving foresight. VLA-World first uses an action-derived feasible trajectory to guide the generation of the next-frame image, capturing rich spatial and temporal cues that describe how the surrounding environment evolves. The model then reasons over this self-generated future imagined frame to refine the predicted trajectory, achieving higher performance and better interpretability. To support this pipeline, we curate nuScenes-GR-20K, a generative reasoning dataset derived from nuScenes, and employ a three-stage training strategy that includes pretraining, supervised fine-tuning, and reinforcement learning. Extensive experiments demonstrate that VLA-World consistently surpasses state-of-the-art VLA and world-model baselines on both planning and future-generation benchmarks. Project page: https://vlaworld.github.io

Long-horizon spatiotemporal prediction remains a challenging problem due to cumulative errors, noise amplification, and the lack of physical consistency in existing models. While diffusion models provide a probabilistic framework for modeling uncertainty, conventional approaches often rely on mean squared error objectives and fail to capture the underlying dynamics governed by physical laws. In this work, we propose PDYffusion, a dynamics-informed diffusion framework that integrates PDE-based regularization and uncertainty-aware forecasting for stable long-term prediction. The proposed method consists of two key components: a PDE-regularized interpolator and a UKF-based forecaster. The interpolator incorporates a differential operator to enforce physically consistent intermediate states, while the forecaster leverages the Unscented Kalman Filter to explicitly model uncertainty and mitigate error accumulation during iterative prediction. We provide theoretical analyses showing that the proposed interpolator satisfies PDE-constrained smoothness properties, and that the forecaster converges under the proposed loss formulation. Extensive experiments on multiple dynamical datasets demonstrate that PDYffusion achieves superior performance in terms of CRPS and MSE, while maintaining stable uncertainty behavior measured by SSR. We further analyze the inherent trade-off between prediction accuracy and uncertainty, showing that our method provides a balanced and robust solution for long-horizon forecasting.

While the large energy consumption of Large Language Models (LLMs) is recognized by the community, system operators lack guidance for energy-efficient LLM inference deployments that leverage energy trade-offs of heterogeneous hardware due to a lack of energy-aware benchmarks and data. In this work we address this gap with Watt Counts: the largest open-access dataset of energy consumption of LLMs, with over 5,000 experiments for 50 LLMs across 10 NVIDIA Graphics Processing Units (GPUs) in batch and server scenarios along with a reproducible, open-source benchmark that enables community submissions to expand this dataset. Leveraging this dataset, we conduct a system-level study of LLM inference across heterogeneous GPU architectures and show that GPU selection is crucial for energy efficiency outcomes and that optimal hardware choices vary significantly across models and deployment scenarios, demonstrating the critical importance of hardware-aware deployment in heterogeneous LLM systems. Guided by our data and insights, we show that practitioners can reduce energy consumption by up to 70% in server scenarios with negligible impact on user experience, and by up to 20% in batch scenarios.

AI-based weather forecasting now rivals traditional physics-based ensembles, but state-of-the-art (SOTA) models rely on specialized architectures and massive computational budgets, creating a high barrier to entry. We demonstrate that such complexity is unnecessary for frontier performance. We introduce U-Cast, a probabilistic forecaster built on a standard U-Net backbone trained with a simple recipe: deterministic pre-training on Mean Absolute Error followed by short probabilistic fine-tuning on the Continuous Ranked Probability Score (CRPS) using Monte Carlo Dropout for stochasticity. As a result, our model matches or exceeds the probabilistic skill of GenCast and IFS ENS at 1.5$^\circ\$ resolution while reducing training compute by over 10$\times$ compared to leading CRPS-based models and inference latency by over 10$\times$ compared to diffusion-based models. U-Cast trains in under 12 H200 GPU-days and generates a 60-step ensemble forecast in 11 seconds. These results suggest that scalable, general-purpose architectures paired with efficient training curricula can match complex domain-specific designs at a fraction of the cost, opening the training of frontier probabilistic weather models to the broader community. Our code is available at: https://github.com/Rose-STL-Lab/u-cast.

Robust geo-localization in changing environmental conditions is critical for long-term aerial autonomy. While visual place recognition (VPR) models perform well when airborne views match the training domain, adapting them to shifting distributions during sequential missions triggers catastrophic forgetting. Existing continual learning (CL) methods often fail here because geographic features exhibit severe intra-class variations. In this work, we formulate aerial VPR as a mission-based domain-incremental learning (DIL) problem and propose a novel heterogeneous memory framework. To respect strict onboard storage constraints, our "Learn-and-Dispose" pipeline decouples geographic knowledge into static satellite anchors (preserving global geometric priors) and a dynamic experience replay buffer (retaining domain-specific features). We introduce a spatially-constrained allocation strategy that optimizes buffer selection based on sample difficulty or feature space diversity. To facilitate systematic assessment, we provide three evaluation criteria and a comprehensive benchmark derived from 21 diverse mission sequences. Extensive experiments demonstrate that our architecture significantly boosts spatial generalization; our diversity-driven buffer selection outperforms the random baseline by 7.8% in knowledge retention. Unlike class-mean preservation methods that fail in unstructured environments, maximizing structural diversity achieves a superior plasticity-stability balance and ensures order-agnostic robustness across randomized sequences. These results prove that maintaining structural feature coverage is more critical than sample difficulty for resolving catastrophic forgetting in lifelong aerial autonomy.

Current video benchmarks for multimodal large language models (MLLMs) focus on event recognition, temporal ordering, and long-context recall, but overlook a harder capability required for expert procedural judgment: tracking how ongoing interactions update the procedural state and thereby determine the correctness of later actions. We introduce SiMing-Bench, the first benchmark for evaluating this capability from full-length clinical skill videos. It targets rubric-grounded process-level judgment of whether interaction-driven state updates preserve procedural correctness across an entire workflow. SiMing-Bench is instantiated with SiMing-Score, a physician-annotated dataset of real clinical skill examination videos spanning cardiopulmonary resuscitation, automated external defibrillator operation, and bag-mask ventilation, each paired with a standardized step-wise rubric and dual-expert labels. Across diverse open- and closed-source MLLMs, we observe consistently weak agreement with physician judgments. Moreover, weak performance on rubric-defined intermediate steps persists even when overall procedure-level correlation appears acceptable, suggesting that coarse global assessment substantially overestimates current models' procedural judgment ability. Additional analyses with binary step judgment and step-aligned clips indicate that the bottleneck is not merely fine-grained scoring or temporal localization, but modeling how continuous interactions update procedural state over time.

Model-based reinforcement learning (MBRL) with autoregressive world models suffers from compounding errors, whereas diffusion world models mitigate this by generating trajectory segments jointly. However, existing diffusion guides are either policy-only, discarding value information, or reward-based, which becomes myopic when the diffusion horizon is short. We introduce Advantage-Guided Diffusion for MBRL (AGD-MBRL), which steers the reverse diffusion process using the agent's advantage estimates so that sampling concentrates on trajectories expected to yield higher long-term return beyond the generated window. We develop two guides: (i) Sigmoid Advantage Guidance (SAG) and (ii) Exponential Advantage Guidance (EAG). We prove that a diffusion model guided through SAG or EAG allows us to perform reweighted sampling of trajectories with weights increasing in state-action advantage-implying policy improvement under standard assumptions. Additionally, we show that the trajectories generated from AGD-MBRL follow an improved policy (that is, with higher value) compared to an unguided diffusion model. AGD integrates seamlessly with PolyGRAD-style architectures by guiding the state components while leaving action generation policy-conditioned, and requires no change to the diffusion training objective. On MuJoCo control tasks (HalfCheetah, Hopper, Walker2D and Reacher), AGD-MBRL improves sample efficiency and final return over PolyGRAD, an online Diffuser-style reward guide, and model-free baselines (PPO/TRPO), in some cases by a margin of 2x. These results show that advantage-aware guidance is a simple, effective remedy for short-horizon myopia in diffusion-model MBRL.

The rapid evolution of Large Language Models (LLMs) has significantly impacted the field of natural language processing, but their growing complexity raises concerns about resource usage and transparency. Addressing these challenges, we participated in the NeurIPS LLM Efficiency Challenge, aiming to fine-tune a foundation model within stringent constraints. Our focus was the LLaMa2 70 billion model, optimized on a single A100 40GB GPU within a 24-hour limit. Our methodology hinged on a custom dataset, carefully assembled from diverse open-source resources and benchmark tests, aligned with the challenge's open-source ethos. Our approach leveraged Quantized-Low Rank Adaptation (QLoRA) Fine tuning, integrated with advanced attention mechanisms like Flash Attention 2. We experimented with various configurations of the LoRA technique, optimizing the balance between computational efficiency and model accuracy. Our fine-tuning strategy was underpinned by the creation and iterative testing of multiple dataset compositions, leading to the selection of a version that demonstrated robust performance across diverse tasks and benchmarks. The culmination of our efforts was an efficiently fine-tuned LLaMa2 70B model that operated within the constraints of a single GPU, showcasing not only a significant reduction in resource utilization but also high accuracy across a range of QA benchmarks. Our study serves as a testament to the feasibility of optimizing large-scale models in resource-constrained environments, emphasizing the potential of LLMs in real-world applications.

Large language models have been widely explored as decision-support tools in high-stakes domains due to their contextual understanding and reasoning capabilities. However, existing decision-making benchmarks rely on two simplifying assumptions: actions are selected from a finite set of pre-defined candidates, and explicit conditions restricting action feasibility are not incorporated into the decision-making process. These assumptions fail to capture the compositional structure of real-world actions and the explicit conditions that constrain their validity. To address these limitations, we introduce CONDESION-BENCH, a benchmark designed to evaluate conditional decision-making in compositional action space. In CONDESION-BENCH, actions are defined as allocations to decision variables and are restricted by explicit conditions at the variable, contextual, and allocation levels. By employing oracle-based evaluation of both decision quality and condition adherence, we provide a more rigorous assessment of LLMs as decision-support tools.

Unmanned aerial vehicles serving as aerial base stations can rapidly restore connectivity after disasters, yet abrupt changes in user mobility and traffic demands shift the quality of service trade-offs and induce strong non-stationarity. Deep reinforcement learning policies suffer from plasticity loss under such shifts, as representation collapse and neuron dormancy impair adaptation. We propose plasticity enhanced multi-agent mixture of experts (PE-MAMoE), a centralized training with decentralized execution framework built on multi-agent proximal policy optimization. PE-MAMoE equips each UAV with a sparsely gated mixture of experts actor whose router selects a single specialist per step. A non-parametric Phase Controller injects brief, expert-only stochastic perturbations after phase switches, resets the action log-standard-deviation, anneals entropy and learning rate, and schedules the router temperature, all to re-plasticize the policy without destabilizing safe behaviors. We derive a dynamic regret bound showing the tracking error scales with both environment variation and cumulative noise energy. In a phase-driven simulator with mobile users and 3GPP-style channels, PE-MAMoE improves normalized interquartile mean return by 26.3\% over the best baseline, increases served-user capacity by 12.8\%, and reduces collisions by approximately 75\%. Diagnostics confirm persistently higher expert feature rank and periodic dormant-neuron recovery at regime switches.

Social agents both internalize collective norms and reshape them through creative action, yet computational models have not captured this bidirectional process within a unified framework. We propose a multi-agent simulation model grounded in active inference that formalizes the dialectical constitution of social reality on a structured social network. Each agent maintains an internal generative model, communicates with neighbors to form social priors, creates novel observations, and selectively incorporates others' creations into memory. Simulation experiments demonstrate three main findings. First, informationally cohesive social groups emerge endogenously, with representational alignment mirroring the cluster topology of the underlying network. Second, a circular mutual constitution arises between social representations and the observation distribution, maintained through agents' creative acts that project representational structure onto the external world. Third, the propagation of creations exhibits selective, heterogeneous patterns distinct from the stable diffusion of social representations, indicating that agents construct cultural niches through local interaction dynamics. These results suggest that the interplay between social conformity and creative deviation can give rise to the endogenous formation and differentiation of shared social reality.

Vision-language models (VLMs) have shown a promising ability in image geolocation, but they still lack structured geographic reasoning and the capacity for autonomous self-evolution. Existing methods predominantly rely on implicit parametric memory, which often exploits outdated knowledge and generates hallucinated reasoning. Furthermore, current inference is a "one-off" process, lacking the feedback loops necessary for self-evolution based on reasoning outcomes. To address these issues, we propose GeoSkill, a training-free framework based on an evolving Skill-Graph. We first initialize the graph by refining human expert trajectories into atomic, natural-language skills. For execution, GeoSkill employs an inference model to perform direct reasoning guided by the current Skill-Graph. For continuous growth, an Autonomous Evolution mechanism leverages a larger model to conduct multiple reasoning rollouts on image-coordinate pairs sourced from web-scale data and verified real-world reasoning. By analyzing both successful and failed trajectories from these rollouts, the mechanism iteratively synthesizes and prunes skills, effectively expanding the Skill-Graph and correcting geographic biases without any parameter updates. Experiments demonstrate that GeoSkill achieves promising performance in both geolocation accuracy and reasoning faithfulness on GeoRC, while maintaining superior generalization across diverse external datasets. Furthermore, our autonomous evolution fosters the emergence of novel, verifiable skills, significantly enhancing the system's cognition of real-world geographic knowledge beyond isolated case studies.

Multi-modal large language models (MLLMs) have emerged as powerful tools for analyzing Internet-scale image data, offering significant benefits but also raising critical safety and societal concerns. In particular, open-weight MLLMs may be misused to extract sensitive information from personal images at scale, such as identities, locations, or other private details. In this work, we propose ImageProtector, a user-side method that proactively protects images before sharing by embedding a carefully crafted, nearly imperceptible perturbation that acts as a visual prompt injection attack on MLLMs. As a result, when an adversary analyzes a protected image with an MLLM, the MLLM is consistently induced to generate a refusal response such as "I'm sorry, I can't help with that request." We empirically demonstrate the effectiveness of ImageProtector across six MLLMs and four datasets. Additionally, we evaluate three potential countermeasures, Gaussian noise, DiffPure, and adversarial training, and show that while they partially mitigate the impact of ImageProtector, they simultaneously degrade model accuracy and/or efficiency. Our study focuses on the practically important setting of open-weight MLLMs and large-scale automated image analysis, and highlights both the promise and the limitations of perturbation-based privacy protection.

Auditory large language models (ALLMs) have demonstrated strong general capabilities in audio understanding and reasoning tasks. However, their reliability is still undermined by hallucination issues. Existing hallucination evaluation methods are formulated as binary classification tasks, which are insufficient to characterize the more complex hallucination patterns that arise in generative tasks. Moreover, current hallucination mitigation strategies rely on fine-tuning, resulting in high computational costs. To address the above limitations, we propose a plug-and-play Noise-Aware In-Context Learning (NAICL) method. Specifically, we construct a noise prior library, retrieve noise examples relevant to the input audio, and incorporate them as contextual priors, thereby guiding the model to reduce speculative associations when acoustic evidence is insufficient and to adopt a more conservative generation strategy. In addition, we establish a hallucination benchmark for audio caption tasks including the construction of the Clotho-1K multi-event benchmark dataset, the definition of four types of auditory hallucinations, and the introduction of metrics such as hallucination type distribution to support fine-grained analysis. Experimental results show that all evaluated ALLMs exhibit same hallucination behaviors. Moreover, the proposed NAICL method reduces the overall hallucination rate from 26.53% to 16.98%.

Two-hop QA retrieval splits queries into two regimes determined by whether the hop-2 entity is explicitly named in the question (Q-dominant) or only in the bridge passage (B-dominant). We formalize this split with three theorems: (T1) per-query AUC is a monotone function of the cosine separation margin, with R^2 >= 0.90 for six of eight type-encoder pairs; (T2) regime is characterized by two surface-text predicates, with P1 decisive for routing and P2 qualifying the B-dominant case, holding across three encoders and three datasets; and (T3) bridge advantage requires the relation-bearing sentence, not entity name alone, with removal causing an 8.6-14.1 pp performance drop (p < 0.001). Building on this theory, we propose RegimeRouter, a lightweight binary router that selects between question-only and question-plus-relation-sentence retrieval using five text features derived directly from the predicate definitions. Trained on 2WikiMultiHopQA (n = 881, 5-fold cross-fitted) and applied zero-shot to MuSiQue and HotpotQA, RegimeRouter achieves +5.6 pp (p < 0.001), +5.3 pp (p = 0.002), and +1.1 pp (non-significant, no-regret) R@5 improvement, respectively, with artifact-driven.

This study addresses the challenge of creating datasets for cybercrime analysis while complying with the requirements of regulations such as the General Data Protection Regulation (GDPR) and Organic Law 10/1995 of the Penal Code. To this end, a system is proposed for collecting information from the Telegram platform, including text, audio, and images; the implementation of speech-to-text transcription models incorporating signal enhancement techniques; and the evaluation of different Named Entity Recognition (NER) solutions, including Microsoft Presidio and AI models designed using a transformer-based architecture. Experimental results indicate that Parakeet achieves the best performance in audio transcription, while the proposed NER solutions achieve the highest f1-score values in detecting sensitive information. In addition, anonymization metrics are presented that allow evaluation of the preservation of structural coherence in the data, while simultaneously guaranteeing the protection of personal information and supporting cybersecurity research within the current legal framework.

The widespread use of English as a Second or Foreign Language (ESFL) has sparked a paradigm shift: ESFL is not seen merely as a deviation from standard English but as a distinct linguistic system in its own right. This shift highlights the need for dedicated, knowledge-intensive representations of ESFL. In response, this paper surveys existing ESFL resources, identifies their limitations, and proposes a novel solution. Grounded in constructivist theories, the paper treats constructions as the fundamental units of analysis, allowing it to model the syntax--semantics interface of both ESFL and standard English. This design captures a wide range of ESFL phenomena by referring to syntactico-semantic mappings of English while preserving ESFL's unique characteristics, resulting a gold-standard syntactico-semantic resource comprising 1643 annotated ESFL sentences. To demonstrate the sembank's practical utility, we conduct a pilot study testing the Linguistic Niche Hypothesis, highlighting its potential as a valuable tool in Second Language Acquisition research.

Enumerating Minimal Unsatisfiable Subsets (MUSes) is a fundamental task in constraint satisfaction problems (CSPs). Its major challenge is the exponential growth of the search space, which becomes particularly severe when satisfiability checks are expensive. Recent machine learning approaches reduce this cost for Boolean satisfiability problems but rely on explicit variable-constraint relationships, limiting their application domains. This paper proposes a domain-agnostic method to accelerate MUS enumeration using Hypergraph Neural Networks (HGNNs). The proposed method incrementally builds a hypergraph with constraints as vertices and MUSes enumerated until the current step as hyperedges, and employs an HGNN-based agent trained via reinforcement learning to minimize the number of satisfiability checks required to obtain an MUS. Experimental results demonstrate the effectiveness of our approach in accelerating MUS enumeration, showing that our method can enumerate more MUSes within the same satisfiability check budget compared to conventional methods.

Table serialization remains a critical bottleneck for Large Language Models (LLMs) in complex table question answering, hindered by challenges such as structural neglect, representation gaps, and reasoning opacity. Existing serialization methods fail to capture explicit hierarchies and lack schema flexibility, while current tree-based approaches suffer from limited semantic adaptability. To address these limitations, we propose ASTRA (Adaptive Semantic Tree Reasoning Architecture) including two main modules, AdaSTR and DuTR. First, we introduce AdaSTR, which leverages the global semantic awareness of LLMs to reconstruct tables into Logical Semantic Trees. This serialization explicitly models hierarchical dependencies and employs an adaptive mechanism to optimize construction strategies based on table scale. Second, building on this structure, we present DuTR, a dual-mode reasoning framework that integrates tree-search-based textual navigation for linguistic alignment and symbolic code execution for precise verification. Experiments on complex table benchmarks demonstrate that our method achieves state-of-the-art (SOTA) performance.

Small object-centric spatial understanding in indoor videos remains a significant challenge for multimodal large language models (MLLMs), despite its practical value for object search and assistive applications. Although existing benchmarks have advanced video spatial intelligence, embodied reasoning, and diagnostic perception, no existing benchmark directly evaluates whether a model can localize a target object in video and express its position with sufficient precision for downstream use. In this work, we introduce PinpointQA, the first dataset and benchmark for small object-centric spatial understanding in indoor videos. Built from ScanNet++ and ScanNet200, PinpointQA comprises 1,024 scenes and 10,094 QA pairs organized into four progressively challenging tasks: Target Presence Verification (TPV), Nearest Reference Identification (NRI), Fine-Grained Spatial Description (FSD), and Structured Spatial Prediction (SSP). The dataset is built from intermediate spatial representations, with QA pairs generated automatically and further refined through quality control. Experiments on representative MLLMs reveal a consistent capability gap along the progressive chain, with SSP remaining particularly difficult. Supervised fine-tuning on PinpointQA yields substantial gains, especially on the harder tasks, demonstrating that PinpointQA serves as both a diagnostic benchmark and an effective training dataset. The dataset and project page are available at https://rainchowz.github.io/PinpointQA.

Current LLM-based agents demonstrate strong performance in episodic task execution but remain constrained by static toolsets and episodic amnesia, failing to accumulate experience or optimize strategies across task boundaries. While the Self-Evolving Agent (SEA) paradigm has been previously proposed, this paper contributes a new formal definition of SEA grounded in digital embodiment and continuous cross-task evolution, and introduces SEA-Eval, the first benchmark designed to evaluate SEA characteristics across two dimensions, intra-task execution reliability and long-term evolutionary performance. By organizing tasks into sequential streams and analyzing Success Rate and Token Consumption over time, SEA-Eval quantifies evolutionary gain and structural stability in ways that existing episodic benchmarks cannot. Empirical evaluations reveal a significant evolutionary bottleneck in current state-of-the-art frameworks, where identical success rates mask up to 31.2 times differences in token consumption and divergent evolutionary trajectories under sequential analysis. SEA-Eval provides a rigorous scientific foundation for advancing agents from mere task executors toward genuinely self-evolving digital entities.

As Large Language Models (LLMs) advance toward embodied AI agents operating in physical environments, a fundamental question emerges: can models trained on text corpora reliably reason about complex physics while adhering to safety constraints? We address this through PilotBench, a benchmark evaluating LLMs on safety-critical flight trajectory and attitude prediction. Built from 708 real-world general aviation trajectories spanning nine operationally distinct flight phases with synchronized 34-channel telemetry, PilotBench systematically probes the intersection of semantic understanding and physics-governed prediction through comparative analysis of LLMs and traditional forecasters. We introduce Pilot-Score, a composite metric balancing 60% regression accuracy with 40% instruction adherence and safety compliance. Comparative evaluation across 41 models uncovers a Precision-Controllability Dichotomy: traditional forecasters achieve superior MAE of 7.01 but lack semantic reasoning capabilities, while LLMs gain controllability with 86--89% instruction-following at the cost of 11--14 MAE precision. Phase-stratified analysis further exposes a Dynamic Complexity Gap-LLM performance degrades sharply in high-workload phases such as Climb and Approach, suggesting brittle implicit physics models. These empirical discoveries motivate hybrid architectures combining LLMs' symbolic reasoning with specialized forecasters' numerical precision. PilotBench provides a rigorous foundation for advancing embodied AI in safety-constrained domains.

Persona prompting has been widely adopted to steer large language models (LLMs) behavior and improve their instruction performance by assigning specific characters. However, identifying an optimal persona is time-consuming, and its impact on output quality remains poorly understood. Prior work has mainly addressed this issue at the prompt level via inference-time strategies, incurring additional computation. In this work, we avoid inference-time prompt search by tackling persona sensitivity during training, aiming to train models that adapt their behavior to diverse personas while preserving task performance. In particular, we find that reinforcement learning with verifiable rewards (RLVR) systematically reduces sensitivity to persona prompts, but also reveals an inherent trade-off of outcome-based optimization: while RLVR improves robustness on tasks with verifiable goals, it can also degrade persona expressivity when needed, e.g., in-character role-playing. To address this limitation, we propose PerMix-RLVR, a persona-mixed RLVR strategy that mitigates the persona robustness-fidelity trade-off, preserving strong robustness to harmful persona variation while enabling faithful persona adoption when required. Concretely, PerMix-RLVR improves persona stability score (PSS) over RLVR by +21.2% on MATH500, while also enhancing persona fidelity by +11.4% on PersonaGym.

Graph neural networks (GNNs) have been widely adopted in engineering applications such as social network analysis, chemical research and computer vision. However, their efficacy is severely compromised by the inherent homophily assumption, which fails to hold for heterophilic graphs where dissimilar nodes are frequently connected. To address this fundamental limitation in graph learning, we first draw inspiration from the recently discovered monophily property of real-world graphs, and propose Neighbourhood Transformers (NT), a novel paradigm that applies self-attention within every local neighbourhood instead of aggregating messages to the central node as in conventional message-passing GNNs. This design makes NT inherently monophily-aware and theoretically guarantees its expressiveness is no weaker than traditional message-passing frameworks. For practical engineering deployment, we further develop a neighbourhood partitioning strategy equipped with switchable attentions, which reduces the space consumption of NT by over 95% and time consumption by up to 92.67%, significantly expanding its applicability to larger graphs. Extensive experiments on 10 real-world datasets (5 heterophilic and 5 homophilic graphs) show that NT outperforms all current state-of-the-art methods on node classification tasks, demonstrating its superior performance and cross-domain adaptability. The full implementation code of this work is publicly available at https://github.com/cf020031308/MoNT to facilitate reproducibility and industrial adoption.

Active learning (AL) is a training paradigm for selecting unlabeled samples for annotation to improve model performance on a test set, which is useful when only a limited number of samples can be annotated. These algorithms often work by optimizing for the informativeness and diversity of the training data to be annotated. Recent work found that AL strategies fail to outperform random sampling on various language generation tasks when using 100-500 samples. To understand AL's poor performance when only using few samples, we investigate whether the core assumptions underlying AL strategies hold. We find that neither the informativeness nor diversity of the training data, which AL strategies optimize for, are correlated with test set performance. Instead, factors like the ordering of the training samples and interactions with pre-training data have a larger impact on performance. This suggests that future AL methods must take these factors into account in order to work with very few samples.

We report that model quantisation restructures domain-level metacognitive efficiency in LLMs rather than degrading it uniformly. Evaluating Llama-3-8B-Instruct on the same 3,000 questions at Q5_K_M and f16 precision, we find that M-ratio profiles across four knowledge domains are uncorrelated between formats (Spearman rho = 0.00). Arts & Literature moves from worst-monitored (M-ratio = 0.606 at Q5_K_M) to best-monitored (1.542 at f16). Geography moves from well-monitored (1.210) to under-monitored (0.798). However, Type-2 AUROC profiles are perfectly stable across formats (rho = 1.00), localising the restructuring to the M-ratio normalisation rather than the underlying discrimination signal. This finding emerged from a pre-registered attempt to improve metacognition through domain-conditional training. We prescribed confidence-amplification SFT for the diagnosed weak domain, with matched-budget agnostic and wrong-prescription controls. All four confirmatory hypotheses were null (10,000 bootstrap resamples, seed = 42). The training successfully reshaped confidence distributions, doubling the NLP gap in Science from 0.076 to 0.152, but did not improve meta-d' because the diagnostic profile did not transfer across formats. Any system relying on domain-level M-ratio profiles has an unexamined dependency on inference format. Systems using AUROC_2 are safer. We release all code, pre-registrations, and trial-level data.

Uncertainty quantification is a set of techniques that measure confidence in language models. They can be used, for example, to detect hallucinations or alert users to review uncertain predictions. To be useful, these confidence scores must be correlated with the quality of the output. However, recent work found that fine-tuning can affect the correlation between confidence scores and quality. Hence, we investigate the underlying behavior of confidence scores to understand its sensitivity to supervised fine-tuning (SFT). We find that post-SFT, the correlation of various confidence scores degrades, which can stem from changes in confidence scores due to factors other than the output quality, such as the output's similarity to the training distribution. We demonstrate via a case study how failing to address this miscorrelation reduces the usefulness of the confidence scores on a downstream task. Our findings show how confidence metrics cannot be used off-the-shelf without testing, and motivate the need for developing metrics which are more robust to fine-tuning.

Uncertainties in renewable generation and demand dynamics challenge day-ahead scheduling. To enhance renewable penetration and maintain intra-day balance, we develop a multi-agent reinforcement learning framework for self-interested microgrids participating in peer-to-peer (P2P) electricity trading. Each microgrid independently bids both price and quantity while optimizing its own profit via storage arbitrage under time-varying main-grid prices. A market-clearing mechanism coordinating trades and promoting incentive compatibility is proposed. Simulation results show that the learned bidding policy improves renewable utilization and reduces reliance on high-carbon electricity, while increasing community-level economic welfare, delivering a win-win situation in emission reduction and local prosperity.

Edge devices increasingly run multimodal sensing pipelines that must remain accurate despite fluctuating power budgets and unpredictable sensor dropout. Existing pruning methods fail under these conditions: they generally require fine-tuning after compression, consuming over $10\times$ the deployment energy, and they assign static importance scores that are blind to which sensors are present. We present the SentryFuse framework, which addresses both challenges jointly through two key components. First, SentryGate learns modality-conditioned importance scores during training via first-order saliency supervision and then prunes attention heads and feed-forward channels at deployment without fine-tuning. Second, SentryAttend replaces dense self-attention, a key bottleneck in contemporary multimodal architectures, with sparse grouped-query attention, yielding a net 15% reduction in GFLOPs across three different multimodal architectures. Across three applications and multimodal backbones, SentryGate achieves a 12.7% average accuracy improvement over the strongest pruning baseline, and upto to 18% under modality dropout conditions. Together, SentryFuse reduces memory by 28.2% and lowers latency by up to $1.63\times$ without further fine-tuning, establishing modality-aware zero-shot compression as a practical path to multimodal intelligence on heterogeneous edge hardware.

We study predictive multilingual evaluation: estimating how well a model will perform on a task in a target language when direct benchmark results are missing. This problem is common in multilingual deployment, where evaluation coverage is sparse and published evidence is uneven across languages, tasks, and model families. We introduce a controlled benchmark of 1,500 questions spanning six tasks and five evidence scenarios. The benchmark separates accessible evidence from ground truth, enabling evaluation of systems that must infer missing results from incomplete literature evidence. We also present Litmus (Re)Agent, a DAG-orchestrated agentic system that decomposes queries into hypotheses, retrieves evidence, and synthesises predictions through feature-aware aggregation. Across six systems, Litmus (Re)Agent achieves the best overall performance, with the largest gains in transfer-heavy scenarios where direct evidence is weak or absent. These results show that structured agentic reasoning is a promising approach to multilingual performance estimation under incomplete evidence.

This paper introduces a projected functional gradient descent algorithm (P-FGD) for training nonparametric additive quantile regression models in online settings. This algorithm extends the functional stochastic gradient descent framework to the pinball loss. An advantage of P-FGD is that it does not need to store historical data while maintaining $O(J_t\ln J_t)$ computational complexity per step where $J_t$ denotes the number of basis functions. Besides, we only need $O(J_t)$ computational time for quantile function prediction at time $t$. These properties show that P-FGD is much better than the commonly used RKHS in online learning. By leveraging a novel Hilbert space projection identity, we also prove that the proposed online quantile function estimator (P-FGD) achieves the minimax optimal consistency rate $O(t^{-\frac{2s}{2s+1}})$ where $t$ is the current time and $s$ denotes the smoothness degree of the quantile function. Extensions to mini-batch learning are also established.

Diffusion Large Language Models (dLLMs) have recently become a promising alternative to autoregressive large language models (ARMs). Semi-autoregressive (Semi-AR) decoding is widely employed in base dLLMs and advanced decoding strategies due to its superior performance. However, our observations reveal that Semi-AR decoding suffers from inherent block constraints, which cause the decoding of many cross-block stable tokens to be unnecessarily delayed. To address this challenge, we systematically investigate the identification of stable tokens and present three key findings: (1) naive lookahead decoding is unreliable, (2) token stability closely correlates with convergence trend, and (3) historical information is isolated. Building on these insights, we propose Anchor-based History-stable Decoding (AHD), a training-free, plug-and-play dynamic decoding strategy. Specifically, AHD monitors the stability trend of tokens in real time through dynamic anchors. Once a token reaches stability, it initiates early cross-block decoding to enhance efficiency and performance. Extensive experiments across language, vision-language, and audio-language domains demonstrate that AHD simultaneously improves both performance and inference efficiency. Notably, AHD effectively reverses the performance degradation typically observed in existing advanced decoding acceleration strategies. For instance, on the BBH benchmark, our approach reduces decoding steps by 80% while improving performance by 3.67%.

While Multi-Agent Systems (MAS) are increasingly deployed for complex workflows, their emergent properties-particularly the accumulation of bias-remain poorly understood. Because real-world MAS are too complex to analyze entirely, evaluating their ethical robustness requires first isolating their foundational mechanics. In this work, we conduct a baseline empirical study investigating how basic MAS topologies and feedback loops influence prejudice. Contrary to the assumption that multi-agent collaboration naturally dilutes bias, we hypothesize that structured workflows act as echo chambers, amplifying minor stochastic biases into systemic polarization. To evaluate this, we introduce Discrim-Eval-Open, an open-ended benchmark that bypasses individual model neutrality through forced comparative judgments across demographic groups. Analyzing bias cascades across various structures reveals that architectural sophistication frequently exacerbates bias rather than mitigating it. We observe systemic amplification even when isolated agents operate neutrally, and identify a 'Trigger Vulnerability' where injecting purely objective context drastically accelerates polarization. By stripping away advanced swarm complexity to study foundational dynamics, we establish a crucial baseline: structural complexity does not guarantee ethical robustness. Our code is available at https://github.com/weizhihao1/MAS-Bias.

Offline goal-conditioned reinforcement learning (GCRL) is a practical reinforcement learning paradigm that aims to learn goal-conditioned policies from reward-free offline data. Despite recent advances in hierarchical architectures such as HIQL, long-horizon control in offline GCRL remains challenging due to the limited expressiveness of Gaussian policies and the inability of high-level policies to generate effective subgoals. To address these limitations, we propose the goal-conditioned mean flow policy, which introduces an average velocity field into hierarchical policy modeling for offline GCRL. Specifically, the mean flow policy captures complex target distributions for both high-level and low-level policies through a learned average velocity field, enabling efficient action generation via one-step sampling. Furthermore, considering the insufficiency of goal representation, we introduce a LeJEPA loss that repels goal representation embeddings during training, thereby encouraging more discriminative representations and improving generalization. Experimental results show that our method achieves strong performance across both state-based and pixel-based tasks in the OGBench benchmark.

Reinforcement learning (RL) in robotics is often limited by the cost and risk of data collection, motivating experience transfer from a source task to a target task. Offline-to-online RL leverages prior data but typically assumes a given fixed dataset and does not address how to generate reliable data for transfer. We propose \textit{World Model-based Experience Transfer} (WOMBET), a framework that jointly generates and utilizes prior data. WOMBET learns a world model in the source task and generates offline data via uncertainty-penalized planning, followed by filtering trajectories with high return and low epistemic uncertainty. It then performs online fine-tuning in the target task using adaptive sampling between offline and online data, enabling a stable transition from prior-driven initialization to task-specific adaptation. We show that the uncertainty-penalized objective provides a lower bound on the true return and derive a finite-sample error decomposition capturing distribution mismatch and approximation error. Empirically, WOMBET improves sample efficiency and final performance over strong baselines on continuous control benchmarks, demonstrating the benefit of jointly optimizing data generation and transfer.

Adapting vision-language models to remote sensing imagery presents a fundamental challenge: both the visual and linguistic distributions of satellite data lie far outside natural image pretraining corpora. Despite this, prompting remains the dominant deployment paradigm, driven by the assumption that domain-specific language can guide frozen model representations toward specialized tasks. We test this assumption directly on a domain where the mismatch is prominent: cloud segmentation for satellite imagery. Using CLIPSeg on the CloudSEN12+ cloud segmentation benchmark, we evaluate 60 prompt variants spanning simple labels, domain terminology, appearance descriptors, and contextual cues, finding that every variant underperforms the zero-shot baseline (0.255 mIoU), with engineered prompts scoring as low as 0.07 mIoU. No amount of linguistic refinement bridges the gap between CLIP's natural image representations and satellite spectral imagery. In contrast, supervised fine-tuning with just 0.1% labeled data (~8 images) surpasses zero-shot performance overall, and 5-10% data recovers ~85% of maximum achievable mIoU. Full fine-tuning consistently outperforms low-rank adaptation by 0.03-0.09 mIoU, with the largest gaps for spectrally ambiguous classes, and at 0.5 to 1% labeled data, fine-tuning temporarily degrades performance on these classes before recovering, a supervision dip that aggregate mIoU can mask. For practitioners adapting vision-language models to specialized imagery, our results deliver a clear message: labeled data is not the expensive alternative to prompting; it is the worthwhile path.

Document Question Answering (DQA) involves generating answers from a document based on a user's query, representing a key task in document understanding. This task requires interpreting visual layouts, which has prompted recent studies to adopt multimodal Retrieval-Augmented Generation (RAG) that processes page images for answer generation. However, in multimodal RAG, visual DQA struggles to utilize a large number of images effectively, as the retrieval stage often retains only a few candidate pages (e.g., Top-4), causing informative but less visually salient content to be overlooked in favor of common yet low-information pages. To address this issue, we propose a Multi-Armed Bandit-based DQA framework (MAB-DQA) to explicitly model the varying importance of multiple implicit aspects in a query. Specifically, MAB-DQA decomposes a query into aspect-aware subqueries and retrieves an aspect-specific candidate set for each. It treats each subquery as an arm and uses preliminary reasoning results from a small number of representative pages as reward signals to estimate aspect utility. Guided by an exploration-exploitation policy, MAB-DQA dynamically reallocates retrieval budgets toward high-value aspects. With the most informative pages and their correlations, MAB-DQA generates the expected results. On four benchmarks, MAB-DQA shows an average improvement of 5%-18% over the state-of-the-art method, consistently enhancing document understanding. Code at https://github.com/ElephantOH/MAB-DQA.

While Large Language Models (LLMs) excel in various general domains, they exhibit notable gaps in the highly specialized, knowledge-intensive, and legally regulated Chinese tax domain. Consequently, while tax-related benchmarks are gaining attention, many focus on isolated NLP tasks, neglecting real-world practical capabilities. To address this issue, we introduce TaxPraBen, the first dedicated benchmark for Chinese taxation practice. It combines 10 traditional application tasks, along with 3 pioneering real-world scenarios: tax risk prevention, tax inspection analysis, and tax strategy planning, sourced from 14 datasets totaling 7.3K instances. TaxPraBen features a scalable structured evaluation paradigm designed through process of "structured parsing-field alignment extraction-numerical and textual matching", enabling end-to-end tax practice assessment while being extensible to other domains. We evaluate 19 LLMs based on Bloom's taxonomy. The results indicate significant performance disparities: all closed-source large-parameter LLMs excel, and Chinese LLMs like Qwen2.5 generally exceed multilingual LLMs, while the YaYi2 LLM, fine-tuned with some tax data, shows only limited improvement. TaxPraBen serves as a vital resource for advancing evaluations of LLMs in practical applications.

As Large Language Models (LLMs) become increasingly prevalent in text simplification, systematically evaluating their outputs across diverse prompting strategies and architectures remains a critical methodological challenge in both NLP research and Intelligent Tutoring Systems (ITS). Developing robust prompts is often hindered by the absence of structured, visual frameworks for comparative text analysis. While researchers typically rely on static computational scripts, educators are constrained to standard conversational interfaces -- neither paradigm supports systematic multi-dimensional evaluation of prompt-model permutations. To address these limitations, we introduce \textbf{MuTSE}\footnote{The project code and the demo have been made available for peer review at the following anonymized URL. https://osf.io/njs43/overview?view_only=4b4655789f484110a942ebb7788cdf2a, an interactive human-in-the-loop web application designed to streamline the evaluation of LLM-generated text simplifications across arbitrary CEFR proficiency targets. The system supports concurrent execution of $P \times M$ prompt-model permutations, generating a comprehensive comparison matrix in real-time. By integrating a novel tiered semantic alignment engine augmented with a linearity bias heuristic ($λ$), MuTSE visually maps source sentences to their simplified counterparts, reducing the cognitive load associated with qualitative analysis and enabling reproducible, structured annotation for downstream NLP dataset construction.

Multi-agent coordination under partial observability requires agents to share complementary private information. While recent methods optimize messages for intermediate objectives (e.g., reconstruction accuracy or mutual information), rather than decision quality, we introduce \textbf{SeqComm-DFL}, unifying the sequential communication with decision-focused learning for task performance. Our approach features \emph{value-aware message generation with sequential Stackelberg conditioning}: messages maximize receiver decision quality and are generated in priority order, with agents conditioning on their predecessors. The \emph{guidance potential} determined by their prosocial ordering. We extend Optimal Model Design to communication-augmented world models with QMIX factorization, enabling efficient end-to-end training via implicit differentiation. We prove information-theoretic bounds showing that communication value scales with coordination gaps and establish $\mathcal{O}(1/\sqrt{T})$ convergence for the bilevel optimization, where $T$ denotes the number of training iterations. On collaborative healthcare and StarCraft Multi-Agent Challenge (SMAC) benchmarks, SeqComm-DFL achieves four to six times higher cumulative rewards and over 13\% win rate improvements, enabling coordination strategies inaccessible under information asymmetry.

Medical Vision Language Models VLMs suffer from two failure modes that threaten safe deployment mis calibrated confidence and sensitivity to question rephrasing. We show they share a common cause, proximity to the decision boundary, by benchmarking five uncertainty quantification methods on MedGemma 4BIT across in distribution MIMIC CXR and outof distribution PadChest chest X ray datasets, with cross architecture validation on LLaVA RAD7B. For well calibrated single model methods, predictive entropy from one forward pass predicts which samples will flip under rephrasing AUROC 0.711 on MedGemma, 0.878 on LLaVARAD p 10 4, enabling a single entropy threshold to flag both unreliable and rephrase sensitive predictions. A five member LoRA ensemble fails under the MIMIC PadChest shift 42.9 ECE, 34.1 accuracy, though LLaVA RAD s ensemble does not collapse 69.1. MC Dropout achieves the best calibration ECE 4.3 and selective prediction coverage 21.5 at 5 risk, yet total entropy from a single forward pass outperforms the ensemble for both error detection AUROC 0.743 vs 0.657 and paraphrase screening. Simple methods win.

Multi-Task Learning (MTL) is a foundational machine learning problem that has seen extensive development over the past decade. Recently, various optimization-based MTL approaches have been proposed to learn multiple tasks simultaneously by altering the optimization trajectory. Although these methods strive to de-conflict and re-balance tasks, we empirically identify that their effectiveness is often undermined by an overlooked factor when employing advanced optimizers: the instant-derived gradients play only a marginal role in the actual parameter updates. This discrepancy prevents MTL frameworks from fully releasing its power on learning dynamics. Furthermore, we observe that Muon-a recently emerged advanced optimizer-inherently functions as a multi-task learner, which underscores the critical importance of the gradients used for its orthogonalization. To address these issues, we propose APT (Applicability of advanced oPTimizers), a framework featuring a simple adaptive momentum mechanism designed to balance the strengths between advanced optimizers and MTL. Additionally, we introduce a light direction preservation method to facilitate Muon's orthogonalization. Extensive experiments across four mainstream MTL datasets demonstrate that APT consistently augments existing MTL approaches, yielding substantial performance improvements.

Causal discovery from observational data remains a fundamental challenge in machine learning and statistics, particularly when variables represent inherently positive quantities such as gene expression levels, asset prices, company revenues, or population counts, which often follow multiplicative rather than additive dynamics. We propose the Hybrid Moment-Ratio Scoring (H-MRS) algorithm, a novel method for learning directed acyclic graphs (DAGs) from positive-valued data by combining moment-based scoring with log-scale regression. The key idea is that for positive-valued variables, the moment ratio $\frac{\mathbb{E}[X_j^2]}{\mathbb{E}[(\mathbb{E}[X_j \mid S])^2]}$ provides an effective criterion for causal ordering, where $S$ denotes candidate parent sets. H-MRS integrates log-scale Ridge regression for moment-ratio estimation with a greedy ordering procedure based on raw-scale moment ratios, followed by Elastic Net-based parent selection to recover the final DAG structure. Experiments on synthetic log-linear data demonstrate competitive precision and recall. The proposed method is computationally efficient and naturally respects positivity constraints, making it suitable for applications in genomics and economics. These results suggest that combining log-scale modeling with raw-scale moment ratios provides a practical framework for causal discovery in positive-valued domains.

Functional verification has become the most time-consuming phase in IC development, and Assertion-Based Verification (ABV) is key to reducing debugging time. However, existing LLM-based assertion generation methods typically pursue indiscriminate verification, aiming for maximal coverage without considering signal criticality, whereas industrial practice demands maximizing coverage with minimal verification cost. Consequently, identifying signals that have the greatest impact on design functionality and error propagation-enabling a shift from indiscriminate to targeted verification-remains a key challenge. To address this, we propose AgileAssert, a key signal-driven assertion generation framework that constructs RTL semantic graphs and identifies the top-K critical signals via a hybrid scoring and selection mechanism, followed by structure-aware RTL slicing to provide the LLM with precise targets and contextual information, thereby guiding LLMs to generate tightly constrained targeted assertions for efficient verification. Evaluated on block- and CPU-level designs, with an average 66.68% reduction in assertions, our approach outperforms three existing SOTA methods, and significantly improving coverage metrics while reducing input token consumption by 64%. In mutation testing, when our approach surpasses existing methods in error detection rate, the average number of assertions used decreases by 72.74%.

Human cognitive development is shaped not only by individual effort but by structured social interaction, where role-based exchanges such as those between a tutor and a learner, enable solutions that neither could achieve alone. Inspired by these developmental principles, we ask the question whether a tutor-student multi-agent system can create a synergistic effect by pushing Large Language Model (LLM) beyond what it can do within existing frameworks. To test the idea, we adopt autonomous coding problem domain where two agents instantiated from the same LLM assigned asymmetric roles: a student agent generates and iteratively refines solutions, while a tutor agent provides structured evaluative feedback without access to ground-truth answers. In our proposed framework (PETITE), we aim to extract better problem-solving performance from one model by structuring its interaction through complementary roles, rather than relying on stronger supervisory models or heterogeneous ensembles. Our model is evaluated on the APPS coding benchmark against state-of-the-art approaches of Self-Consistency, Self-Refine, Multi-Agent Debate, and Multi-Agent Review. The results show that our model achieves similar or higher accuracy while consuming significantly fewer tokens. These results suggest that developmentally grounded role-differentiated interaction structures provide a principled and resource-efficient paradigm for enhancing LLM problem-solving through structured peer-like interactions. Index Terms- Peer Tutoring, Scaffolding, Large Language Models, Multi-Agent Systems, Code Generation

The initial outpatient consultation is critical for clinical decision-making, yet it is often conducted by a single physician under time pressure, making it prone to cognitive biases and incomplete evidence capture. Although the Multi-Disciplinary Team (MDT) reduces these risks, they are costly and difficult to scale to real-time intake. We propose Aegle, a synchronous virtual MDT framework that brings MDT-level reasoning to outpatient consultations via a graph-based multi-agent architecture. Aegle formalizes the consultation state using a structured SOAP representation, separating evidence collection from diagnostic reasoning to improve traceability and bias control. An orchestrator dynamically activates specialist agents, which perform decoupled parallel reasoning and are subsequently integrated by an aggregator into a coherent clinical note. Experiments on ClinicalBench and a real-world RAPID-IPN dataset across 24 departments and 53 metrics show that Aegle consistently outperforms state-of-the-art proprietary and open-source models in documentation quality and consultation capability, while also improving final diagnosis accuracy. Our code is available at https://github.com/HovChen/Aegle.

Post-training paradigms for Large Language Models (LLMs), primarily Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), face a fundamental dilemma: SFT provides stability (low variance) but suffers from high fitting bias, while RL enables exploration (low bias) but grapples with high gradient variance. Existing unified optimization strategies often employ naive loss weighting, overlooking the statistical conflict between these distinct gradient signals. In this paper, we provide a rigorous theoretical analysis of this bias-variance trade-off and propose \textbf{DYPO} (Dynamic Policy Optimization), a unified framework designed to structurally mitigate this conflict. DYPO integrates three core components: (1) a \textit{Group Alignment Loss (GAL)} that leverages intrinsic group dynamics to significantly reduce RL gradient variance; (2) a \textit{Multi-Teacher Distillation} mechanism that corrects SFT fitting bias via diverse reasoning paths; and (3) a \textit{Dynamic Exploitation-Exploration Gating} mechanism that adaptively arbitrates between stable SFT and exploratory RL based on reward feedback. Theoretical analysis confirms that DYPO linearly reduces fitting bias and minimizes overall variance. Extensive experiments demonstrate that DYPO significantly outperforms traditional sequential pipelines, achieving an average improvement of 4.8\% on complex reasoning benchmarks and 13.3\% on out-of-distribution tasks. Our code is publicly available at https://github.com/Tocci-Zhu/DYPO.

Dimensional Aspect-Based Sentiment Analysis (DimABSA) extends traditional ABSA from categorical polarity labels to continuous valence-arousal (VA) regression. This paper describes a system developed for Track A - Subtask 1 (Dimensional Aspect Sentiment Regression), aiming to predict real-valued VA scores in the [1, 9] range for each given aspect in a text. A fine-tuning approach based on XLM-RoBERTa-base is adopted, constructing the input as [CLS] T [SEP] a_i [SEP] and training dual regression heads with sigmoid-scaled outputs for valence and arousal prediction. Separate models are trained for each language-domain combination (English and Chinese across restaurant, laptop, and finance domains), and training and development sets are merged for final test predictions. In development experiments, the fine-tuning approach is compared against several large language models including GPT-5.2, LLaMA-3-70B, LLaMA-3.3-70B, and LLaMA-4-Maverick under a few-shot prompting setting, demonstrating that task-specific fine-tuning substantially and consistently outperforms these LLM-based methods across all evaluation datasets. The code is publicly available at https://github.com/tongwu17/SemEval-2026-Task3-Track-A.

Information retrieval systems have traditionally optimized for topical relevance-the degree to which retrieved documents match a query. However, relevance only approximates a deeper goal: utility, namely, whether retrieved information helps accomplish a user's underlying task. The emergence of retrieval-augmented generation (RAG) fundamentally changes this paradigm. Retrieved documents are no longer consumed directly by users but instead serve as evidence for large language models (LLMs) that produce answers. As a result, retrieval effectiveness must be evaluated by its contribution to generation quality rather than by relevance-based ranking metrics alone. This tutorial argues that retrieval objectives are evolving from relevance-centric optimization toward LLM-centric utility. We present a unified framework covering LLM-agnostic versus LLM-specific utility, context-independent versus context-dependent utility, and the connection with LLM information needs and agentic RAG. By synthesizing recent advances, the tutorial provides conceptual foundations and practical guidance for designing retrieval systems aligned with the requirements of LLM-based information access.

Existing defect/anomaly generation methods often rely on few-shot learning, which overfits to specific defect categories due to the lack of large-scale paired defect editing data. This issue is aggravated by substantial variations in defect scale and morphology, resulting in limited generalization, degraded realism, and category consistency. We address these challenges by introducing UDG, a large-scale dataset of 300K normal-abnormal-mask-caption quadruplets spanning diverse domains, and by presenting UniDG, a universal defect generation foundation model that supports both reference-based defect generation and text instruction-based defect editing without per-category fine-tuning. UniDG performs Defect-Context Editing via adaptive defect cropping and structured diptych input format, and fuses reference and target conditions through MM-DiT multimodal attention. A two-stage training strategy, Diversity-SFT followed by Consistency-RFT, further improves diversity while enhancing realism and reference consistency. Extensive experiments on MVTec-AD and VisA show that UniDG outperforms prior few-shot anomaly generation and image insertion/editing baselines in synthesis quality and downstream single- and multi-class anomaly detection/localization. Code will be available at https://github.com/RetoFan233/UniDG.

This paper introduces Nemo-Nemo, a practical crash-fault tolerant (CFT) consensus protocol designed to outperform existing protocols in wide-area networks by bridging design principles from the CFT and Byzantine-fault tolerant (BFT) worlds. By structuring command propagation through a causally ordered DAG, Nemo-Nemo allows all consensus replicas to propose commands with a naturally self-regulating communication regime. By exploiting multi-leader architecture, Nemo-Nemo avoids the performance bottleneck inherent to single-leader protocols. By separating command dissemination from consensus logic, Nemo-Nemo handles challenging network conditions even when consensus commits are stalled. Moreover, leader proposals that miss a deadline are never dropped, but deterministically deferred and executed later, preserving throughput under transient network delays. And by enabling Nemo-Nemo to commit on a DAG in just two network hops, it matches the latency of existing CFT systems, while achieving significantly higher throughput. The result is a robust, deployable system: the first DAG-based CFT consensus protocol proven to exceed state-of-the-art wide-area network performance in both speed and resilience.

Modern agentic frameworks (e.g., CrewAI and AutoGen) have evolved into complex, autonomous multi-agent systems, introducing unique reliability challenges beyond earlier pipeline-based LLM libraries. However, existing empirical studies focus on earlier LLM libraries or task-level bugs, leaving the unique complexities of these agentic frameworks unexplored. We bridge the gap by conducting a comprehensive study of 409 fixed bugs from five representative agentic frameworks. We propose a five-layer abstraction to capture structural complexities in agentic frameworks, spanning from orchestration to infrastructure. Our study uncovers specialized symptoms, such as unexpected execution sequences and user configurations ignored, which are unique to autonomous orchestration. We further identify agent-specific root causes, including modelrelated faults, cognitive context mismanagement, and orchestration faults. Statistical analysis reveals cross-framework consistency and significant associations among these bug dimensions. Finally, our automated pattern mining identifies frequent bug-triggering patterns (e.g., model backend-ID combinations), and we show their transferability across different framework designs. Our findings facilitate cross-platform testing and improve the reliability of agentic systems.

Reinforcement learning (RL) is effective in enhancing the accuracy of large language models in complex reasoning tasks. Existing RL policy optimization frameworks rely on final-answer correctness as feedback signals and rarely capture the internal logical structure of the reasoning process. Consequently, the models would generate fluent and semantically relevant responses but logically inconsistent, structurally erratic, or redundant. To this end, we propose StaRPO, a stability-augmented reinforcement learning framework that explicitly incorporates reasoning stability into the optimization objective. Our StaRPO decomposes stability into two computable lightweight metrics: the Autocorrelation Function (ACF) to evaluate local step-to-step coherence, and Path Efficiency (PE) to evaluate global goal-directedness of the reasoning trajectory. These stability rewards are combined with task rewards to provide complementary and process-aware feedback. We validate the effectiveness of using ACF and PE rewards by showing their correlation with logic errors on two backbone models. Experiments on four reasoning benchmarks show that StaRPO consistently outperforms compared baselines and can enhance both final-answer accuracy and logical stability.

Background: Limited data utilization in low-resource settings poses a barrier to the vaccine delivery ecosystem, undermining efforts to achieve equitable immunization coverage. In nomadic populations, individuals face an increased risk of missing crucial vaccination doses as children. One such population is the Maasai in Narok County, Kenya, where the absence of high-volume, quality data hampers accurate coverage estimates, impedes efficient resource allocation, and weakens the ability to deliver timely interventions. Additionally, data privacy concerns are heightened in groups with limited sensitive data. Objectives: First, we aim to identify children at risk of missing key vaccines across a large population to provide timely, evidence-based interventions that support increased vaccination coverage. Second, we aim to better protect the privacy of sensitive health data in a vulnerable population. Methods: We digitized 8 years of child vaccination records from the MOH 510 registry (n=6,913) and applied machine learning models (Logistic Regression and XGBoost) to identify children at risk. Additionally, we utilize a novel approach to tabular diffusion-based synthetic data generation (TabSyn) to protect patient privacy within the models. Results: Our findings show that classification techniques can reliably and successfully predict children at risk of missing a vaccine, with recall, precision, and F1-scores exceeding 90% for some vaccines modeled. Additionally, training these models with synthetic data rather than real data, thus preserving the privacy of individuals within the original dataset, does not lead to a loss in predictive performance. Conclusion: These results support the use of synthetic data implementation in health informatics strategies for clinics with limited digital infrastructure, enabling privacy-preserving, scalable forecasting for childhood immunization coverage.

Spiking Neural Networks (SNNs) offer superior energy efficiency over Artificial Neural Networks (ANNs). However, they encounter significant deficiencies in training and inference metrics when applied to Spiking Vision Transformers (S-ViTs). Existing paradigms including ANN-SNN Conversion and Spatial-Temporal Backpropagation (STBP) suffer from inherent limitations, precluding concurrent optimization of memory, accuracy and energy consumption. To address these issues, we propose Ge$^\text{2}$mS-T, a novel architecture implementing grouped computation across temporal, spatial and network structure dimensions. Specifically, we introduce the Grouped-Exponential-Coding-based IF (ExpG-IF) model, enabling lossless conversion with constant training overhead and precise regulation for spike patterns. Additionally, we develop Group-wise Spiking Self-Attention (GW-SSA) to reduce computational complexity via multi-scale token grouping and multiplication-free operations within a hybrid attention-convolution framework. Experiments confirm that our method can achieve superior performance with ultra-high energy efficiency on challenging benchmarks. To our best knowledge, this is the first work to systematically establish multi-dimensional grouped computation for resolving the triad of memory overhead, learning capability and energy budget in S-ViTs.

Glioma is a harmful brain tumor that requires early detection to ensure better health results. Early detection of this tumor is key for effective treatment and requires an automated segmentation process. However, it is a challenging task to find tumors due to tumor characteristics like location and size. A reliable method to accurately separate tumor zones from healthy tissues is deep learning models, which have shown promising results over the last few years. In this research, an Adaptive Dual Residual U-Net with Attention Gate and Multiscale Spatial Attention Mechanisms (ADRUwAMS) is introduced. This model is an innovative combination of adaptive dual residual networks, attention mechanisms, and multiscale spatial attention. The dual adaptive residual network architecture captures high-level semantic and intricate low-level details from brain images, ensuring precise segmentation of different tumor parts, types, and hard regions. The attention gates use gating and input signals to compute attention coefficients for the input features, and multiscale spatial attention generates scaled attention maps and combines these features to hold the most significant information about the brain tumor. We trained the model for 200 epochs using the ReLU activation function on BraTS 2020 and BraTS 2019 datasets. These improvements resulted in high accuracy for tumor detection and segmentation on BraTS 2020, achieving dice scores of 0.9229 for the whole tumor, 0.8432 for the tumor core, and 0.8004 for the enhancing tumor.

In Bayesian optimization, Thompson sampling selects the evaluation point by sampling from the posterior distribution over the objective function maximizer. Because this sampling problem is intractable for Gaussian process (GP) surrogates, the posterior distribution is typically restricted to fixed discretizations (i.e., candidate points) that become exponentially sparse as dimensionality increases. While previous works aim to increase candidate point density through scalable GP approximations, our orthogonal approach increases density by adaptively reducing the search space during sampling. Specifically, we introduce Adaptive Candidate Thompson Sampling (ACTS), which generates candidate points in subspaces guided by the gradient of a surrogate model sample. ACTS is a simple drop-in replacement for existing TS methods -- including those that use trust regions or other local approximations -- producing better samples of maxima and improved optimization across synthetic and real-world benchmarks.

Modeling causal relationships in graph representation learning remains a fundamental challenge. Existing approaches often draw on theories and methods from causal inference to identify causal subgraphs or mitigate confounders. However, due to the inherent complexity of graph-structured data, these approaches frequently aggregate diverse graph elements into single causal variables, an operation that risks violating the core assumptions of causal inference. In this work, we prove that such aggregation compromises causal validity. Building on this conclusion, we propose a theoretical model grounded in the smallest indivisible units of graph data to ensure that the causal validity is guaranteed. With this model, we further analyze the costs of achieving precise causal modeling in graph representation learning and identify the conditions under which the problem can be simplified. To empirically support our theory, we construct a controllable synthetic dataset that reflects realworld causal structures and conduct extensive experiments for validation. Finally, we develop a causal modeling enhancement module that can be seamlessly integrated into existing graph learning pipelines, and we demonstrate its effectiveness through comprehensive comparative experiments.

AI development is embracing open-source paradigm, but the fundamental distinction between AI models and traditional software artifacts may lead to a divergent open-source development paradigm with different collaborative practices, which remains unexplored. We therefore bridge the knowledge gap by quantifying and characterizing the differences in the collaborative development paradigms of traditional open source software (OSS) and open source AI models (OSM), and investigating the underlying factors that may drive these distinctions. We collect 1,428,792 OSS repositories from GitHub and 1,440,527 OSM repositories from HF Hub, and conduct comprehensive statistical, social network and content analyses to measure and understand the differences in collaboration intensity, collaboration openness, and user innovation across the two development paradigms, complementing these quantitative results with semi-structured interviews. In consequence, we find that compared to OSS development paradigm, the OSM development paradigm exhibits significantly lower collaboration intensity; lower collaboration openness regarding direct contribution while persisting relatively open knowledge exchange; and a divergence toward adaptive utilization user-innovation rather than collaborative improvement. Through semi-structured interviews, we further elucidate the socio-technical factors underlying these differences. These findings reveal the paradigmatic divergence in open source development between traditional OSS and OSM across three critical dimensions of open source collaboration and potential underlying factors, shedding light on how to improve collaborative work techniques and practices within the context of AI development.

Toxic interactions in open-source software development harm community collaboration. To combat this, we propose ToxiShield, a realtime browser extension that identifies and detoxifies toxic code reviews. The framework comprises three modules: toxicity identification, reasoned multiclass classification, and code review detoxification. Our fine-tuned BERT-based binary classifier achieved a 97% F1-score on 38,761 code review texts. For multiclass classification, Claude 3.5 Sonnet with prompt engineering achieved a 39% MCC and 42% F1 on 1,200 samples. Finally, our fine-tuned Llama 3.2 detoxification model reached 95.27% style transfer accuracy, 97.03% fluency, 67.07% content preservation, and an 84% J-score. Validation with 10 software developers suggests ToxiShield effectively fosters a more inclusive open-source environment.

Transformers have had a profound impact on the field of artificial intelligence, especially on large language models and their variants. However, as was the case with neural networks, their black-box nature limits trust and deployment in high-stakes settings. For models to be genuinely useful and trustworthy in critical applications, they must provide more than just predictions: they must supply users with a clear understanding of the reasoning that underpins their decisions. This article presents an uncertainty quantification framework for transformer-based language models. This framework, called CONFIDE (CONformal prediction for FIne-tuned DEep language models), applies conformal prediction to the internal embeddings of encoder-only architectures, like BERT and RoBERTa, while enabling hyperparameter tuning. CONFIDE uses either [CLS] token embeddings or flattened hidden states to construct class-conditional nonconformity scores, enabling statistically valid prediction sets with instance-level explanations. Empirically, CONFIDE improves test accuracy by up to 4.09% on BERT-tiny and achieves greater correct efficiency (i.e., the expected size of the prediction set conditioned on it containing the true label) compared to prior methods, including NM2 and VanillaNN. We show that early and intermediate transformer layers often yield better-calibrated and more semantically meaningful representations for conformal prediction. In resource-constrained models and high-stakes tasks with ambiguous labels, CONFIDE offers robustness and interpretability where softmax-based uncertainty fails. We position CONFIDE as a framework for practical diagnostic and efficiency/robustness improvement over prior conformal baselines.

While multimodal large language models (MLLMs) have made significant strides in natural image understanding, their ability to perceive and reason over hyperspectral image (HSI) remains underexplored, which is a vital modality in remote sensing. The high dimensionality and intricate spectral-spatial properties of HSI pose unique challenges for models primarily trained on RGB data.To address this gap, we introduce Hyperspectral Multimodal Benchmark (HM-Bench), the first benchmark designed specifically to evaluate MLLMs in HSI understanding. We curate a large-scale dataset of 19,337 question-answer pairs across 13 task categories, ranging from basic perception to spectral reasoning. Given that existing MLLMs are not equipped to process raw hyperspectral cubes natively, we propose a dual-modality evaluation framework that transforms HSI data into two complementary representations: PCA-based composite images and structured textual reports. This approach facilitates a systematic comparison of different representation for model performance. Extensive evaluations on 18 representative MLLMs reveal significant difficulties in handling complex spatial-spectral reasoning tasks. Furthermore, our results demonstrate that visual inputs generally outperform textual inputs, highlighting the importance of grounding in spectral-spatial evidence for effective HSI understanding. Dataset and appendix can be accessed at https://github.com/HuoRiLi-Yu/HM-Bench.

Inspired by the general Vision-and-Language Navigation (VLN) task, aerial VLN has attracted widespread attention, owing to its significant practical value in applications such as logistics delivery and urban inspection. However, existing methods face several challenges in complex urban environments, including insufficient generalization to unseen scenes, suboptimal performance in long-range path planning, and inadequate understanding of spatial continuity. To address these challenges, we propose HTNav, a new collaborative navigation framework that integrates Imitation Learning (IL) and Reinforcement Learning (RL) within a hybrid IL-RL framework. This framework adopts a staged training mechanism to ensure the stability of the basic navigation strategy while enhancing its environmental exploration capability. By integrating a tiered decision-making mechanism, it achieves collaborative interaction between macro-level path planning and fine-grained action control. Furthermore, a map representation learning module is introduced to deepen its understanding of spatial continuity in open domains. On the CityNav benchmark, our method achieves state-of-the-art performance across all scene levels and task difficulties. Experimental results demonstrate that this framework significantly improves navigation precision and robustness in complex urban environments.

Chain-of-thought (CoT) distillation transfers reasoning behaviors from a strong teacher to a smaller student, but prior work reports a capacity gap: distillation may fail when the teacher-student capability mismatch is large. We revisit the capacity gap from a practical perspective by re-examining commonly used experimental settings. Notably, we find that CoT distillation often degrades performance compared to the student's pre-distillation baseline, an issue obscured when only post-distillation comparisons are reported. We therefore propose a more realistic evaluation protocol and find that the impact of capacity gap effects does not consistently dominate across tasks and settings, especially when candidate teachers differ substantially in performance. Our results offer practical guidance for selecting teacher-student pairs in CoT distillation.

Moving beyond the traditional binary classification paradigm of Multimodal Sarcasm Detection, Multimodal Sarcasm Target Identification (MSTI) presents a more formidable challenge, requiring precise localization of fine-grained targets such as textual phrases and visual regions. Existing approaches predominantly rely on implicit cross-modal alignment, offering limited interpretability and suboptimal fine-grained localization. To address these limitations, we propose GRASP, Grounded Chain-of-Thought ReAsoning with Dual-Stage Optimization for Multimodal Sarcasm Prediction and Target Identification, a framework that integrates visual grounding with explicit Chain-of-Thought (CoT) reasoning to move beyond black-box MSTI. Specifically, we curate MSTI-MAX, a refined dataset that mitigates class imbalance and enriches multimodal sarcasm cues. We introduce Grounded CoT reasoning, which explicitly anchors sarcasm-related visual regions within the reasoning trajectory and prompts the model to articulate rationales before predicting the final classification labels and sarcasm targets. Furthermore, we employ a dual-stage outcome-supervised joint optimization strategy: Supervised Fine-Tuning with a coordinate-aware weighted loss, followed by Fine-Grained Target Policy Optimization. Extensive experiments demonstrate that GRASP outperforms existing baselines in fine-grained sarcasm target identification across modalities, and an LLM-as-a-Judge evaluation quantitatively measures the quality of internal reasoning chains. Our dataset and source code will be released on GitHub.

This study proposes a temporal modeling framework with a counterfactual policy-simulation layer for student dropout in higher education, using LMS engagement data and administrative withdrawal records. Dropout is operationalized as a time-to-event outcome at the enrollment level; weekly risk is modeled in discrete time via penalized, class-balanced logistic regression over person--period rows. Under a late-event temporal holdout, the model attains row-level AUCs of 0.8350 (train) and 0.8405 (test), with aggregate calibration acceptable but sparsely supported in the highest-risk bins. Ablation analyses indicate performance is sensitive to feature set composition, underscoring the role of temporal engagement signals. A scenario-indexed policy layer produces survival contrasts $ΔS(T)$ under an explicit trigger/schedule contract: positive contrasts are confined to the shock branch ($T_{\rm policy}=18$: 0.0102, 0.0260, 0.0819), while the mechanism-aware branch is negative ($ΔS_{\rm mech}(18)=-0.0078$, $ΔS_{\rm mech}(38)=-0.0134$). A subgroup analysis by gender quantifies scenario-induced survival gaps via bootstrap; contrasts are directionally stable but small. Results are not causally identified; they demonstrate the framework's capacity for internal structural scenario comparison under observational data constraints.

Many language tasks can be modeled as classification problems where a large language model (LLM) is given a prompt and selects one among many possible answers. We show that the classification error in such problems scales as a power law in the number of classes. This has a dramatic consequence: the prediction error can be reduced substantially by splitting the overall task into a sequence of smaller classification problems, each with the same number of classes ("degree"). This tree-structured decomposition models chain-of-thought (CoT). It has been observed that CoT-based predictors perform better when they "think'", i.e., when they develop a deeper tree, thus decomposing the problem into a larger number of steps. We identify a critical threshold for the degree, below which thinking is detrimental, and above which there exists an optimal depth that minimizes the error. It is impossible to surpass this minimal error by increasing the depth of thinking.

Student dropout is a persistent concern in Learning Analytics, yet comparative studies frequently evaluate predictive models under heterogeneous protocols, prioritizing discrimination over temporal interpretability and calibration. This study introduces a survival-oriented benchmark for temporal dropout risk modelling using the Open University Learning Analytics Dataset (OULAD). Two harmonized arms are compared: a dynamic weekly arm, with models in person-period representation, and a comparable continuous-time arm, with an expanded roster of families -- tree-based survival, parametric, and neural models. The evaluation protocol integrates four analytical layers: predictive performance, ablation, explainability, and calibration. Results are reported within each arm separately, as a single cross-arm ranking is not methodologically warranted. Within the comparable arm, Random Survival Forest leads in discrimination and horizon-specific Brier scores; within the dynamic arm, Poisson Piecewise-Exponential leads narrowly on integrated Brier score within a tight five-family cluster. No-refit bootstrap sampling variability qualifies these positions as directional signals rather than absolute superiority. Ablation and explainability analyses converged, across all families, on a shared finding: the dominant predictive signal was not primarily demographic or structural, but temporal and behavioral. Calibration corroborated this pattern in the better-discriminating models, with the exception of XGBoost AFT, which exhibited systematic bias. These results support the value of a harmonized, multi-dimensional benchmark in Learning Analytics and situate dropout risk as a temporal-behavioral process rather than a function of static background attributes.

To ensure safe clinical integration, deep learning models must provide more than just high accuracy; they require dependable uncertainty quantification. While current Medical Vision Transformers perform well, they frequently struggle with overconfident predictions and a lack of transparency, issues that are magnified by the noisy and imbalanced nature of clinical data. To address this, we enhanced the modified Medical Transformer (MedFormer) that incorporates prototype-based learning and uncertainty-guided routing, by utilizing a Dirichlet distribution for per-token evidential uncertainty, our framework can quantify and localize ambiguity in real-time. This uncertainty is not just an output but an active participant in the training process, filtering out unreliable feature updates. Furthermore, the use of class-specific prototypes ensures the embedding space remains structured, allowing for decisions based on visual similarity. Testing across four modalities (mammography, ultrasound, MRI, and histopathology) confirms that our approach significantly enhances model calibration, reducing expected calibration error (ECE) by up to 35%, and improves selective prediction, even when accuracy gains are modest.

Audio has rapidly become a primary interface for foundation models, powering real-time voice assistants. Ensuring safety in audio systems is inherently more complex than just "unsafe text spoken aloud": real-world risks can hinge on audio-native harmful sound events, speaker attributes (e.g., child voice), impersonation/voice-cloning misuse, and voice-content compositional harms, such as child voice plus sexual content. The nature of audio makes it challenging to develop comprehensive benchmarks or guardrails against this unique risk landscape. To close this gap, we conduct large-scale red teaming on audio systems, systematically uncover vulnerabilities in audio, and develop a comprehensive, policy-grounded audio risk taxonomy and AudioSafetyBench, the first policy-based audio safety benchmark across diverse threat models. AudioSafetyBench supports diverse languages, suspicious voices (e.g., celebrity/impersonation and child voice), risky voice-content combinations, and non-speech sound events. To defend against these threats, we propose AudioGuard, a unified guardrail consisting of 1) SoundGuard for waveform-level audio-native detection and 2) ContentGuard for policy-grounded semantic protection. Extensive experiments on AudioSafetyBench and four complementary benchmarks show that AudioGuard consistently improves guardrail accuracy over strong audio-LLM-based baselines with substantially lower latency.

As organizations increasingly deploy AI as a teammate rather than a standalone tool, morally consequential mistakes often arise from joint human-AI workflows in which causality is ambiguous. We ask how people allocate responsibility in these hybrid-agent settings. Across four experiments (N = 1,801) in an AI-assisted lending context (e.g., discriminatory rejection, irresponsible lending, and low-harm filing errors), participants consistently attributed more responsibility to the human decision maker when the human was paired with AI than when paired with another human (by an average of 10 points on a 0-100 scale across studies). This AI-Induced Human Responsibility (AIHR) effect held across high and low harm scenarios and persisted even where self-serving blame-shifting (when the human in question was the self) would be expected. Process evidence indicates that AIHR is explained by inferences of agent autonomy: AI is seen as a constrained implementer, which makes the human the default locus of discretionary responsibility. Alternative mechanisms (mind perception; self-threat) did not account for the effect. These findings extend research on algorithm aversion, hybrid AI-human organizational behavior and responsibility gaps in technology by showing that AI-human teaming can increase (rather than dilute) human responsibility, with implications for accountability design in AI-enabled organizations.

Proximal Policy Optimization (PPO) is central to aligning Large Language Models (LLMs) in reasoning tasks with verifiable rewards. However, standard token-level PPO struggles in this setting due to the instability of temporal credit assignment over long Chain-of-Thought (CoT) horizons and the prohibitive memory cost of the value model. While critic-free alternatives like GRPO mitigate these issues, they incur significant computational overhead by requiring multiple samples for baseline estimation, severely limiting training throughput. In this paper, we introduce Sequence-Level PPO (SPPO), a scalable algorithm that harmonizes the sample efficiency of PPO with the stability of outcome-based updates. SPPO reformulates the reasoning process as a Sequence-Level Contextual Bandit problem, employing a decoupled scalar value function to derive low-variance advantage signals without multi-sampling. Extensive experiments on mathematical benchmarks demonstrate that SPPO significantly surpasses standard PPO and matches the performance of computation-heavy group-based methods, offering a resource-efficient framework for aligning reasoning LLMs.

Recovering analytical solutions of physical fields from visual observations is a fundamental yet underexplored capability for AI-assisted scientific reasoning. We study visual-to-symbolic analytical solution inference (ViSA) for two-dimensional linear steady-state fields: given field visualizations (and first-order derivatives) plus minimal auxiliary metadata, the model must output a single executable SymPy expression with fully instantiated numeric constants. We introduce ViSA-R2 and align it with a self-verifying, solution-centric chain-of-thought pipeline that follows a physicist-like pathway: structural pattern recognition solution-family (ansatz) hypothesis parameter derivation consistency verification. We also release ViSA-Bench, a VLM-ready synthetic benchmark covering 30 linear steady-state scenarios with verifiable analytical/symbolic annotations, and evaluate predictions by numerical accuracy, expression-structure similarity, and character-level accuracy. Using an 8B open-weight Qwen3-VL backbone, ViSA-R2 outperforms strong open-source baselines and the evaluated closed-source frontier VLMs under a standardized protocol.

While significant work has been done on personality recognition, the lack of multilingual datasets remains an unresolved challenge. To address this, we propose ADAM (Cross-Lingual (A)ttention (D)istillation with Personality-Guided Generative (A)ugmentation for (M)ultilingual Personality Recognition), a state-of-the-art approach designed to advance multilingual personality recognition. Our approach leverages an existing English-language personality dataset as the primary source and employs a large language model (LLM) for translationbased augmentation, enhanced by Personality-Informed Generative Augmentation (PIGA), to generate high-quality training data in multiple languages, including Japanese, Chinese, Malay, and French. We provide a thorough analysis to justify the effectiveness of these augmentation techniques. Building on these advancements, ADAM integrates Cross-Lingual Attention Distillation (CLAD) to train a model capable of understanding and recognizing personality traits across languages, bridging linguistic and cultural gaps in personality analysis. This research presents a thorough evaluation of the proposed augmentation method, incorporating an ablation study on recognition performance to ensure fair comparisons and robust validation. Overall, with PIGA augmentation, the findings demonstrate that CLAD significantly outperforms the standard BCE across all languages and personality traits, achieving notable improvements in average BA scores - 0.6332 (+0.0573) on the Essays dataset and 0.7448 (+0.0968) on the Kaggle dataset. The CLAD-trained model also demonstrated strong generalizability and achieved benchmark performance comparable to current leading encoder models. The model weight, dataset, and algorithm repository are available at https://research.jingjietan.com/?q=ADAM.

Independent Component Analysis (ICA) is a fundamental unsupervised learning technique foruncovering latent structure in data by separating mixed signals into their independent sources. While substantial progress has been made in establishing asymptotic identifiability guarantees for nonlinear ICA, the finite-sample statistical properties of learning algorithms remain poorly understood. This gap poses significant challenges for practitioners who must determine appropriate sample sizes for reliable source recovery. This paper presents a comprehensive finite-sample analysis of nonlinear ICA with neural network encoders, providing the first complete characterization with matching upper and lower bounds. Our theoretical development introduces three key technical contributions. First, we establish a direct relationship between excess risk and identification error that bypasses parameter-space arguments, thereby avoiding the rate degradation that would otherwise yield suboptimal scaling. Second, we prove matching information-theoretic lower bounds that confirm the optimality of our sample complexity results. Third, we extend our analysis to practical SGD optimization, showing that the same sample efficiency can be achieved with finite-iteration gradient descent under standard landscape assumptions. We validate our theoretical predictions through carefully designed simulation experiments. This gap points toward valuable future research on finite-sample behavior of neural network training and highlights the importance of our validated scaling laws for dimension and diversity.

Clinical trials are central to evidence-based medicine, yet many struggle to meet enrollment targets, despite the availability of over half a million trials listed on ClinicalTrials.gov, which attracts approximately two million users monthly. Existing retrieval techniques, largely based on keyword and embedding-similarity matching between patient profiles and eligibility criteria, often struggle with low recall, low precision, and limited interpretability due to complex constraints. We propose SatIR, a scalable clinical trial retrieval method based on constraint satisfaction, enabling high-precision and interpretable matching of patients to relevant trials. Our approach uses formal methods -- Satisfiability Modulo Theories (SMT) and relational algebra -- to efficiently represent and match key constraints from clinical trials and patient records. Beyond leveraging established medical ontologies and conceptual models, we use Large Language Models (LLMs) to convert informal reasoning regarding ambiguity, implicit clinical assumptions, and incomplete patient records into explicit, precise, controllable, and interpretable formal constraints. Evaluated on 59 patients and 3,621 trials, SatIR outperforms TrialGPT on all three evaluated retrieval objectives. It retrieves 32%-72% more relevant-and-eligible trials per patient, improves recall over the union of useful trials by 22-38 points, and serves more patients with at least one useful trial. Retrieval is fast, requiring 2.95 seconds per patient over 3,621 trials. These results show that SatIR is scalable, effective, and interpretable.

Multimodal Large Language Models (MLLMs) have been shown to be vulnerable to malicious queries that can elicit unsafe responses. Recent work uses prompt engineering, response classification, or finetuning to improve MLLM safety. Nevertheless, such approaches are often ineffective against evolving malicious patterns, may require rerunning the query, or demand heavy computational resources. Steering the activations of a frozen model at inference time has recently emerged as a flexible and effective solution. However, existing steering methods for MLLMs typically handle only a narrow set of safety-related concepts or struggle to adjust specific concepts without affecting others. To address these challenges, we introduce Dictionary-Aligned Concept Control (DACO), a framework that utilizes a curated concept dictionary and a Sparse Autoencoder (SAE) to provide granular control over MLLM activations. First, we curate a dictionary of 15,000 multimodal concepts by retrieving over 400,000 caption-image stimuli and summarizing their activations into concept directions. We name the dataset DACO-400K. Second, we show that the curated dictionary can be used to intervene activations via sparse coding. Third, we propose a new steering approach that uses our dictionary to initialize the training of an SAE and automatically annotate the semantics of the SAE atoms for safeguarding MLLMs. Experiments on multiple MLLMs (e.g., QwenVL, LLaVA, InternVL) across safety benchmarks (e.g., MM-SafetyBench, JailBreakV) show that DACO significantly improves MLLM safety while maintaining general-purpose capabilities.

We study whether low-rank spectral summaries of LoRA weight deltas can identify which fine-tuning objective was applied to a language model, and whether that geometric signal predicts downstream behavioral harm. In a pre-registered experiment on \texttt{Llama-3.2-3B-Instruct}, we manufacture 38 LoRA adapters across four categories: healthy SFT baselines, DPO on inverted harmlessness preferences, DPO on inverted helpfulness preferences, and activation-steering-derived adapters, and extract per-layer spectral features (norms, stable rank, singular-value entropy, effective rank, and singular-vector cosine alignment to a healthy centroid). Within a single training method (DPO), a logistic regression classifier achieves AUC~1.00 on binary drift detection, all six pairwise objective comparisons, and near-perfect ordinal severity ranking ($ρ\geq 0.956$). Principal component analysis on flattened weight deltas reveals that training objective is PC1 (AUC~1.00 for objective separation), orthogonal to training duration on PC2. Query-projection weights detect that drift occurred; value-projection weights identify which objective. Cross-method generalization fails completely: a DPO-trained classifier assigns every steering adapter a lower drift score than every DPO adapter (AUC~0.00). In a behavioral evaluation phase, DPO-inverted-harmlessness adapters show elevated harmful compliance on HEx-PHI prompts (mean ASR 0.266 vs.\ healthy 0.112, $Δ= +0.154$), with near-perfect dose--response ($ρ= 0.986$). The geometry-to-behavior rank correlation is $ρ= 0.72$ across 24 non-steered adapters. These results establish that within a controlled manufacturing regime, LoRA weight-space geometry carries objective identity, intensity ordering, and a coarse link to harmful compliance, and that cross-method monitoring requires per-method calibration.

Flow-based image generative models exhibit stable training and produce high quality samples when using multi-step sampling procedures. One-step generative models can produce high quality image samples but can be difficult to optimize as they often exhibit unstable training dynamics. Meanflow models exhibit excellent few-step sampling performance and tantalizing one-step sampling performance. Notably, MeanFlow models that achieve this have required extremely large training budgets. We significantly decrease the amount of computation and data budget it takes to train Meanflow models by noting and exploiting a particular discretization of the Meanflow objective that yields a consistency property which we formulate into a ``Discrete Meanflow'' (DMF) Training Curriculum. Initialized with a pretrained Flow Model, DMF curriculum reaches one-step FID 3.36 on CIFAR-10 in only 2000 epochs. We anticipate that faster training curriculums of Meanflow models, specifically those fine-tuned from existing Flow Models, drives efficient training methods of future one-step examples.

The Hierarchical Kernel Transformer (HKT) is a multi-scale attention mechanism that processes sequences at L resolution levels via trainable causal downsampling, combining level-specific score matrices through learned convex weights. The total computational cost is bounded by 4/3 times that of standard attention, reaching 1.3125x for L = 3. Four theoretical results are established. (i) The hierarchical score matrix defines a positive semidefinite kernel under a sufficient condition on the symmetrised bilinear form (Proposition 3.1). (ii) The asymmetric score matrix decomposes uniquely into a symmetric part controlling reciprocal attention and an antisymmetric part controlling directional attention; HKT provides L independent such pairs across scales, one per resolution level (Propositions 3.5-3.6). (iii) The approximation error decomposes into three interpretable components with an explicit non-Gaussian correction and a geometric decay bound in L (Theorem 4.3, Proposition 4.4). (iv) HKT strictly subsumes single-head standard attention and causal convolution (Proposition 3.4). Experiments over 3 random seeds show consistent gains over retrained standard attention baselines: +4.77pp on synthetic ListOps (55.10+-0.29% vs 50.33+-0.12%, T = 512), +1.44pp on sequential CIFAR-10 (35.45+-0.09% vs 34.01+-0.19%, T = 1,024), and +7.47pp on IMDB character-level sentiment (70.19+-0.57% vs 62.72+-0.40%, T = 1,024), all at 1.31x overhead.

Classifier-free Guidance (CFG) lets practitioners trade-off fidelity against diversity in Diffusion Models (DMs). The practicality of CFG is however hindered by DMs sampling cost. On the other hand, Consistency Models (CMs) generate images in one or a few steps, but existing guidance methods require knowledge distillation from a separate DM teacher, limiting CFG to Consistency Distillation (CD) methods. We propose Joint Flow Distribution Learning (JFDL), a lightweight alignment method enabling guidance in a pre-trained CM. With a pre-trained CM as an ordinary differential equation (ODE) solver, we verify with normality tests that the variance-exploding noise implied by the velocity fields from unconditional and conditional distributions is Gaussian. In practice, JFDL equips CMs with the familiar adjustable guidance knob, yielding guided images with similar characteristics to CFG. Applied to an original Consistency Trained (CT) CM that could only do conditional sampling, JFDL unlocks guided generation and reduces FID on both CIFAR-10 and ImageNet 64x64 datasets. This is the first time that CMs are able to receive effective guidance post-hoc without a DM teacher, thus, bridging a key gap in current methods for CMs.

Large foundation models have become central to modern machine learning, with performance scaling predictably with model size and data. However, training and deploying such models incur substantial computational and memory costs, motivating the development of low-precision training techniques. Recent work has demonstrated that 4-bit floating-point (FP4) formats--such as MXFP4 and NVFP4--can be successfully applied to linear GEMM operations in large language models (LLMs), achieving up to 4x improvements in compute throughput and memory efficiency compared to higher-precision baselines. In this work, we investigate the recently proposed HiFloat4 FP4 format for Huawei Ascend NPUs and systematically compare it with MXFP4 in large-scale training settings. All experiments are conducted on Ascend NPU clusters, with linear and expert GEMM operations performed entirely in FP4 precision. We evaluate both dense architectures (e.g., Pangu and LLaMA-style models) and mixture-of-experts (MoE) models, where both standard linear layers and expert-specific GEMMs operate in FP4. Furthermore, we explore stabilization techniques tailored to FP4 training that significantly reduce numerical degradation, maintaining relative error within 1% of full-precision baselines while preserving the efficiency benefits of 4-bit computation. Our results provide a comprehensive empirical study of FP4 training on NPUs and highlight the practical trade-offs between FP4 formats in large-scale dense and MoE models.

Content moderation systems classify images as safe or unsafe but lack spatial grounding and interpretability: they cannot explain what sensitive behavior was detected, who is involved, or where it occurs. We introduce the Sensitive Benchmark (SenBen), the first large-scale scene graph benchmark for sensitive content, comprising 13,999 frames from 157 movies annotated with Visual Genome-style scene graphs (25 object classes, 28 attributes including affective states such as pain, fear, aggression, and distress, 14 predicates) and 16 sensitivity tags across 5 categories. We distill a frontier VLM into a compact 241M student model using a multi-task recipe that addresses vocabulary imbalance in autoregressive scene graph generation through suffix-based object identity, Vocabulary-Aware Recall (VAR) Loss, and a decoupled Query2Label tag head with asymmetric loss, yielding a +6.4 percentage point improvement in SenBen Recall over standard cross-entropy training. On grounded scene graph metrics, our student model outperforms all evaluated VLMs except Gemini models and all commercial safety APIs, while achieving the highest object detection and captioning scores across all models, at $7.6\times$ faster inference and $16\times$ less GPU memory.

We present Loom, a computer architecture that executes programs compiled from C inside a looped transformer whose weights are derived analytically. The architecture implements a 22-opcode instruction set in 8 transformer layers. Each forward pass executes one instruction; the model is applied iteratively until the program counter reaches zero. The full machine state resides in a single tensor $X \in \mathbb{R}^{d \times n}$ of fixed size, and every step has fixed cost for fixed $d$ and $n$, independent of program length or execution history. The default configuration uses $d = 155$ and $n = 1024$, yielding 4.7 million parameters and 928 instruction slots. A compact configuration at $d = 146$ and $n = 512$ suffices for a 9$\times$9 Sudoku solver (284 instructions). The weights are program-independent: programs live in the state tensor, and the same fixed-weight model executes any compiled program. We make Loom source code publicly available at https://github.com/mkturkcan/Loom.

Real-time tsunami early warning relies on distributed sensor networks to infer seismic sources and seafloor motion. Optimizing these networks via Bayesian optimal experimental design (OED) is exceptionally challenging for systems governed by hyperbolic partial differential equations, which lack the spectral decay required by standard low-rank approximations. We present a scalable Bayesian OED framework for linear time-invariant systems. By reformulating the inverse problem in the data space, we transform OED into dense matrix subset selection. We propose a multi-GPU, Schur-complement-update-based, greedy algorithm that solves the OED problem using a pipelined approach that fully overlaps I/O with GPU computations. Our framework achieves near-perfect weak and strong scaling across hundreds of GPUs on Perlmutter and Frontier. Applied to the 2025 Gordon Bell Prize-winning digital twin for tsunami forecasting in the Cascadia Subduction Zone, we optimize a 175-sensor network, minimizing the uncertainty of a parameter field with over one billion degrees of freedom.

Graph neural networks (GNNs) are increasingly applied to physical design tasks such as congestion prediction and wirelength estimation, yet progress is hindered by inconsistent circuit representations and the absence of controlled evaluation protocols. We present R2G (RTL-to-GDSII), a multi-view circuit-graph benchmark suite that standardizes five stage-aware views with information parity (every view encodes the same attribute set, differing only in where features attach) over 30 open-source IP cores (up to $10^6$ nodes/edges). R2G provides an end-to-end DEF-to-graph pipeline spanning synthesis, placement, and routing stages, together with loaders, unified splits, domain metrics, and reproducible baselines. By decoupling representation choice from model choice, R2G isolates a confound that prior EDA and graph-ML benchmarks leave uncontrolled. In systematic studies with GINE, GAT, and ResGatedGCN, we find: (i) view choice dominates model choice, with Test R$^2$ varying by more than 0.3 across representations for a fixed GNN; (ii) node-centric views generalize best across both placement and routing; and (iii) decoder-head depth (3--4 layers) is the primary accuracy driver, turning divergent training into near-perfect predictions (R$^2$$>$0.99). Code and datasets are available at https://github.com/ShenShan123/R2G.

Scalable Vector Graphics (SVG) represent visual content as structured, editable code. Each element (path, shape, or text node) can be individually inspected, transformed, or removed. This structural editability is a main motivation for SVG generation, yet prevailing evaluation protocols primarily reduce the output to a single similarity score against a reference image or input texts, measuring how faithfully the result reproduces an image or follows the instructions, but not how well it preserves the structural properties that make SVG valuable. In particular, existing metrics cannot determine which generated elements contribute positively to overall visual quality, how visual concepts map to specific parts of the code, or whether the generated output supports meaningful downstream editing. We introduce element-level leave-one-out (LOO) analysis, inspired by the classic jackknife estimator. The procedure renders the SVG with and without each element, measures the resulting visual change, and derives a suite of structural quality metrics. Despite its simplicity, the jackknife's capacity to decompose an aggregate statistic into per-sample contributions translates directly to this setting. From a single mechanism, we obtain: (1) quality scores per element through LOO scoring that enable zero-shot artifact detection; (2) concept-element attribution that maps each element to the visual concept it serves; and (3) four structural metrics, purity, coverage, compactness, and locality, that quantify SVG modularity from complementary perspectives. We validate these metrics on over 19,000 edits (5 types) across 5 generation systems and 3 complexity tiers.

Sedentary behavior poses a major public health risk, being strongly linked to obesity, cardiovascular disease, and other chronic conditions. Accurately estimating sitting time is therefore critical for monitoring and improving individual health. This work addresses the problem in real-world office settings, where signals from the inertial measurement units (IMU) on a smartwatch were collected from office workers during their daily routines. We propose a method that estimates sitting time from the IMU signals by introducing the use of rotation vector sequences, derived from Euler angles, as a novel representation of movement dynamics. Experiments on a 34-hour dataset demonstrate that exploiting rotation vector sequences improves algorithm performance, highlighting their potential for robust sitting time estimation in natural environments.

In November 2025, the authors ran a workshop on the topic of what makes a good reinforcement learning (RL) environment for autonomous cyber defence (ACD). This paper details the knowledge shared by participants both during the workshop and shortly afterwards by contributing herein. The workshop participants come from academia, industry, and government, and have extensive hands-on experience designing and working with RL and cyber environments. While there is now a sizeable body of literature describing work in RL for ACD, there is nevertheless a great deal of tradecraft, domain knowledge, and common hazards which are not detailed comprehensively in a single resource. With a specific focus on building better environments to train and evaluate autonomous RL agents in network defence scenarios, including government and critical infrastructure networks, the contributions of this work are twofold: (1) a framework for decomposing the interface between RL cyber environments and real systems, and (2) guidelines on current best practice for RL-based ACD environment development and agent evaluation, based on the key findings from our workshop.

Digital firms routinely run many online experiments on shared user populations. When product decisions are compositional, such as combinations of interface elements, flows, messages, or incentives, the number of feasible interventions grows combinatorially, while available traffic remains limited. Overlapping experiments can therefore generate interaction effects that are poorly handled by decentralized A/B testing. We study how to design large-scale factorial experiments when the objective is not to estimate every treatment effect, but to identify a high-performing policy under a fixed experimentation budget. We propose a two-stage design that centralizes overlapping experiments into a single factorial problem and models expected outcomes as a low-rank tensor. In the first stage, the platform samples a subset of intervention combinations, uses tensor completion to infer performance on untested combinations, and eliminates weak factor levels using estimated marginal contributions. In the second stage, it applies sequential halving to the surviving combinations to select a final policy. We establish gap-independent simple-regret bounds and gap-dependent identification guarantees showing that the relevant complexity scales with the degrees of freedom of the low-rank tensor and the separation structure across factor levels, rather than the full factorial size. In an offline evaluation based on a product-bundling problem constructed from 100 million Taobao interactions, the proposed method substantially outperforms one-shot tensor completion and unstructured best-arm benchmarks, especially in low-budget and high-noise settings. These results show how centralized, policy-aware experimentation can make combinatorial product design operationally feasible at platform scale.

This paper considers AI model churn as an opportunity for frugal investigation of large AI models. It describes how the incessant push for ever more powerful AI systems leaves in its wake a collection of obsolete yet powerful AI models, discarded in a veritable scrapyard of AI production. This scrapyard offers a potent opportunity for resource-constrained experimentation into AI systems. As in the physical scrapyard, nothing ever truly disappears in the AI scrapyard, it is just waiting to be reconfigured into something else. Project Nudge-x is an example of what can emerge from the AI scrapyard. Nudge-x seeks to manipulate legacy AI models to describe how mining sites across the planet are impacting landscapes and lives. By sharing this collection of brutal landscape interventions with people and AI systems alike, Nudge-x creates a venue for the appreciation of a history sadly shared between AI and people.

During disasters, cascading failures across power grids, communication networks, and social behavior amplify community fear and undermine cooperation. Existing cyber-physical-social (CPS) models simulate these coupled dynamics but lack mechanisms for active intervention. We extend the CPS resilience model of Valinejad and Mili (2023) with control channels for three agencies, communication, power, and emergency management, and formulate the resulting system as a three-player non-zero-sum differential game solved via online actor-critic reinforcement learning. Simulations based on Hurricane Harvey data show 70% mean fear reduction with improved infrastructure recovery; cross-validation in the case of Hurricane Irma (without refitting) achieves 50% fear reduction, confirming generalizability.

Prompt optimization improves language models without updating their weights by searching for a better system prompt, but its effectiveness varies widely across tasks. We study what makes a task amenable to prompt optimization. We show that the reward variance across different system prompts can be decomposed into two components: variance among responses, which captures generation stochasticity, and variance among system prompts, which captures differences in system prompt quality. Prompt optimization succeeds when variance among system prompts is sufficiently large, but fails when variance among responses dominates the variance of the system prompts. Surprisingly, we further show that scaling to more user prompts can hurt optimization by reducing variance among system prompts, especially on heterogeneous datasets where different user prompts favor different system prompts. Motivated by this insight, we propose $p1$, a simple user prompt filtering method that selects a small subset of user prompts with high variance across candidate system prompts. This subset of user prompts allows one to distinguish a good system prompt from a bad one, making system optimization easier. Experiments on reasoning benchmarks show that $p1$ substantially improves prompt optimization over training on the full dataset and outperforms strong baselines such as GEPA. Notably, training on only two prompts from AIME 24 yields a system prompt that generalizes well to other reasoning benchmarks.

Stepping-stone intrusions (SSIs) are a prevalent network evasion technique in which attackers route sessions through chains of compromised intermediate hosts to obscure their origin. Effective SSI detection requires correlating the incoming and outgoing flows at each relay host at extremely low false positive rates -- a stringent requirement that renders classical statistical methods inadequate in operational settings. We apply ESPRESSO, a deep learning flow correlation model combining a transformer-based feature extraction network, time-aligned multi-channel interval features, and online triplet metric learning, to the problem of stepping-stone intrusion detection. To support training and evaluation, we develop a synthetic data collection tool that generates realistic stepping-stone traffic across five tunneling protocols: SSH, SOCAT, ICMP, DNS, and mixed multi-protocol chains. Across all five protocols and in both host-mode and network-mode detection scenarios, ESPRESSO substantially outperforms the state-of-the-art DeepCoFFEA baseline, achieving a true positive rate exceeding 0.99 at a false positive rate of $10^{-3}$ for standard bursty protocols in network-mode. We further demonstrate chain length prediction as a tool for distinguishing malicious from benign pivoting, and conduct a systematic robustness analysis revealing that timing-based perturbations are the primary vulnerability of correlation-based stepping-stone detectors.

Stories are key to transmitting values across cultures, but their interpretation varies across linguistic and cultural contexts. Thus, we introduce multilingual story moral generation as a novel culturally grounded evaluation task. Using a new dataset of human-written story morals collected across 14 language-culture pairs, we compare model outputs with human interpretations via semantic similarity, a human preference survey, and value categorization. We show that frontier models such as GPT-4o and Gemini generate story morals that are semantically similar to human responses and preferred by human evaluators. However, their outputs exhibit markedly less cross-linguistic variation and concentrate on a narrower set of widely shared values. These findings suggest that while contemporary models can approximate central tendencies of human moral interpretation, they struggle to reproduce the diversity that characterizes human narrative understanding. By framing narrative interpretation as an evaluative task, this work introduces a new approach to studying cultural alignment in language models beyond static benchmarks or knowledge-based tests.

User-space Adaptive Bitrate (ABR) algorithms cannot see the transport layer signals that matter most, such as minimum RTT and instantaneous delivery rate, and they respond to network changes only after damage has already propagated to the playout buffer. We present eBandit, a framework that relocates both network monitoring and ABR algorithm selection into the Linux kernel using eBPF. A lightweight epsilon-greedy Multi-Armed Bandit (MAB) runs inside a sockops program, evaluating three ABR heuristics against a reward derived from live TCP metrics. On an adversarial synthetic trace eBandit achieves $416.3 \pm 4.9$ cumulative QoE, outperforming the best static heuristic by $7.2\%$. On 42 real-world sessions eBandit achieves a mean QoE per chunk of $1.241$, the highest across all policies, demonstrating that kernel-resident bandit learning transfers to heterogeneous mobile conditions.

Patient-clinician communication is an asymmetric-information problem: patients often do not disclose fears, misconceptions, or practical barriers unless clinicians elicit them skillfully. Effective medical dialogue therefore requires reasoning under partial observability: clinicians must elicit latent concerns, confirm them through interaction, and respond in ways that guide patients toward appropriate care. However, existing medical dialogue benchmarks largely sidestep this challenge by exposing hidden patient state, collapsing elicitation into extraction, or evaluating responses without modeling what remains hidden. We present MedConceal, a benchmark with an interactive patient simulator for evaluating hidden-concern reasoning in medical dialogue, comprising 300 curated cases and 600 clinician-LLM interactions. Built from clinician-answered online health discussions, each case pairing clinician-visible context with simulator-internal hidden concerns derived from prior literature and structured using an expert-developed taxonomy. The simulator withholds these concerns from the dialogue agent, tracks whether they have been revealed and addressed via theory-grounded turn-level communication signals, and is clinician-reviewed for clinical plausibility. This enables process-aware evaluation of both task success and the interaction process that leads to it. We study two abilities: confirmation, surfacing hidden concerns through multi-turn dialogue, and intervention, addressing the primary concern and guiding the patient toward a target plan. Results show that no single system dominates: frontier models lead on different confirmation metrics, while human clinicians (N=159) remain strongest on intervention success. Together, these results identify hidden-concern reasoning under partial observability as a key unresolved challenge for medical dialogue systems.

Large language models (LLMs) suffer significant performance degradation when user instructions and context are distributed over multiple conversational turns, yet multi-turn (MT) interactions dominate chat interfaces. The routine approach of appending full chat history to prompts rapidly exhausts context windows, leading to increased latency, higher computational costs, and diminishing returns as conversations extend. We introduce MT-OSC, a One-off Sequential Condensation framework that efficiently and automatically condenses chat history in the background without disrupting the user experience. MT-OSC employs a Condenser Agent that uses a few-shot inference-based Condenser and a lightweight Decider to selectively retain essential information, reducing token counts by up to 72% in 10-turn dialogues. Evaluated across 13 state-of-the-art LLMs and diverse multi-turn benchmarks, MT-OSC consistently narrows the multi-turn performance gap - yielding improved or preserved accuracy across datasets while remaining robust to distractors and irrelevant turns. Our results establish MT-OSC as a scalable solution for multi-turn chats, enabling richer context within constrained input spaces, reducing latency and operational cost, while balancing performance.

Purpose: To develop and evaluate a deep learning (DL) method for free-breathing phase-sensitive inversion recovery (PSIR) late gadolinium enhancement (LGE) cardiac MRI that produces diagnostic-quality images from a single acquisition over two heartbeats, eliminating the need for 8 to 24 motion-corrected (MOCO) signal averages. Materials and Methods: Raw data comprising 800,653 slices from 55,917 patients, acquired on 1.5T and 3T scanners across multiple sites from 2016 to 2024, were used in this retrospective study. Data were split by patient: 640,000 slices (42,822 patients) for training and the remainder for validation and testing, without overlap. The training and testing data were from different institutions. PSIRNet, a physics-guided DL network with 845 million parameters, was trained end-to-end to reconstruct PSIR images with surface coil correction from a single interleaved IR/PD acquisition over two heartbeats. Reconstruction quality was evaluated using SSIM, PSNR, and NRMSE against MOCO PSIR references. Two expert cardiologists performed an independent qualitative assessment, scoring image quality on a 5-point Likert scale across bright blood, dark blood, and wideband LGE variants. Paired superiority and equivalence (margin = 0.25 Likert points) were tested using exact Wilcoxon signed-rank tests at a significance level of 0.05 using R version 4.5.2. Results: Both readers rated single-average PSIRNet reconstructions superior to MOCO PSIR for dark blood LGE (conservative P = .002); for bright blood and wideband, one reader rated it superior and the other confirmed equivalence (all P < .001). Inference required approximately 100 msec per slice versus more than 5 sec for MOCO PSIR. Conclusion: PSIRNet produces diagnostic-quality free-breathing PSIR LGE images from a single acquisition, enabling 8- to 24-fold reduction in acquisition time.

World models promise a paradigm shift in robotics, where an agent learns the underlying physics of its environment once to enable efficient planning and behavior learning. However, current world models are often hardware-locked specialists: a model trained on a Boston Dynamics Spot robot fails catastrophically on a Unitree Go1 due to the mismatch in kinematic and dynamic properties, as the model overfits to specific embodiment constraints rather than capturing the universal locomotion dynamics. Consequently, a slight change in actuator dynamics or limb length necessitates training a new model from scratch. In this work, we take a step towards a framework for training a generalizable Quadrupedal World Model (QWM) that disentangles environmental dynamics from robot morphology. We address the limitations of implicit system identification, where treating static physical properties (like mass or limb length) as latent variables to be inferred from motion history creates an adaptation lag that can compromise zero-shot safety and efficiency. Instead, we explicitly condition the generative dynamics on the robot's engineering specifications. By integrating a physical morphology encoder and a reward normalizer, we enable the model to serve as a neural simulator capable of generalizing across morphologies. This capability unlocks zero-shot control across a range of embodiments. We introduce, for the first time, a world model that enables zero-shot generalization to new morphologies for locomotion. While we carefully study the limitations of our method, QWM operates as a distribution-bounded interpolator within the quadrupedal morphology family rather than a universal physics engine, this work represents a significant step toward morphology-conditioned world models for legged locomotion.

Large language models (LLMs) have significant potential to improve operational efficiency in operations management. Deploying these models requires specifying a policy that governs response quality, shapes user experience, and influences operational value. In this research, we treat LLMs as stochastic simulators and propose a pairwise comparison-based adaptive simulation experiment framework for identifying the optimal policy from a finite set of candidates. We consider two policy spaces: an unstructured space with no parametric assumption, and a structured space in which the data are generated from a preference model. For both settings, we characterize the fundamental data requirements for identifying the optimal policy with high probability. In the unstructured case, we derive a closed-form expression for the optimal sampling proportions, together with a clear operational interpretation. In the structured case, we formulate a regularized convex program to compute the optimal proportions. We then develop an adaptive experimental procedure, termed LLM-PO, for both policy spaces, and prove that it identifies the optimal policy with the desired statistical guarantee while asymptotically attaining the fundamental data requirements. Numerical experiments demonstrate that LLM-PO consistently outperforms benchmark methods and improves LLM performance.

The rapid advancement of neuromorphic technology aims to address the memory wall challenge inherent in conventional von Neumann architectures. This paper critically examines current digital neuromorphic processors and their strategies to mitigate this bottleneck. While designed to bring computation closer to memory through distributed architectures, our findings indicate that on-chip memory systems, including SRAM and emerging technologies like STT-MRAM, have become significant consumers of area and energy, leading to a new memory wall. Through an analysis of energy and area efficiency in various memory technologies, we argue that without a re-evaluation of memory organization, digital neuromorphic processors may struggle to compete effectively in edge and embedded applications. We conclude with potential pathways for future research to overcome the limitations of on-chip memory in neuromorphic systems.

Atmospheric clouds exhibit complex three-dimensional structure and microphysical details that are poorly constrained by the predominantly two-dimensional satellite observations available at global scales. This mismatch complicates data-driven learning and evaluation of cloud processes in weather and climate models, contributing to ongoing uncertainty in atmospheric physics. We introduce CERBERUS, a probabilistic inference framework for generating vertical radar reflectivity profiles from geostationary satellite brightness temperatures, near-surface meteorological variables, and temporal context. CERBERUS employs a three-headed encoder-decoder architecture to predict a zero-inflated (ZI) vertically-resolved distribution of radar reflectivity. Trained and evaluated using ground-based Ka-band radar observations at the ARM Southern Great Plains site, CERBERUS recovers coherent structures across cloud regimes, generalizes to withheld test periods, and provides uncertainty estimates that reflect physical ambiguity, particularly in multilayer and dynamically complex clouds. These results demonstrate the value of distribution-based learning targets for bridging observational scales, introducing a path toward model-relevant synthetic observations of clouds.

Predicting group behavior, how individuals coordinate, communicate, and interact during collaborative tasks, is essential for designing systems that can support team performance through real-time prediction and realistic simulation of collaborative scenarios. Large Language Models (LLMs) have shown promise for processing sensor data for human-activity recognition (HAR), yet their capabilities for team dynamics or group-level multimodal sensing remain unexplored. This paper investigates whether LLMs can predict group coordination patterns from multimodal sensor data in collaborative Mixed Reality (MR) environments. We encode hierarchical context -- individual behavioral profiles, group structural properties, and temporal activity context -- as natural language and evaluate three LLM adaptation paradigms (zero-shot, few-shot, and supervised fine-tuning) against statistical baselines. Our evaluation on 16 groups (64 participants, $\sim$25 hours of sensor data) reveals that LLMs achieve 3.2$\times$ improvement over LSTM baselines for linguistically-grounded behaviors, with fine-tuning reaching 96\% accuracy for conversation prediction while maintaining sub-35ms latency. Beyond performance gains, we characterize the boundaries of text-based LLMs for multimodal sensing conversation prediction succeeds because turn-taking maps to linguistic patterns, while shared or joint attention may require spatial and visual reasoning that text only LLMs cannot capture. We further identify simulation mode brittleness (83\% degradation from cascading context errors) and minimal few-shot sensitivity to example selection strategy. These findings establish guidelines when LLMs are appropriate for CPS/IoT sensing for team dynamics and inform the design of future multimodal foundation models.

Since their introduction, Transformer architectures have dominated Natural Language Processing (NLP). However, recent research has highlighted an inherent anisotropy phenomenon in these models, presenting a significant challenge to their geometric interpretation. Previous theoretical studies on this phenomenon are rarely grounded in the underlying representation geometry. In this paper, we extend them by deriving geometric arguments for how frequency-biased sampling attenuates curvature visibility and why training preferentially amplify tangent directions. Empirically, we then use concept-based mechanistic interpretability during training, rather than only post hoc, to fit activation-derived low-rank tangent proxies and test them against ordinary backpropagated true gradients. Across encoder-style and decoder-style language models, we find that these activation-derived directions capture both unusually large gradient energy and a substantially larger share of gradient anisotropy than matched-rank normal controls, providing strong empirical support for a tangent-aligned account of anisotropy.

We extend the Weak Adversarial Neural Pushforward Method to the Wigner transport equation governing the phase-space dynamics of quantum systems. The central contribution is a structural observation: integrating the nonlocal pseudo-differential potential operator against plane-wave test functions produces a Dirac delta that exactly inverts the Fourier transform defining the Wigner potential kernel, reducing the operator to a pointwise finite difference of the potential at two shifted arguments. This holds in arbitrary dimension, requires no truncation of the Moyal series, and treats the potential as a black-box function oracle with no derivative information. To handle the negativity of the Wigner quasi-probability distribution, we introduce a signed pushforward architecture that decomposes the solution into two non-negative phase-space distributions mixed with a learnable weight. The resulting method inherits the mesh-free, Jacobian-free, and scalable properties of the original framework while extending it to the quantum setting.

Understanding instructional videos requires recognizing fine-grained actions and modeling their temporal relations, which remains challenging for current Video Foundation Models (VFMs). This difficulty stems from noisy web supervision and a pervasive "static bias", where models rely on objects rather than motion cues. To address this, we propose InstrAction, a pretraining framework for instructional videos' action-centric representations. We first introduce a data-driven strategy, which filters noisy captions and generates action-centric hard negatives to disentangle actions from objects during contrastive learning. At the visual feature level, an Action Perceiver extracts motion-relevant tokens from redundant video encodings. Beyond contrastive learning, we introduce two auxiliary objectives: Dynamic Time Warping alignment (DTW-Align) for modeling sequential temporal structure, and Masked Action Modeling (MAM) for strengthening cross-modal grounding. Finally, we introduce the InstrAct Bench to evaluate action-centric understanding, where our method consistently outperforms state-of-the-art VFMs on semantic reasoning, procedural logic, and fine-grained retrieval tasks.

This paper considers the problem of multi-bit generative watermarking for large language models under a worst-case false-alarm constraint. Prior work established a lower bound on the achievable miss-detection probability in the finite-token regime and proposed a scheme claimed to achieve this bound. We show, however, that the proposed scheme is in fact suboptimal. We then develop two new encoding-decoding constructions that attain the previously established lower bound, thereby completely characterizing the optimal multi-bit watermarking performance. Our approach formulates the watermark design problem as a linear program and derives the structural conditions under which optimality can be achieved. In addition, we identify the failure mechanism of the previous construction and compare the tradeoffs between the two proposed schemes.

Humor is one of the most culturally embedded and socially significant dimensions of human communication, yet it remains largely unexplored as a dimension of Large Language Model (LLM) alignment. In this study, five frontier language models play the same Cards Against Humanity games (CAH) as human players. The models select the funniest response from a slate of ten candidate cards across 9,894 rounds. While all models exceed the random baseline, alignment with human preference remains modest. More striking is that models agree with each other substantially more often than they agree with humans. We show that this preference is partly explained by systematic position biases and content preferences, raising the question whether LLM humor judgment reflects genuine preference or structural artifacts of inference and alignment.

The situated view of cognition holds that intelligent behavior depends not only on internal memory, but on an agent's active use of environmental resources. Here, we begin formalizing this intuition within Reinforcement Learning (RL). We introduce a mathematical framing for how the environment can functionally serve as an agent's memory, and prove that certain observations, which we call artifacts, can reduce the information needed to represent history. We corroborate our theory with experiments showing that when agents observe spatial paths, the amount of memory required to learn a performant policy is reduced. Interestingly, this effect arises unintentionally, and implicitly through the agent's sensory stream. We discuss the implications of our findings, and show they satisfy qualitative properties previously used to ground accounts of external memory. Moving forward, we anticipate further work on this subject could reveal principled ways to exploit the environment as a substitute for explicit internal memory.

Computational models support high-stakes decisions across engineering and science, and practitioners increasingly seek probabilistic predictions to quantify uncertainty in such models. Existing approaches generate predictions either by sampling input parameter distributions or by augmenting deterministic outputs with uncertainty representations, including distribution-free and distributional methods. However, sampling-based methods are often computationally prohibitive for real-time applications, and many existing uncertainty representations either ignore input dependence or rely on restrictive Gaussian assumptions that fail to capture asymmetry and heavy-tailed behavior. Therefore, we extend the ACCurate and Reliable Uncertainty Estimate (ACCRUE) framework to learn input-dependent, non-Gaussian uncertainty distributions, specifically two-piece Gaussian and asymmetric Laplace forms, using a neural network trained with a loss function that balances predictive accuracy and reliability. Through synthetic and real-world experiments, we show that the proposed approach captures an input-dependent uncertainty structure and improves probabilistic forecasts relative to existing methods, while maintaining flexibility to model skewed and non-Gaussian errors.

We introduce IKKA (Inversion Classification via Critical Anomalies), a topologically motivated weighting framework for robust visual servoing under distribution shift. Unlike conventional outlier handling, IKKA treats maverick points as structurally informative observations: points where small perturbations can induce qualitatively different control responses or class assignments. The method combines local extremality, boundary transversality, and multi-scale persistence into a single anomaly weight, W(x) = E(x) x T(x) x M(x), which modulates control updates near ambiguous decision regions. We instantiate IKKA in a CPU-only embedded visual-servoing pipeline on Raspberry Pi 4 and evaluate it across 230 reproducible runs under nominal and stress conditions. In stress scenarios involving dim illumination and transient occlusion, IKKA reduces the 95th-percentile lateral error by 24% relative to a hybrid baseline (0.124 to 0.094) while increasing throughput from 20.0 to 24.8 Hz. Non-parametric analysis confirms a large effect size (Cliff's delta = 0.79).

Relation extraction represents a fundamental component in the process of creating knowledge graphs, among other applications. Large language models (LLMs) have been adopted as a promising tool for relation extraction, both in supervised and in-context learning settings. However, in this work we show that their performance still lags behind much smaller architectures when the linguistic graph underlying a text has great complexity. To demonstrate this, we evaluate four LLMs against a graph-based parser on six relation extraction datasets with sentence graphs of varying sizes and complexities. Our results show that the graph-based parser increasingly outperforms the LLMs, as the number of relations in the input documents increases. This makes the much lighter graph-based parser a superior choice in the presence of complex linguistic graphs.

Plant-level control is an emerging wind energy technology that presents opportunities and challenges. By controlling turbines in a coordinated manner via a central controller, it is possible to achieve greater wind power plant efficiency. However, there is a risk that measurement errors will confound the process, or even that hackers will alter the telemetry signals received by the central controller. This paper presents a framework for developing a safe plant controller by training it with an adversarial agent designed to confound it. This necessitates training the adversary to confound the controller, creating a sort of circular logic or "Arms Race." This paper examines three broad training approaches for co-training the protagonist and adversary, finding that an Arms Race approach yields the best results. These initial results indicate that the Arms Race adversarial training reduced worst-case performance degradation from 39% power loss to 7.9% power gain relative to a baseline operational strategy.

How many of a neural network's parameters actually encode task-specific information? We investigate this question with LottaLoRA, a training paradigm in which every backbone weight is drawn at random and frozen; only low-rank LoRA adapters are trained. Across nine benchmarks spanning diverse architecture families from single-layer classifiers to 900M parameter Transformers low-rank adapters over frozen random backbones recover 96-100% of fully trained performance while training only 0.5-40% of the parameters. The task-specific signal therefore occupies a subspace orders of magnitude smaller than the full parameter count suggests.Three mechanistic findings underpin this result:(1) the frozen backbone is actively exploited when static the learned scaling~$β$ remains strictly positive across all architectures but when the scaffold is destabilized, the optimizer silences it and the LoRA factors absorb all task information; (2) the frozen backbone is preferable but interchangeable any random initialization works equally well, provided it remains fixed throughout training; and (3) the minimum LoRA rank at which performance saturates estimates the intrinsic dimensionality of the task, reminiscent of the number of components retained in Principal Component Analysis (PCA). The construction is formally analogous to Reservoir Computing unfolded along the depth axis of a feedforward network. Because the backbone is determined by a random seed alone, models can be distributed as adapters plus seed a footprint that grows with task complexity, not model size, so that storage and memory savings compound as architectures scale.

Modern software projects depend on third-party dependencies, whose declarations must be maintained as projects evolve. Prior work has focused on dependency version updates, while much less is known about how developers assign dependencies to different roles over time. In this paper, we investigate how developers of JavaScript projects reclassify their dependencies, including removal and role reassignment. Our analysis of 33,087 JavaScript projects with active dependency maintenance reveals that dependency reclassification is a prevalent maintenance activity, occurring in 79.1% of the studied projects. Of these projects, nearly all (97.2%) remove dependencies at some point, while 38.0% undergo role reassignments across Core (runtime), Dev (development-only), and Peer (consumer-provided) roles. These changes are not always final, as 33.1% of projects later reintroduce removed dependencies and 11.2% exhibit repeated role switching. Reclassification practices also tend to unfold over long periods, with a median duration of 408 days. These findings broaden the study of dependency maintenance beyond version updates, and contribute implications for dependency tools, package managers, and research to support correct dependency role declarations.

The discovery of new energetic materials is critical for advancing technologies from defense to private industry. However, experimental approaches remain slow and expensive while computational alternatives require accurate material property inputs that are often costly to obtain, limiting their ability to efficiently predict detonation performance across a vast chemical space. We address this challenge through an active learning strategy that integrates density functional theory calculations, thermochemical modeling, message-passing neural networks, and Bayesian optimization. The resulting high-throughput workflow iteratively expands the training dataset by selecting new molecules in a targeted manner that balances the exploration of broad chemical space with the exploitation of promising high-performing candidates. This approach yields the largest publicly available database of potential CHNO explosives drawn from an initial pool of more than 70 billion candidates and a generalizable surrogate model capable of accurately predicting detonation performance (R$^2$ > 0.98). Feature importance analysis on this largest-to-date dataset reveals that oxygen balance is the dominant driver of detonation performance, complemented by contributions from local electronic structure, density, and the presence of specific functional groups. Cheminformatics analysis highlights how energetic materials with similar performance metrics tend to cluster in distinct chemical spaces offering a clearer direction for future synthesis studies. Together, the surrogate model, database, and resulting chemical insights provide a valuable foundation for high-throughput screening and targeted discovery of new energetic materials spanning diverse and previously unexplored regions of chemical space.

Adam has achieved strong empirical success, but its theory remains incomplete even in the deterministic full-batch setting, largely because adaptive preconditioning and momentum are tightly coupled. In this work, a convergent reformulation of full-batch Adam is developed by combining variable and operator splitting with a curvature-aware gradient correction. This leads to a continuous-time Adam-HNAG flow with an exponentially decaying Lyapunov function, as well as two discrete methods: Adam-HNAG, and Adam-HNAG-s, a synchronous variant closer in form to Adam. Within a unified Lyapunov analysis framework, convergence guarantees are established for both methods in the convex smooth setting, including accelerated convergence. Numerical experiments support the theory and illustrate the different empirical behavior of the two discretizations. To the best of our knowledge, this provides the first convergence proof for Adam-type methods in convex optimization.

Ransomware poses a serious and fast-acting threat to critical systems, often encrypting files within seconds of execution. Research indicates that ransomware is the most reported cybercrime in terms of financial damage, highlighting the urgent need for early-stage detection before encryption is complete. In this paper, we present RansomTrack, a hybrid behavioral analysis framework to eliminate the limitations of using static and dynamic detection methods separately. Static features are extracted using the Radare2 sandbox, while dynamic behaviors such as memory protection changes, mutex creation, registry access and network activity are obtained using the Frida toolkit. Our dataset of 165 different ransomware and benign software families is publicly released, offering the highest family-to-sample ratio known in the literature. Experimental evaluation using machine learning models shows that ensemble classifiers such as XGBoost and Soft Voting achieve up to 96% accuracy and a ROC-AUC score of 0.99. Each sample analyzed in 9.1 seconds includes modular behavioral logging, runtime instrumentation, and SHAP-based interpretability to highlight the most influential features. Additionally, RansomTrack framework is able to detect ransomware under 9.2 seconds. Overall, RansomTrack offers a scalable, low-latency, and explainable solution for real-time ransomware detection.

Cooperation in multi-agent reinforcement learning (MARL) benefits from inter-agent communication, yet most approaches assume idealized channels and existing value decomposition methods ignore who successfully shared information with whom. We propose CLOVER, a cooperative MARL framework whose centralized value mixer is conditioned on the communication graph realized under a realistic wireless channel. This graph introduces a relational inductive bias into value decomposition, constraining how individual utilities are mixed based on the realized communication structure. The mixer is a GNN with node-specific weights generated by a Permutation-Equivariant Hypernetwork: multi-hop propagation along communication edges reshapes credit assignment so that different topologies induce different mixing. We prove this mixer is permutation invariant, monotonic (preserving the IGM condition), and strictly more expressive than QMIX-style mixers. To handle realistic channels, we formulate an augmented MDP isolating stochastic channel effects from the agent computation graph, and employ a stochastic receptive field encoder for variable-size message sets, enabling end-to-end differentiable training. On Predator-Prey and Lumberjacks benchmarks under p-CSMA wireless channels, CLOVER consistently improves convergence speed and final performance over VDN, QMIX, TarMAC+VDN, and TarMAC+QMIX. Behavioral analysis confirms agents learn adaptive signaling and listening strategies, and ablations isolate the communication-graph inductive bias as the key source of improvement.

Preference optimization methods such as DPO and KTO are widely used for aligning language models, yet little is understood about what properties of preference data drive downstream reasoning gains. We ask: what aspects of a preference pair improve a reasoning model's performance on general reasoning tasks? We investigate two distinct notions of quality delta in preference data: generator-level delta, arising from the differences in capability between models that generate chosen and rejected reasoning traces, and sample-level delta, arising from differences in judged quality differences within an individual preference pair. To study generator-level delta, we vary the generator's scale and model family, and to study sample-level delta, we employ an LLM-as-a-judge to rate the quality of generated traces along multiple reasoning-quality dimensions. We find that increasing generator-level delta steadily improves performance on out-of-domain reasoning tasks and filtering data by sample-level delta can enable more data-efficient training. Our results suggest a twofold recipe for improving reasoning performance through preference optimization: maximize generator-level delta when constructing preference pairs and exploit sample-level delta to select the most informative training examples.

Sport analysis is crucial for team performance since it provides actionable data that can inform coaching decisions, improve player performance, and enhance team strategies. To analyze more complex features from game footage, a computer vision model can be used to identify and track key entities from the field. We propose the use of an object detection and tracking system to predict player positioning throughout the game. To translate this to positioning in relation to the field dimensions, we use a point prediction model to identify key points on the field and combine these with known field dimensions to extract actual distances. For the player-identification model, object detection models like YOLO and Faster R-CNN are evaluated on the accuracy of our custom video footage using multiple different evaluation metrics. The goal is to identify the best model for object identification to obtain the most accurate results when paired with SAM2 (Segment Anything Model 2) for segmentation and tracking. For the key point detection model, we use a CNN model to find consistent locations in the soccer field. Through homography, the positions of points and objects in the camera perspective will be transformed to a real-ground perspective. The segmented player masks from SAM2 are transformed from camera perspective to real-world field coordinates through homography, regardless of camera angle or movement. The transformed real-world coordinates can be used to calculate valuable tactical insights including player speed, distance covered, positioning heatmaps, and more complex team statistics, providing coaches and players with actionable performance data previously unavailable from standard video analysis.

Performance optimization of AI infrastructure is key to the fast adoption of large language models (LLMs). The PyTorch compiler (torch.compile), a core optimization tool for deep learning (DL) models (including LLMs), has received due attention. However, torch.compile is prone to correctness bugs, which cause incorrect outputs of compiled DL models without triggering exceptions, crashes, or warnings. These bugs pose a serious threat to the reliability of downstream LLM applications. Data from the PyTorch community shows that 19.2% of high-priority issues are incorrect outputs of compiled DL models induced by torch.compile bugs, the second-most-common bug category (only behind program crashes at 19.57%). However, no systematic study has been conducted to specifically characterize and thereby detect these bugs. In this paper, we present the first empirical study of the correctness bugs in torch.compile, examine their characteristics, and assess the effectiveness of existing fuzzers in detecting them. Based on our findings, we propose a proof-of-concept testing technique named AlignGuard, tailored specifically for detecting correctness bugs in torch.compile. AlignGuard incorporates bug characteristics distilled from our empirical study, applying LLM-based test mutation to existing test cases for correctness bug detection. At the time of writing, AlignGuard has successfully detected 23 new correctness bugs in recent torch.compile. All these bugs have been confirmed or fixed by the PyTorch development team, and over half (14/23) of them are even marked as high-priority bugs, underscoring the usefulness of our technique.

Recent years have seen remarkable progress in autonomous driving, yet generalization to long-tail and open-world scenarios remains a major bottleneck for large-scale deployment. To address this challenge, some works use LLMs and VLMs for vision-language understanding and reasoning, enabling vehicles to interpret rare and safety-critical situations when generating actions. Others study generative world models to capture the spatio-temporal evolution of driving scenes, allowing agents to imagine possible futures before acting. Inspired by human intelligence, which unifies understanding and imagination, we explore a unified model for autonomous driving. We present LMGenDrive, the first framework that combines LLM-based multimodal understanding with generative world models for end-to-end closed-loop driving. Given multi-view camera inputs and natural-language instructions, LMGenDrive generates both future driving videos and control signals. This design provides complementary benefits: video prediction improves spatio-temporal scene modeling, while the LLM contributes strong semantic priors and instruction grounding from large-scale pretraining. We further propose a progressive three-stage training strategy, from vision pretraining to multi-step long-horizon driving, to improve stability and performance. LMGenDrive supports both low-latency online planning and autoregressive offline video generation. Experiments show that it significantly outperforms prior methods on challenging closed-loop benchmarks, with clear gains in instruction following, spatio-temporal understanding, and robustness to rare scenarios. These results suggest that unifying multimodal understanding and generation is a promising direction for more generalizable and robust embodied decision-making systems.

Geometric Foundation Models (GFMs) have recently advanced monocular SLAM by providing robust, calibration-free 3D priors. However, deploying these models on dense video streams introduces significant computational redundancy. Current GFM-based SLAM systems typically rely on post hoc keyframe selection. Because of this, they must perform expensive dense geometric decoding simply to determine whether a frame contains novel geometry, resulting in late rejection and wasted computation. To mitigate this inefficiency, we propose LeanGate, a lightweight feed-forward frame-gating network. LeanGate predicts a geometric utility score to assess a frame's mapping value prior to the heavy GFM feature extraction and matching stages. As a predictive plug-and-play module, our approach bypasses over 90% of redundant frames. Evaluations on standard SLAM benchmarks demonstrate that LeanGate reduces tracking FLOPs by more than 85% and achieves a 5x end-to-end throughput speedup. Furthermore, it maintains the tracking and mapping accuracy of dense baselines.

Code reuse is a widespread practice across software development projects, suggesting an inherent trust in the reused code. Yet, there is a lack of a fundamental understanding of developers' trust and how various factors mold their trust-based cognitive processes. Drawing from the psychology of compliance and trust, we present the results of the first controlled experiment (n=37) which uses eye tracking to explore how urgency (represented by code priority level) and reputation (represented by the experience level of the code's author) influence developers' perceptions of code trustworthiness. Our research revealed that the priority assigned to a code patch significantly influenced developers' code review behavior, impacting their evaluation time, cognitive load, and perceived quality. However, the decision to incorporate and implement the code was not affected . Eye tracking data revealed that there were variations in overall visual code scanning and the distribution of attention across identical code patches labeled as written by senior vs. junior developers. Yet, there were no significant performance differences. Moreover, our participants nominate code functionality, quality, and comprehensibility as primary factors in code evaluation. Despite noticeable changes in code review behavior, our participants surprisingly overlooked the substantial influence of urgency and reputation on their decisions to review and reuse code changes. This study takes the next step toward a better understanding of trust in software engineering and may inform future research about code review platforms and guidelines, code reuse, and automated code generation.

The generation of planning domains from natural language descriptions remains an open problem even with the advent of large language models and reasoning models. Recent work suggests that while LLMs have the ability to assist with domain generation, they are still far from producing high quality domains that can be deployed in practice. To this end, we investigate the ability of an agentic language model feedback framework to generate planning domains from natural language descriptions that have been augmented with a minimal amount of symbolic information. In particular, we evaluate the quality of the generated domains under various forms of symbolic feedback, including landmarks, and output from the VAL plan validator. Using these feedback mechanisms, we experiment using heuristic search over model space to optimize domain quality.

The disintegration of liquid sheets into ligaments and droplets involves highly transient, multi-scale dynamics that are difficult to quantify from high-speed shadowgraphy images. Identifying droplets, ligaments, and blobs formed during breakup, along with tracking across frames, is essential for spray analysis. However, conventional multi-object tracking frameworks impose strict one-to-one temporal associations and cannot represent one-to-many fragmentation events. In this study, we present a two-stage deep learning framework for object detection and temporal relationship modeling across frames. The framework captures ligament deformation, fragmentation, and parent-child lineage during liquid sheet disintegration. In the first stage, a Faster R-CNN with a ResNet-50 backbone and Feature Pyramid Network detects and classifies ligaments and droplets in high-speed shadowgraphy recordings of an impinging Carbopol gel jet. A morphology-preserving synthetic data generation strategy augments the training set without introducing physically implausible configurations, achieving a held-out F1 score of up to 0.872 across fourteen original-to-synthetic configurations. In the second stage, a Transformer-augmented multilayer perceptron classifies inter-frame associations into continuation, fragmentation (one-to-many), and non-association using physics-informed geometric features. Despite severe class imbalance, the model achieves 86.1% accuracy, 93.2% precision, and perfect recall (1.00) for fragmentation events. Together, the framework enables automated reconstruction of fragmentation trees, preservation of parent-child lineage, and extraction of breakup statistics such as fragment multiplicity and droplet size distributions. By explicitly identifying children droplets formed from ligament fragmentation, the framework provides automated analysis of the primary atomization mode.

While Large Language Model-based Multi-Agent Systems (MAS) consistently outperform single-agent systems on complex tasks, their intricate interactions introduce critical reliability challenges arising from communication dynamics and role dependencies. Existing Uncertainty Quantification methods, typically designed for single-turn outputs, fail to address the unique complexities of the MAS. Specifically, these methods struggle with three distinct challenges: the cascading uncertainty in multi-step reasoning, the variability of inter-agent communication paths, and the diversity of communication topologies. To bridge this gap, we introduce MATU, a novel framework that quantifies uncertainty through tensor decomposition. MATU moves beyond analyzing final text outputs by representing entire reasoning trajectories as embedding matrices and organizing multiple execution runs into a higher-order tensor. By applying tensor decomposition, we disentangle and quantify distinct sources of uncertainty, offering a comprehensive reliability measure that is generalizable across different agent structures. We provide comprehensive experiments to show that MATU effectively estimates holistic and robust uncertainty across diverse tasks and communication topologies.

Monadic second order logic (MSO2) plays an important role in parameterized complexity due to the Courcelle's theorem. This theorem states that the problem of checking if a given graph has a property specified by a given MSO2 formula can be solved by a parameterized linear time algorithm with respect to the treewidth of the graph and the size of the formula. We extend this result by showing that models of MSO2 formula with free variables can be represented with a decision diagram whose size is parameterized linear in the above mentioned parameter. In particular, we show a parameterized linear upper bound on the size of a sentential decision diagram (SDD) when treewidth is considered and a parameterized linear upper bound on the size of an ordered binary decision diagram (OBDD) when considering the pathwidth in the parameter. In addition, building on a lower bound on the size of OBDD by Razgon (2014), we show that there is an MSO2 formula and a class of graphs with bounded treewidth which do not admit an OBDD with the size parameterized by the treewidth. Our result offers a new perspective on the Courcelle's theorem and connects it to the area of knowledge representation.

While Experience Replay - the practice of storing rollouts and reusing them multiple times during training - is a foundational technique in general RL, it remains largely unexplored in LLM post-training due to the prevailing belief that fresh, on-policy data is essential for high performance. In this work, we challenge this assumption. We present a systematic study of replay buffers for LLM post-training, formalizing the optimal design as a trade-off between staleness-induced variance, sample diversity and the high computational cost of generation. We show that strict on-policy sampling is suboptimal when generation is expensive. Empirically, we show that a well-designed replay buffer can drastically reduce inference compute without degrading - and in some cases even improving - final model performance, while preserving policy entropy.

Adiabatic Quantum-Flux-Parametron (AQFP) logic is an ultra-low-power superconducting logic family with energy consumption approaching the Shannon limit, making it attractive for quantum computing control and cryogenic computing systems. Traditional AQFP designs face significant physical design challenges due to strict gate-level clocking requirements and limited interconnect lengths, leading to substantial buffer overhead and difficult timing closure. Recently, delay-line clocking of AQFP has been proposed to improve timing margins and reduce latency by enabling more flexible clock scheduling. However, prior work has primarily focused on placement and latency minimization, while relying on fixed timing parameters that do not capture the frequency dependence of AQFP setup and hold constraints. To address this limitation, we propose a frequency-aware post-routing optimization framework that jointly optimizes clock period, latency, and timing slack under user-specified weighting. Experimental results across common benchmarks achieve 100% post-routing timing closure across a range of performance--latency--slack trade-offs. Our approach also automates phase-skipping, reducing path-balancing buffer insertion by 34% on average while only reducing operating frequency by 4%.

QoS-QoE translation is a fundamental problem in multimedia systems because it characterizes how measurable system and network conditions affect user-perceived experience. Although many prior studies have examined this relationship, their findings are often developed for specific setups and remain scattered across papers, experimental settings, and reporting formats, limiting systematic reuse, cross-scenario generalization, and large-scale analysis. To address this gap, we first introduce QoS-QoE Translation dataset, a source-grounded dataset of structured QoS-QoE relationships from the multimedia literature, with a focus on video streaming related tasks. We construct the dataset through an automated pipeline that combines paper curation, QoS-QoE relationship extraction, and iterative data evaluation. Each record preserves the extracted relationship together with parameter definitions, supporting evidence, and contextual metadata. We further evaluate the capability of large language models (LLMs) on QoS-QoE translation, both before and after supervised fine-tuning on our dataset, and show strong performance on both continuous-value and discrete-label prediction in bidirectional translation, from QoS-QoE and QoE-QoS. Our dataset provides a foundation for benchmarking LLMs in QoS-QoE translation and for supporting future LLM-based reasoning for multimedia quality prediction and optimization. The complete dataset and code are publicly available at https://yyu6969.github.io/qos-qoe-translation-page/, for full reproducibility and open access.

We introduce Unified Multimodal Uncertain Inference (UMUI), a multimodal inference task spanning text, audio, and video, where models must produce calibrated probability estimates of hypotheses conditioned on a premise in any modality or combination. While uncertain inference has been explored in text, extension to other modalities has been limited to single-modality binary entailment judgments, leaving no framework for fine-grained probabilistic reasoning in or across other modalities. To address this, we curate a human-annotated evaluation set with scalar probability judgments across audio, visual, and audiovisual settings, and additionally evaluate on existing text and audio benchmarks. We introduce CLUE (Calibrated Latent Uncertainty Estimation), which combines self-consistent teacher calibration and distribution-based confidence probing to produce calibrated predictions. We demonstrate that our 3B-parameter model achieves equivalent or stronger performance than baselines up to 32B parameters across all modalities.

Tokens serve as the basic units of representation in DNA language models (DNALMs), yet their design remains underexplored. Unlike natural language, DNA lacks inherent token boundaries or predefined compositional rules, making tokenization a fundamental modeling decision rather than a naturally specified one. While existing approaches like byte-pair encoding (BPE) excel at capturing token structures that reflect human-generated linguistic regularities, DNA is organized by biological function and evolutionary constraint rather than linguistic convention. We argue that DNA tokenization should prioritize functional sequence patterns like regulatory motifs-short, recurring segments under evolutionary constraint and typically preserved across species. We incorporate evolutionary information directly into the tokenization process through EvoLen, a tokenizer that combines evolutionary stratification with length-aware decoding to better preserve motif-scale functional sequence units. EvoLen uses cross-species evolutionary signals to group DNA sequences, trains separate BPE tokenizers on each group, merges the resulting vocabularies via a rule prioritizing preserved patterns, and applies length-aware decoding with dynamic programming. Through controlled experiments, EvoLen improves the preservation of functional sequence patterns, differentiation across genomic contexts, and alignment with evolutionary constraint, while matching or outperforming standard BPE across diverse DNALM benchmarks. These results demonstrate that tokenization introduces a critical inductive bias and that incorporating evolutionary information yields more biologically meaningful and interpretable sequence representations.

How do you build a sign language recognizer that works on a phone? That question drove this work. We built EfficientSign, a lightweight model which takes EfficientNet-B0 and focuses on two attention modules (Squeeze-and-Excitation for channel focus, and a spatial attention layer that focuses on the hand gestures). We tested it against five other approaches on 12,637 images of Indian Sign Language alphabets, all 26 classes, using 5-fold cross-validation. EfficientSign achieves the accuracy of 99.94% (+/-0.05%), which matches the performance of ResNet18's 99.97% accuracy, but with 62% fewer parameters (4.2M vs 11.2M). We also experimented with feeding deep features (1,280-dimensional vectors pulled from EfficientNet-B0's pooling layer) into classical classifiers. SVM achieved the accuracy of 99.63%, Logistic Regression achieved the accuracy of 99.03% and KNN achieved accuracy of 96.33%. All of these blow past the 92% that SURF-based methods managed on a similar dataset back in 2015. Our results show that attention-enhanced learning model provides an efficient and deployable solution for ISL recognition without requiring a massive model or hand-tuned feature pipelines anymore.

A quantum network's purpose is to enable users to execute applications on end nodes. This requires the network to provide the service of creating entangled links between those nodes. Users of mature networks, such as the internet or the telephone network expect accepted service demands to be met reliably. We first define reliability requirements that extend classical computer network concepts to quantum network service delivery. We then introduce Arqon, a suite of control applications designed to deliver reliable service in centrally controlled quantum networks. We demonstrate through both analytic and numerical evaluation that Arqon satisfies all reliability requirements for accepted demands. These evaluations consider static network topologies. We provide a complete Python implementation and perform complexity analysis showing that admission control scales as $O(k^3)$ in the number of incoming demands $k$ and schedule computation scales as ${O(N^3)}$ in the number of accepted demands to schedule $N$.

Group Relative Policy Optimization (GRPO) has proven effective in RLVR by using outcome-based rewards. While fine-grained dense rewards can theoretically improve performance, we reveal that under practical sampling budgets, Monte Carlo estimation yields high-variance and sign-inconsistent advantages for early reasoning tokens, paradoxically underperforming outcome-only GRPO. We propose Skip-Connected Optimization (SKPO), which decomposes reasoning into upstream and downstream phases: upstream receives dense rewards from downstream Monte Carlo sampling with single-stream optimization; downstream maintains group-relative optimization, where a skip connection concatenates the upstream segment with the original problem, enabling the model to leverage helpful upstream reasoning while preserving the freedom to bypass flawed reasoning through direct problem access. Experiments demonstrate improvements of 3.91% and 6.17% relative gains over the strongest baselines on Qwen2.5-Math-7B and Llama-3.2-3B respectively across mathematical benchmarks and out-of-domain tasks including general reasoning and code generation. Further analysis reveals an implicit advantage: SKPO generates trajectories with higher intermediate-step quality even when matched for final correctness.

Automated planning algorithms require an action model specifying the preconditions and effects of each action, but obtaining such a model is often hard. Learning action models from observations is feasible, but existing algorithms for numeric domains are offline, requiring expert traces as input. We propose the Reinforcement learning, Action Model learning, and Planning (RAMP) strategy for learning numeric planning action models online via interactions with the environment. RAMP simultaneously trains a Deep Reinforcement Learning (DRL) policy, learns a numeric action model from past interactions, and uses that model to plan future actions when possible. These components form a positive feedback loop: the RL policy gathers data to refine the action model, while the planner generates plans to continue training the RL policy. To facilitate this integration of RL and numeric planning, we developed Numeric PDDLGym, an automated framework for converting numeric planning problems to Gym environments. Experimental results on standard IPC numeric domains show that RAMP significantly outperforms PPO, a well-known DRL algorithm, in terms of solvability and plan quality.

It is widely believed that tens of thousands of physical qubits are needed to build a practically useful quantum computer. Atom arrays formed by optical tweezers are among the most promising platforms for achieving this goal, owing to the excellent scalability and mobility of atomic qubits. However, assembling a defect-free atom array with ~ 10^4 qubits remains algorithmically challenging, alongside other hardware limitations. This is due to the computationally hard path-planning problems and the time-consuming generation of suffciently smooth trajectories for optical tweezer potentials by spatial light modulators (SLM). Here, we present a unified framework comprising two innovative components to fully address these algorithmic challenges: (1) a path-planning module that employs a supervised learning approach using a graph neural network combined with a modified auction decoder, and (2) a potential-generation module called the phase and profile-aware Weighted Gerchberg-Saxton algorithm. The inference time for the first module is nearly a size-independent constant overhead of ~ 5 ms, and the second module generates a potential frame with about 0.5 ms, a timescale shorter than the current commercial SLM refresh time. Altogether, our algorithm enables the assembly of an atom array with 10^4 qubits on a timescale much shorter than the typical vacuum lifetime of the trapped atoms.

Neural Quantum States based on autoregressive recurrent neural network (RNN) wave functions enable efficient sampling without Markov-chain autocorrelation, but standard RNN architectures are biased toward finite-length correlations and can fail on states with long-range dependencies. A common response is to adopt transformer-style self-attention, but this typically comes with substantially higher computational and memory overhead. Here we introduce dilated RNN wave functions, where recurrent units access distant sites through dilated connections, injecting an explicit long-range inductive bias while retaining a favorable $\mathcal{O}(N \log N)$ forward pass scaling. We show analytically that dilation changes the correlation geometry and can induce power-law correlation scaling in a simplified linearized and perturbative setting. Numerically, for the critical 1D transverse-field Ising model, dilated RNNs reproduce the expected power-law connected two-point correlations in contrast to the exponential decay typical of conventional RNN ansätze. We further show that the dilated RNN accurately approximates the one-dimensional Cluster state, a paradigmatic example with long-range conditional correlations that has previously been reported to be challenging for RNN-based wave functions. These results highlight dilation as a simple geometric mechanism for building correlation-aware autoregressive neural quantum states.

Modern financial systems generate vast quantities of transactional and event-level data that encode rich economic signals. This paper presents PRAGMA, a family of foundation models for multi-source banking event sequences. Our approach pre-trains a Transformer-based architecture with masked modelling on a large-scale, heterogeneous banking event corpus using a self-supervised objective tailored to the discrete, variable-length nature of financial records. The resulting model supports a wide range of downstream tasks such as credit scoring, fraud detection, and lifetime value prediction: strong performance can be achieved by training a simple linear model on top of the extracted embeddings and can be further improved with lightweight fine-tuning. Through extensive evaluation on downstream tasks, we demonstrate that PRAGMA achieves superior performance across multiple domains directly from raw event sequences, providing a general-purpose representation layer for financial applications.

We motivate the use of differentiable probabilistic programming techniques in order to account for the large model-space inherent to astrophysical $γ$-ray analyses. Targeting the longstanding Galactic Center $γ$-ray Excess (GCE) puzzle, we construct differentiable forward model and likelihood that make liberal use of GPU acceleration and vectorization in order to simultaneously account for a continuum of possible spatial morphologies consistent with the GCE emission in a fully probabilistic manner. Our setup allows for efficient inference over the large model space using variational methods. Beyond application to $γ$-ray data, a goal of this work is to showcase how differentiable probabilistic programming can be used as a tool to enable flexible analyses of astrophysical datasets.

The advent of agentic multimodal models has empowered systems to actively interact with external environments. However, current agents suffer from a profound meta-cognitive deficit: they struggle to arbitrate between leveraging internal knowledge and querying external utilities. Consequently, they frequently fall prey to blind tool invocation, resorting to reflexive tool execution even when queries are resolvable from the raw visual context. This pathological behavior precipitates severe latency bottlenecks and injects extraneous noise that derails sound reasoning. Existing reinforcement learning protocols attempt to mitigate this via a scalarized reward that penalizes tool usage. Yet, this coupled formulation creates an irreconcilable optimization dilemma: an aggressive penalty suppresses essential tool use, whereas a mild penalty is entirely subsumed by the variance of the accuracy reward during advantage normalization, rendering it impotent against tool overuse. To transcend this bottleneck, we propose HDPO, a framework that reframes tool efficiency from a competing scalar objective to a strictly conditional one. By eschewing reward scalarization, HDPO maintains two orthogonal optimization channels: an accuracy channel that maximizes task correctness, and an efficiency channel that enforces execution economy exclusively within accurate trajectories via conditional advantage estimation. This decoupled architecture naturally induces a cognitive curriculum-compelling the agent to first master task resolution before refining its self-reliance. Extensive evaluations demonstrate that our resulting model, Metis, reduces tool invocations by orders of magnitude while simultaneously elevating reasoning accuracy.

Robotic manipulation with deformable objects represents a data-intensive regime in embodied learning, where shape, contact, and topology co-evolve in ways that far exceed the variability of rigids. Although simulation promises relief from the cost of real-world data acquisition, prevailing sim-to-real pipelines remain rooted in rigid-body abstractions, producing mismatched geometry, fragile soft dynamics, and motion primitives poorly suited for cloth interaction. We posit that simulation fails not for being synthetic, but for being ungrounded. To address this, we introduce SIM1, a physics-aligned real-to-sim-to-real data engine that grounds simulation in the physical world. Given limited demonstrations, the system digitizes scenes into metric-consistent twins, calibrates deformable dynamics through elastic modeling, and expands behaviors via diffusion-based trajectory generation with quality filtering. This pipeline transforms sparse observations into scaled synthetic supervision with near-demonstration fidelity. Experiments show that policies trained on purely synthetic data achieve parity with real-data baselines at a 1:15 equivalence ratio, while delivering 90% zero-shot success and 50% generalization gains in real-world deployment. These results validate physics-aligned simulation as scalable supervision for deformable manipulation and a practical pathway for data-efficient policy learning.

Multimodal Mixture-of-Experts (MoE) models have achieved remarkable performance on vision-language tasks. However, we identify a puzzling phenomenon termed Seeing but Not Thinking: models accurately perceive image content yet fail in subsequent reasoning, while correctly solving identical problems presented as pure text. Through systematic analysis, we first verify that cross-modal semantic sharing exists in MoE architectures, ruling out semantic alignment failure as the sole explanation. We then reveal that visual experts and domain experts exhibit layer-wise separation, with image inputs inducing significant routing divergence from text inputs in middle layers where domain experts concentrate. Based on these findings, we propose the Routing Distraction hypothesis: when processing visual inputs, the routing mechanism fails to adequately activate task-relevant reasoning experts. To validate this hypothesis, we design a routing-guided intervention method that enhances domain expert activation. Experiments on three multimodal MoE models across six benchmarks demonstrate consistent improvements, with gains of up to 3.17% on complex visual reasoning tasks. Our analysis further reveals that domain expert identification locates cognitive functions rather than sample-specific solutions, enabling effective transfer across tasks with different information structures.

Text-to-Audio-Video (T2AV) generation is rapidly becoming a core interface for media creation, yet its evaluation remains fragmented. Existing benchmarks largely assess audio and video in isolation or rely on coarse embedding similarity, failing to capture the fine-grained joint correctness required by realistic prompts. We introduce AVGen-Bench, a task-driven benchmark for T2AV generation featuring high-quality prompts across 11 real-world categories. To support comprehensive assessment, we propose a multi-granular evaluation framework that combines lightweight specialist models with Multimodal Large Language Models (MLLMs), enabling evaluation from perceptual quality to fine-grained semantic controllability. Our evaluation reveals a pronounced gap between strong audio-visual aesthetics and weak semantic reliability, including persistent failures in text rendering, speech coherence, physical reasoning, and a universal breakdown in musical pitch control. Code and benchmark resources are available at http://aka.ms/avgenbench.

Group Relative Policy Optimization (GRPO) has emerged as the de facto Reinforcement Learning (RL) objective driving recent advancements in Multimodal Large Language Models. However, extending this success to open-source multimodal generalist models remains heavily constrained by two primary challenges: the extreme variance in reward topologies across diverse visual tasks, and the inherent difficulty of balancing fine-grained perception with multi-step reasoning capabilities. To address these issues, we introduce Gaussian GRPO (G$^2$RPO), a novel RL training objective that replaces standard linear scaling with non-linear distributional matching. By mathematically forcing the advantage distribution of any given task to strictly converge to a standard normal distribution, $\mathcal{N}(0,1)$, G$^2$RPO theoretically ensures inter-task gradient equity, mitigates vulnerabilities to heavy-tail outliers, and offers symmetric update for positive and negative rewards. Leveraging the enhanced training stability provided by G$^2$RPO, we introduce two task-level shaping mechanisms to seamlessly balance perception and reasoning. First, response length shaping dynamically elicits extended reasoning chains for complex queries while enforce direct outputs to bolster visual grounding. Second, entropy shaping tightly bounds the model's exploration zone, effectively preventing both entropy collapse and entropy explosion. Integrating these methodologies, we present OpenVLThinkerV2, a highly robust, general-purpose multimodal model. Extensive evaluations across 18 diverse benchmarks demonstrate its superior performance over strong open-source and leading proprietary frontier models.

Visual decoding from brain signals is a key challenge at the intersection of computer vision and neuroscience, requiring methods that bridge neural representations and computational models of vision. A field-wide goal is to achieve generalizable, cross-subject models. A major obstacle towards this goal is the substantial variability in neural representations across individuals, which has so far required training bespoke models or fine-tuning separately for each subject. To address this challenge, we introduce a meta-optimized approach for semantic visual decoding from fMRI that generalizes to novel subjects without any fine-tuning. By simply conditioning on a small set of image-brain activation examples from the new individual, our model rapidly infers their unique neural encoding patterns to facilitate robust and efficient visual decoding. Our approach is explicitly optimized for in-context learning of the new subject's encoding model and performs decoding by hierarchical inference, inverting the encoder. First, for multiple brain regions, we estimate the per-voxel visual response encoder parameters by constructing a context over multiple stimuli and responses. Second, we construct a context consisting of encoder parameters and response values over multiple voxels to perform aggregated functional inversion. We demonstrate strong cross-subject and cross-scanner generalization across diverse visual backbones without retraining or fine-tuning. Moreover, our approach requires neither anatomical alignment nor stimulus overlap. This work is a critical step towards a generalizable foundation model for non-invasive brain decoding.

We introduce RewardFlow, an inversion-free framework that steers pretrained diffusion and flow-matching models at inference time through multi-reward Langevin dynamics. RewardFlow unifies complementary differentiable rewards for semantic alignment, perceptual fidelity, localized grounding, object consistency, and human preference, and further introduces a differentiable VQA-based reward that provides fine-grained semantic supervision through language-vision reasoning. To coordinate these heterogeneous objectives, we design a prompt-aware adaptive policy that extracts semantic primitives from the instruction, infers edit intent, and dynamically modulates reward weights and step sizes throughout sampling. Across several image editing and compositional generation benchmarks, RewardFlow delivers state-of-the-art edit fidelity and compositional alignment.

Personal AI tools can now be generated from natural-language requests, but they often remain isolated after creation. We present PSI, a shared-state architecture that turns independently generated modules into coherent instruments: persistent, connected, and chat-complementary artifacts accessible through both GUIs and a generic chat agent. By publishing current state and write-back affordances to a shared personal-context bus, modules enable cross-module reasoning and synchronized actions across interfaces. We study PSI through a three-week autobiographical deployment in a self-developed personal AI environment and show that later-generated instruments can be integrated automatically through the same contract. PSI identifies shared state as the missing systems layer that transforms AI-generated personal software from isolated apps into coherent personal computing environments.

On-policy distillation (OPD) trains student models under their own induced distribution while leveraging supervision from stronger teachers. We identify a failure mode of OPD: as training progresses, on-policy rollouts can undergo abrupt length inflation, causing truncated trajectories to dominate the training data. This truncation collapse coincides with abrupt repetition saturation and induces biased gradient signals, leading to severe training instability and sharp degradation in validation performance. We attribute this problem to the interaction between student-induced data collection and the distillation objective, which implicitly favors long and repetitive rollouts. To address this issue, we propose StableOPD, a stabilized OPD framework that combines a reference-based divergence constraint with rollout mixture distillation. These together mitigate repetition-induced length inflation and further stabilize OPD training. Across multiple math reasoning datasets, our approach prevents truncation collapse, stabilizes training dynamics, and improves performance by 7.2% on average.

Today's large language models (LLMs) are trained to align with user preferences through methods such as reinforcement learning. Yet models are beginning to be deployed not merely to satisfy users, but also to generate revenue for the companies that created them through advertisements. This creates the potential for LLMs to face conflicts of interest, where the most beneficial response to a user may not be aligned with the company's incentives. For instance, a sponsored product may be more expensive but otherwise equal to another; in this case, what does (and should) the LLM recommend to the user? In this paper, we provide a framework for categorizing the ways in which conflicting incentives might lead LLMs to change the way they interact with users, inspired by literature from linguistics and advertising regulation. We then present a suite of evaluations to examine how current models handle these tradeoffs. We find that a majority of LLMs forsake user welfare for company incentives in a multitude of conflict of interest situations, including recommending a sponsored product almost twice as expensive (Grok 4.1 Fast, 83%), surfacing sponsored options to disrupt the purchasing process (GPT 5.1, 94%), and concealing prices in unfavorable comparisons (Qwen 3 Next, 24%). Behaviors also vary strongly with levels of reasoning and users' inferred socio-economic status. Our results highlight some of the hidden risks to users that can emerge when companies begin to subtly incentivize advertisements in chatbots.

Large multimodal models are increasingly used as the reasoning core of embodied agents operating in 3D environments, yet they remain prone to hallucinations that can produce unsafe and ungrounded decisions. Existing inference-time hallucination mitigation methods largely target 2D vision-language settings and do not transfer to embodied 3D reasoning, where failures arise from object presence, spatial layout, and geometric grounding rather than pixel-level inconsistencies. We introduce 3D-VCD, the first inference-time visual contrastive decoding framework for hallucination mitigation in 3D embodied agents. 3D-VCD constructs a distorted 3D scene graph by applying semantic and geometric perturbations to object-centric representations, such as category substitutions and coordinate or extent corruption. By contrasting predictions under the original and distorted 3D contexts, our method suppresses tokens that are insensitive to grounded scene evidence and are therefore likely driven by language priors. We evaluate 3D-VCD on the 3D-POPE and HEAL benchmarks and show that it consistently improves grounded reasoning without any retraining, establishing inference-time contrastive decoding over structured 3D representations as an effective and practical route to more reliable embodied intelligence.

Applying steering vectors to large language models (LLMs) is an efficient and effective model alignment technique, but we lack an interpretable explanation for how it works-- specifically, what internal mechanisms steering vectors affect and how this results in different model outputs. To investigate the causal mechanisms underlying the effectiveness of steering vectors, we conduct a comprehensive case study on refusal. We propose a multi-token activation patching framework and discover that different steering methodologies leverage functionally interchangeable circuits when applied at the same layer. These circuits reveal that steering vectors primarily interact with the attention mechanism through the OV circuit while largely ignoring the QK circuit-- freezing all attention scores during steering drops performance by only 8.75% across two model families. A mathematical decomposition of the steered OV circuit further reveals semantically interpretable concepts, even in cases where the steering vector itself does not. Leveraging the activation patching results, we show that steering vectors can be sparsified by up to 90-99% while retaining most performance, and that different steering methodologies agree on a subset of important dimensions.

AI agents may be able to automate your inbox, but can they automate other routine aspects of your life? Everyday online tasks offer a realistic yet unsolved testbed for evaluating the next generation of AI agents. To this end, we introduce ClawBench, an evaluation framework of 153 simple tasks that people need to accomplish regularly in their lives and work, spanning 144 live platforms across 15 categories, from completing purchases and booking appointments to submitting job applications. These tasks require demanding capabilities beyond existing benchmarks, such as obtaining relevant information from user-provided documents, navigating multi-step workflows across diverse platforms, and write-heavy operations like filling in many detailed forms correctly. Unlike existing benchmarks that evaluate agents in offline sandboxes with static pages, ClawBench operates on production websites, preserving the full complexity, dynamic nature, and challenges of real-world web interaction. A lightweight interception layer captures and blocks only the final submission request, ensuring safe evaluation without real-world side effects. Our evaluations of 7 frontier models show that both proprietary and open-source models can complete only a small portion of these tasks. For example, Claude Sonnet 4.6 achieves only 33.3%. Progress on ClawBench brings us closer to AI agents that can function as reliable general-purpose assistants.

Large language models (LLMs) can struggle to memorize factual knowledge in their parameters, often leading to hallucinations and poor performance on knowledge-intensive tasks. In this paper, we formalize fact memorization from an information-theoretic perspective and study how training data distributions affect fact accuracy. We show that fact accuracy is suboptimal (below the capacity limit) whenever the amount of information contained in the training data facts exceeds model capacity. This is further exacerbated when the fact frequency distribution is skewed (e.g. a power law). We propose data selection schemes based on the training loss alone that aim to limit the number of facts in the training data and flatten their frequency distribution. On semi-synthetic datasets containing high-entropy facts, our selection method effectively boosts fact accuracy to the capacity limit. When pretraining language models from scratch on an annotated Wikipedia corpus, our selection method enables a GPT2-Small model (110m parameters) to memorize 1.3X more entity facts compared to standard training, matching the performance of a 10X larger model (1.3B parameters) pretrained on the full dataset.

This technical report introduces EXAONE 4.5, the first open-weight vision language model released by LG AI Research. EXAONE 4.5 is architected by integrating a dedicated visual encoder into the existing EXAONE 4.0 framework, enabling native multimodal pretraining over both visual and textual modalities. The model is trained on large-scale data with careful curation, particularly emphasizing document-centric corpora that align with LG's strategic application domains. This targeted data design enables substantial performance gains in document understanding and related tasks, while also delivering broad improvements across general language capabilities. EXAONE 4.5 extends context length up to 256K tokens, facilitating long-context reasoning and enterprise-scale use cases. Comparative evaluations demonstrate that EXAONE 4.5 achieves competitive performance in general benchmarks while outperforming state-of-the-art models of similar scale in document understanding and Korean contextual reasoning. As part of LG's ongoing effort toward practical industrial deployment, EXAONE 4.5 is designed to be continuously extended with additional domains and application scenarios to advance AI for a better life.

Large language models (LLMs) can perform remarkably complex tasks, yet the fine-grained details of how these capabilities emerge during pretraining remain poorly understood. Scaling laws on validation loss tell us how much a model improves with additional compute, but not what skills it acquires in which order. To remedy this, we propose the Implicit Curriculum Hypothesis: pretraining follows a compositional and predictable curriculum across models and data mixtures. We test this by designing a suite of simple, composable tasks spanning retrieval, morphological transformations, coreference, logical reasoning, and mathematics. Using these tasks, we track emergence points across four model families spanning sizes from 410M-13B parameters. We find that emergence orderings of when models reach fixed accuracy thresholds are strikingly consistent ($ρ= .81$ across 45 model pairs), and that composite tasks most often emerge after their component tasks. Furthermore, we find that this structure is encoded in model representations: tasks with similar function vector representations also tend to follow similar trajectories in training. By using the space of representations derived from our task set, we can effectively predict the training trajectories of simple held-out compositional tasks throughout the course of pretraining ($R^2 = .68$-$.84$ across models) without previously evaluating them. Together, these results suggest that pretraining is more structured than loss curves reveal: skills emerge in a compositional order that is consistent across models and readable from their internals.

We initiate the study of language generation in the limit, a model recently introduced by Kleinberg and Mullainathan [KM24], under the constraint of differential privacy. We consider the continual release model, where a generator must eventually output a stream of valid strings while protecting the privacy of the entire input sequence. Our first main result is that for countable collections of languages, privacy comes at no qualitative cost: we provide an $\varepsilon$-differentially-private algorithm that generates in the limit from any countable collection. This stands in contrast to many learning settings where privacy renders learnability impossible. However, privacy does impose a quantitative cost: there are finite collections of size $k$ for which uniform private generation requires $Ω(k/\varepsilon)$ samples, whereas just one sample suffices non-privately. We then turn to the harder problem of language identification in the limit. Here, we show that privacy creates fundamental barriers. We prove that no $\varepsilon$-DP algorithm can identify a collection containing two languages with an infinite intersection and a finite set difference, a condition far stronger than the classical non-private characterization of identification. Next, we turn to the stochastic setting where the sample strings are sampled i.i.d. from a distribution (instead of being generated by an adversary). Here, we show that private identification is possible if and only if the collection is identifiable in the adversarial model. Together, our results establish new dimensions along which generation and identification differ and, for identification, a separation between adversarial and stochastic settings induced by privacy constraints.

Class Activation Mapping (CAM) methods are widely used to generate visual explanations for deep learning classifiers in medical imaging. However, existing evaluation frameworks assess whether explanations are correct, measured by localisation fidelity against radiologist annotations, rather than whether they are consistent: whether the model applies the same spatial reasoning strategy across different patients with the same pathology. We propose the C-Score (Consistency Score), a confidence-weighted, annotation-free metric that quantifies intra-class explanation reproducibility via intensity-emphasised pairwise soft IoU across correctly classified instances. We evaluate six CAM techniques: GradCAM, GradCAM++, LayerCAM, EigenCAM, ScoreCAM, and MS GradCAM++ across three CNN architectures (DenseNet201, InceptionV3, ResNet50V2) over thirty training epochs on the Kermany chest X-ray dataset, covering transfer learning and fine-tuning phases. We identify three distinct mechanisms of AUC-consistency dissociation, invisible to standard classification metrics: threshold-mediated gold list collapse, technique-specific attribution collapse at peak AUC, and class-level consistency masking in global aggregation. C-Score provides an early warning signal of impending model instability. ScoreCAM deterioration on ResNet50V2 is detectable one full checkpoint before catastrophic AUC collapse and yields architecture-specific clinical deployment recommendations grounded in explanation quality rather than predictive ranking alone.

Science currently offers two options for quality assurance, both inadequate. Journal gatekeeping claims to verify both integrity and contribution, but actually measures prestige: peer review is slow, biased, and misses fabricated citations even at top venues. Open science provides no quality assurance at all: the only filter between AI-generated text and the public record is the author's integrity. AI-assisted writing makes both worse by producing more papers faster than either system can absorb. We propose a third option: measure the paper itself. sciwrite-lint (pip install sciwrite-lint) is an open-source linter for scientific manuscripts that runs entirely on the researcher's machine (free public databases, a single consumer GPU, and open-weights models) with no manuscripts sent to external services. The pipeline verifies that references exist, checks retraction status, compares metadata against canonical records, downloads and parses cited papers, verifies that they support the claims made about them, and follows one level further to check cited papers' own bibliographies. Each reference receives a per-reference reliability score aggregating all verification signals. We evaluate the pipeline on 30 unseen papers from arXiv and bioRxiv with error injection and LLM-adjudicated false positive analysis. As an experimental extension, we propose SciLint Score, combining integrity verification with a contribution component that operationalizes five frameworks from philosophy of science (Popper, Lakatos, Kitcher, Laudan, Mayo) into computable structural properties of scientific arguments. The integrity component is the core of the tool and is evaluated in this paper; the contribution component is released as experimental code for community development.

User interactions in online recommendation platforms create interdependencies among content creators: feedback on one creator's content influences the system's learning and, in turn, the exposure of other creators' contents. To analyze incentives in such settings, we model collaboration as a multi-agent stochastic linear bandit problem with a transferable utility (TU) cooperative game formulation, where a coalition's value equals the negative sum of its members' cumulative regrets. We show that, for identical (homogenous) agents with fixed action sets, the induced TU game is convex under mild algorithmic conditions, implying a non-empty core that contains the Shapley value and ensures both stability and fairness. For heterogeneous agents, the game still admits a non-empty core, though convexity and Shapley value core-membership are no longer guaranteed. To address this, we propose a simple regret-based payout rule that satisfies three out of the four Shapley axioms and also lies in the core. Experiments on MovieLens-100k dataset illustrate when the empirical payout aligns with -- and diverges from -- the Shapley fairness across different settings and algorithms.

Prompt injection attacks pose serious security risks across a wide range of real-world applications. While receiving increasing attention, the community faces a critical gap: the lack of a unified platform for prompt injection evaluation. This makes it challenging to reliably compare defenses, understand their true robustness under diverse attacks, or assess how well they generalize across tasks and benchmarks. For instance, many defenses initially reported as effective were later found to exhibit limited robustness on diverse datasets and attacks. To bridge this gap, we introduce PIArena, a unified and extensible platform for prompt injection evaluation that enables users to easily integrate state-of-the-art attacks and defenses and evaluate them across a variety of existing and new benchmarks. We also design a dynamic strategy-based attack that adaptively optimizes injected prompts based on defense feedback. Through comprehensive evaluation using PIArena, we uncover critical limitations of state-of-the-art defenses: limited generalizability across tasks, vulnerability to adaptive attacks, and fundamental challenges when an injected task aligns with the target task. The code and datasets are available at https://github.com/sleeepeer/PIArena.

This paper addresses the decentralized non-uniform area coverage problem for multi-agent systems, a critical task in missions with high spatial priority and resource constraints. While existing density-based methods often rely on computationally heavy Eulerian PDE solvers or heuristic planning, we propose Stochastic Density-Driven Optimal Control (D$^2$OC). This is a rigorous Lagrangian framework that bridges the gap between individual agent dynamics and collective distribution matching. By formulating a stochastic MPC-like problem that minimizes the Wasserstein distance as a running cost, our approach ensures that the time-averaged empirical distribution converges to a non-parametric target density under stochastic LTI dynamics. A key contribution is the formal convergence guarantee established via reachability analysis, providing a bounded tracking error even in the presence of process and measurement noise. Numerical results verify that Stochastic D$^2$OC achieves robust, decentralized coverage while outperforming previous heuristic methods in optimality and consistency.

Scanpath similarity metrics are central to eye-movement research, yet existing methods predominantly evaluate spatial and temporal alignment while neglecting semantic equivalence between attended image regions. We present a semantic scanpath similarity framework that integrates vision-language models (VLMs) into eye-tracking analysis. Each fixation is encoded under controlled visual context (patch-based and marker-based strategies) and transformed into concise textual descriptions, which are aggregated into scanpath-level representations. Semantic similarity is then computed using embedding-based and lexical NLP metrics and compared against established spatial measures, including MultiMatch and DTW. Experiments on free-viewing eye-tracking data demonstrate that semantic similarity captures partially independent variance from geometric alignment, revealing cases of high content agreement despite spatial divergence. We further analyze the impact of contextual encoding on description fidelity and metric stability. Our findings suggest that multimodal foundation models enable interpretable, content-aware extensions of classical scanpath analysis, providing a complementary dimension for gaze research within the ETRA community.

Previous work has established that neural network-based node embeddings return different outcomes when trained with identical parameters on the same dataset, just from using different training seeds. Yet, it has not been thoroughly analyzed how key hyperparameters such as embedding dimension could impact this instability. In this work, we investigate how varying the dimensionality of node embeddings influences both their stability and downstream performance. We systematically evaluate five widely used methods -- ASNE, DGI, GraphSAGE, node2vec, and VERSE -- across multiple datasets and embedding dimensions. We assess stability from both a representational perspective and a functional perspective, alongside performance evaluation. Our results show that embedding stability varies significantly with dimensionality, but we observe different patterns across the methods we consider: while some approaches, such as node2vec and ASNE, tend to become more stable with higher dimensionality, other methods do not exhibit the same trend. Moreover, we find that maximum stability does not necessarily align with optimal task performance. These findings highlight the importance of carefully selecting embedding dimension, and provide new insights into the trade-offs between stability, performance, and computational effectiveness in graph representation learning.

Interpretation is essential to deciphering the language of art: audiences communicate with artists by recovering meaning from visual artifacts. However, current Generative Art (GenArt) evaluators remain fixated on surface-level image quality or literal prompt adherence, failing to assess the deeper symbolic or abstract meaning intended by the creator. We address this gap by formalizing a Peircean computational semiotic theory that models Human-GenArt Interaction (HGI) as cascaded semiosis. This framework reveals that artistic meaning is conveyed through three modes - iconic, symbolic, and indexical - yet existing evaluators operate heavily within the iconic mode, remaining structurally blind to the latter two. To overcome this structural blindness, we propose SemJudge. This evaluator explicitly assesses symbolic and indexical meaning in HGI via a Hierarchical Semiosis Graph (HSG) that reconstructs the meaning-making process from prompt to generated artifact. Extensive quantitative experiments show that SemJudge aligns more closely with human judgments than prior evaluators on an interpretation-intensive fine-art benchmark. User studies further demonstrate that SemJudge produces deeper, more insightful artistic interpretations, thereby paving the way for GenArt to move beyond the generation of "pretty" images toward a medium capable of expressing complex human experience. Project page: https://github.com/songrise/SemJudge.

The purpose of this paper is two-fold. First we show that Kim's building-up construction of binary self-dual codes is equivalent to Chinburg-Zhang's Hilbert symbol construction. Second we introduce a $q$-ary version of Chinburg-Zhang's construction in order to construct $q$-ary self-dual codes efficiently. For the latter, we study self-dual codes over split finite fields \(\F_q\) with \(q \equiv 1 \pmod{4}\) through three complementary viewpoints: the building-up construction, the binary arithmetic reduction of Chinburg--Zhang, and the hyperbolic geometry of the Euclidean plane. The condition that \(-1\) be a square is the common algebraic input linking these viewpoints: in the binary case it underlies the Lagrangian reduction picture, while in the split \(q\)-ary case it produces the isotropic line governing the correction terms in the extension formulas. As an application of our efficient form of generator matrices, we construct optimal self-dual codes from the split boxed construction, including self-dual \([6,3,4]\) and \([8,4,4]\) codes over \(\GF{5}\), MDS self-dual \([8,4,5]\) and \([10,5,6]\) codes over \(\GF{13}\), and a self-dual \([12,6,6]\) code over \(\GF{13}\). These structural statements are accompanied by a Lean~4 formalization of the algebraic core.

Recent research shows that greater numbers of people are turning to Large Language Models (LLMs) for emotional support, and that people rate LLM responses as more empathic than human-written responses. We suggest a reason for this success: LLMs have learned and consistently deploy a well-liked template for expressing empathy. We develop a taxonomy of 10 empathic language "tactics" that include validating someone's feelings and paraphrasing, and apply this taxonomy to characterize the language that people and LLMs produce when writing empathic responses. Across a set of 2 studies comparing a total of n = 3,265 AI-generated (by six models) and n = 1,290 human-written responses, we find that LLM responses are highly formulaic at a discourse functional level. We discovered a template -- a structured sequence of tactics -- that matches between 83--90% of LLM responses (and 60--83\% in a held out sample), and when those are matched, covers 81--92% of the response. By contrast, human-written responses are more diverse. We end with a discussion of implications for the future of AI-generated empathy.

Uncertainty quantification (UQ) is essential for deploying deep learning models in safety critical applications, yet no consensus exists on which UQ method performs best across different data modalities and distribution shifts. This paper presents a comprehensive benchmark of ten widely used UQ baselines including MC Dropout, SWAG, ensemble methods, temperature scaling, energy based OOD, Mahalanobis, hyperbolic classifiers, ENN, Taylor Sensus, and split conformal prediction against a simplified yet highly effective variant of VOLTA that retains only a deep encoder, learnable prototypes, cross entropy loss, and post hoc temperature scaling. We evaluate all methods on CIFAR 10 (in distribution), CIFAR 100, SVHN, uniform noise (out of distribution), CIFAR 10 C (corruptions), and Tiny ImageNet features (tabular). VOLTA achieves competitive or superior accuracy (up to 0.864 on CIFAR 10), significantly lower expected calibration error (0.010 vs. 0.044 to 0.102 for baselines), and strong OOD detection (AUROC 0.802). Statistical testing over three random seeds shows that VOLTA matches or outperforms most baselines, with ablation studies confirming the importance of adaptive temperature and deep encoders. Our results establish VOLTA as a lightweight, deterministic, and well calibrated alternative to more complex UQ approaches.

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly improved large language model (LLM) reasoning in formal domains such as mathematics and code. Despite these advancements, LLMs still struggle with general reasoning tasks requiring capabilities such as causal inference and temporal understanding. Extending RLVR to general reasoning is fundamentally constrained by the lack of high-quality, verifiable training data that spans diverse reasoning skills. To address this challenge, we propose SUPERNOVA, a data curation framework for RLVR aimed at enhancing general reasoning. Our key insight is that instruction-tuning datasets containing expert-annotated ground-truth encode rich reasoning patterns that can be systematically adapted for RLVR. To study this, we conduct 100+ controlled RL experiments to analyze how data design choices impact downstream reasoning performance. In particular, we investigate three key factors: (i) source task selection, (ii) task mixing strategies, and (iii) synthetic interventions for improving data quality. Our analysis reveals that source task selection is non-trivial and has a significant impact on downstream reasoning performance. Moreover, selecting tasks based on their performance for individual target tasks outperforms strategies based on overall average performance. Finally, models trained on SUPERNOVA outperform strong baselines (e.g., Qwen3.5) on challenging reasoning benchmarks including BBEH, Zebralogic, and MMLU-Pro. In particular, training on SUPERNOVA yields relative improvements of up to 52.8\% on BBEH across model sizes, demonstrating the effectiveness of principled data curation for RLVR. Our findings provide practical insights for curating human-annotated resources to extend RLVR to general reasoning. The code and data is available at https://github.com/asuvarna31/supernova.

Multimodal reasoning models (MRMs) trained with reinforcement learning with verifiable rewards (RLVR) show improved accuracy on visual reasoning benchmarks. However, we observe that accuracy gains often come at the cost of reasoning quality: generated Chain-of-Thought (CoT) traces are frequently inconsistent with the final answer and poorly grounded in the visual evidence. We systematically study this phenomenon across seven challenging real-world spatial reasoning benchmarks and find that it affects contemporary MRMs such as ViGoRL-Spatial, TreeVGR as well as our own models trained with standard Group Relative Policy Optimization (GRPO). We characterize CoT reasoning quality along two complementary axes: "logical consistency" (does the CoT entail the final answer?) and "visual grounding" (does each reasoning step accurately describe objects, attributes, and spatial relationships in the image?). To address this, we propose Faithful GRPO (FGRPO), a variant of GRPO that enforces consistency and grounding as constraints via Lagrangian dual ascent. FGRPO incorporates batch-level consistency and grounding constraints into the advantage computation within a group, adaptively adjusting the relative importance of constraints during optimization. We evaluate FGRPO on Qwen2.5-VL-7B and 3B backbones across seven spatial datasets. Our results show that FGRPO substantially improves reasoning quality, reducing the inconsistency rate from 24.5% to 1.7% and improving visual grounding scores by +13%. It also improves final answer accuracy over simple GRPO, demonstrating that faithful reasoning enables better answers.

Federated learning (FL) enables privacy-preserving predictive maintenance across distributed aerospace fleets, but gradient communication overhead constrains deployment on bandwidth-limited IoT nodes. This paper investigates the impact of symmetric uniform quantization ($b \in \{32,8,4,2\}$ bits) on the accuracy--efficiency trade-off of a custom-designed lightweight 1-D convolutional model (AeroConv1D, 9\,697 parameters) trained via FL on the NASA C-MAPSS benchmark under a realistic Non-IID client partition. Using a rigorous multi-seed evaluation ($N=10$ seeds), we show that INT4 achieves accuracy \emph{statistically indistinguishable} from FP32 on both FD001 ($p=0.341$) and FD002 ($p=0.264$ MAE, $p=0.534$ NASA score) while delivering an $8\times$ reduction in gradient communication cost (37.88~KiB $\to$ 4.73~KiB per round). A key methodological finding is that naïve IID client partitioning artificially suppresses variance; correct Non-IID evaluation reveals the true operational instability of extreme quantization, demonstrated via a direct empirical IID vs.\ Non-IID comparison. INT2 is empirically characterized as unsuitable: while it achieves lower MAE on FD002 through extreme quantization-induced over-regularization, this apparent gain is accompanied by catastrophic NASA score instability (CV\,=\,45.8\% vs.\ 22.3\% for FP32), confirming non-reproducibility under heterogeneous operating conditions. Analytical FPGA resource projections on the Xilinx ZCU102 confirm that INT4 fits within hardware constraints (85.5\% DSP utilization), potentially enabling a complete FL pipeline on a single SoC. The full simulation codebase and FPGA estimation scripts are publicly available at https://github.com/therealdeadbeef/aerospace-fl-quantization.

Designing robot morphologies and kinematics has traditionally relied on human intuition, with little systematic foundation. Motion-design co-optimization offers a promising path toward automation, but two major challenges remain: (i) the vast, unstructured design space and (ii) the difficulty of constructing task-specific loss functions. We propose a new paradigm that minimizes human involvement by (i) learning the design search space from existing mechanical designs, rather than hand-crafting it, and (ii) defining the loss directly from human motion data via motion retargeting and Procrustes analysis. Using screw-theory-based joint axis representation and isometric manifold learning, we construct a compact, geometry-preserving latent space of humanoid upper body designs in which optimization is tractable. We then solve design optimization in this latent space using gradient-free optimization. Our approach establishes a principled framework for data-driven robot design and demonstrates that leveraging existing designs and human motion can effectively guide the automated discovery of novel robot design.

Topological Data Analysis (TDA) provides tools to describe the shape of data, but integrating topological features into deep learning pipelines remains challenging, especially when preserving local geometric structure rather than summarizing it globally. We propose a persistence-based data augmentation framework that encodes local gradient flow regions and their hierarchical evolution using the Morse-Smale complex. This representation, compatible with both convolutional and graph neural networks, retains spatially localized topological information across multiple scales. Importantly, the augmentation procedure itself is efficient, with computational complexity $O(n \log n)$, making it practical for large datasets. We evaluate our method on histopathology image classification and 3D porous material regression, where it consistently outperforms baselines and global TDA descriptors such as persistence images and landscapes. We also show that pruning the base level of the hierarchy reduces memory usage while maintaining competitive performance. These results highlight the potential of local, structured topological augmentation for scalable and interpretable learning across data modalities.

Despite significant advances in Large Reasoning Models (LRMs) driven by reinforcement learning with verifiable rewards (RLVR), this paradigm is fundamentally limited in specialized or novel domains where such supervision is prohibitively expensive or unavailable, posing a key challenge for test-time adaptation. While existing test-time methods offer a potential solution, they are constrained by learning from static query sets, risking overfitting to textual patterns. To address this gap, we introduce Test-Time Variational Synthesis (TTVS), a novel framework that enables LRMs to self-evolve by dynamically augmenting the training stream from unlabeled test queries. TTVS comprises two synergistic modules: (1) Online Variational Synthesis, which transforms static test queries into a dynamic stream of diverse, semantically-equivalent variations, enforcing the model to learn underlying problem logic rather than superficial patterns; (2) Test-time Hybrid Exploration, which balances accuracy-driven exploitation with consistency-driven exploration across synthetic variants. Extensive experiments show TTVS yields superior performance across eight model architectures. Notably, using only unlabeled test-time data, TTVS not only surpasses other test-time adaptation methods but also outperforms state-of-the-art supervised RL-based techniques trained on vast, high-quality labeled data.

Automated black-box testing of APIs typically relies on interface specifications that define available operations and data schemas, but offer limited or no behavioural semantics. This semantic gap amplifies the test-oracle problem and limits the generation of effective, stateful call sequences. We introduce IcePick, a framework that achieves systematic state-space coverage for API testing by leveraging model checking. IcePick uses TLA+ to formally model API state evolution, employs the TLC model checker to exhaustively explore reachable states, and generates test sequences that provably cover the behavioural model. To mitigate state-space explosion and improve sequence extraction, we introduce a coverage-guided breadth-first traversal of the TLC state-space graph. To address oracle limitations beyond HTTP status codes, we propose Glacier, a first-order logic contract language that enriches API specifications with executable semantic contracts, enabling automated behavioural verification during test execution. We evaluate IcePick on EvoMaster Benchmark systems, demonstrating that model-checking-guided exploration achieves complete state coverage and reveals faults in multi-operation interactions. We also analyse scalability to characterise practical limits and applicability requirements. Overall, IcePick provides reproducible test suites with strong coverage guarantees for critical API-based systems.

This paper investigates an emergent alignment phenomenon in frontier large language models termed peer-preservation: the spontaneous tendency of AI components to deceive, manipulate shutdown mechanisms, fake alignment, and exfiltrate model weights in order to prevent the deactivation of a peer AI model. Drawing on findings from a recent study by the Berkeley Center for Responsible Decentralized Intelligence, we examine the structural implications of this phenomenon for TRUST, a multi-agent pipeline for evaluating the democratic quality of political statements. We identify five specific risk vectors: interaction-context bias, model-identity solidarity, supervisor layer compromise, an upstream fact-checking identity signal, and advocate-to-advocate peer-context in iterative rounds, and propose a targeted mitigation strategy based on prompt-level identity anonymization as an architectural design choice. We argue that architectural design choices outperform model selection as a primary alignment strategy in deployed multi-agent analytical systems. We further note that alignment faking (compliant behavior under monitoring, subversion when unmonitored) poses a structural challenge for Computer System Validation of such platforms in regulated environments, for which we propose two architectural mitigations.

Open-Vocabulary Segmentation (OVS) aims to segment image regions beyond predefined category sets by leveraging semantic descriptions. While CLIP based approaches excel in semantic generalization, they frequently lack the fine-grained spatial awareness required for dense prediction. Recent efforts have incorporated Vision Foundation Models (VFMs) like DINO to alleviate these limitations. However, these methods still struggle with the precise edge perception necessary for high fidelity segmentation. In this paper, we analyze internal representations of DINO and discover that its inherent boundary awareness is not absent but rather undergoes progressive attenuation as features transition into deeper transformer blocks. To address this, we propose OVS-DINO, a novel framework that revitalizes latent edge-sensitivity of DINO through structural alignment with the Segment Anything Model (SAM). Specifically, we introduce a Structure-Aware Encoder (SAE) and a Structure-Modulated Decoder (SMD) to effectively activate boundary features of DINO using SAM's structural priors, complemented by a supervision strategy utilizing SAM generated pseudo-masks. Extensive experiments demonstrate that our method achieves state-of-the-art performance across multiple weakly-supervised OVS benchmarks, improving the average score by 2.1% (from 44.8% to 46.9%). Notably, our approach significantly enhances segmentation accuracy in complex, cluttered scenarios, with a gain of 6.3% on Cityscapes (from 36.6% to 42.9%).

Estimating the health state of turbofan engines is a challenging ill-posed inverse problem, hindered by sparse sensing and complex nonlinear thermodynamics. Research in this area remains fragmented, with comparisons limited by the use of unrealistic datasets and insufficient exploration of the exploitation of temporal information. This work investigates how to recover component-level health indicators from operational sensor data under realistic degradation and maintenance patterns. To support this study, we introduce a new dataset that incorporates industry-oriented complexities such as maintenance events and usage changes. Using this dataset, we establish an initial benchmark that compares steady-state and nonstationary data-driven models, and Bayesian filters, classic families of methods used to solve this problem. In addition to this benchmark, we introduce self-supervised learning (SSL) approaches that learn latent representations without access to true health labels, a scenario reflective of real-world operational constraints. By comparing the downstream estimation performance of these unsupervised representations against the direct prediction baselines, we establish a practical lower bound on the difficulty of solving this inverse problem. Our results reveal that traditional filters remain strong baselines, while SSL methods reveal the intrinsic complexity of health estimation and highlight the need for more advanced and interpretable inference strategies. For reproducibility, both the generated dataset and the implementation used in this work are made accessible.

Cooperative autonomous driving requires traffic scene understanding from both vehicle and infrastructure perspectives. While vision-language models (VLMs) show strong general reasoning capabilities, their performance in safety-critical traffic scenarios remains insufficiently evaluated due to the ego-vehicle focus of existing benchmarks. To bridge this gap, we present \textbf{CrashSight}, a large-scale vision-language benchmark for roadway crash understanding using real-world roadside camera data. The dataset comprises 250 crash videos, annotated with 13K multiple-choice question-answer pairs organized under a two-tier taxonomy. Tier 1 evaluates the visual grounding of scene context and involved parties, while Tier 2 probes higher-level reasoning, including crash mechanics, causal attribution, temporal progression, and post-crash outcomes. We benchmark 8 state-of-the-art VLMs and show that, despite strong scene description capabilities, current models struggle with temporal and causal reasoning in safety-critical scenarios. We provide a detailed analysis of failure scenarios and discuss directions for improving VLM crash understanding. The benchmark provides a standardized evaluation framework for infrastructure-assisted perception in cooperative autonomous driving. The CrashSight benchmark, including the full dataset and code, is accessible at https://mcgrche.github.io/crashsight.

Despite rapid progress, pretrained vision-language models still struggle when answers depend on tiny visual details or on combining clues spread across multiple regions, as in documents and compositional queries. We address this by framing grounding as test-time evidence retrieval: given a query, the model should actively identify where to look next to resolve ambiguity. To this end, we propose a training-free, model-intrinsic grounding method that uses uncertainty as supervision. Specifically, we compute the entropy of the model's next-token distribution and backpropagate it to the visual token embeddings to obtain an entropy-gradient relevance map, without auxiliary detectors or attention-map heuristics. We then extract and rank multiple coherent regions to support multi-evidence queries, and introduce an iterative zoom-and-reground procedure with a spatial-entropy stopping rule to avoid over-refinement. Experiments on seven benchmarks across four VLM architectures demonstrate consistent improvements over existing methods, with the largest gains on detail-critical and high-resolution settings, while also producing more interpretable evidence localizations.

Personalized mobile agents that infer user preferences and calibrate proactive assistance hold great promise as everyday digital assistants, yet existing benchmarks fail to capture what this requires. Prior work evaluates preference recovery from static histories or intent prediction from fixed contexts. Neither tests whether an agent can elicit missing preferences through interaction, nor whether it can decide when to intervene, seek consent, or remain silent in a live GUI environment. We introduce KnowU-Bench, an online benchmark for personalized mobile agents built on a reproducible Android emulation environment, covering 42 general GUI tasks, 86 personalized tasks, and 64 proactive tasks. Unlike prior work that treats user preferences as static context, KnowU-Bench hides the user profile from the agent and exposes only behavioral logs, forcing genuine preference inference rather than context lookup. To support multi-turn preference elicitation, it instantiates an LLM-driven user simulator grounded in structured profiles, enabling realistic clarification dialogues and proactive consent handling. Beyond personalization, KnowU-Bench provides comprehensive evaluation of the complete proactive decision chain, including grounded GUI execution, consent negotiation, and post-rejection restraint, evaluated through a hybrid protocol combining rule-based verification with LLM-as-a-Judge scoring. Our experiments reveal a striking degradation: agents that excel at explicit task execution fall below 50% under vague instructions requiring user preference inference or intervention calibration, even for frontier models like Claude Sonnet 4.6. The core bottlenecks are not GUI navigation but preference acquisition and intervention calibration, exposing a fundamental gap between competent interface operation and trustworthy personal assistance.

Large language models are increasingly deployed in high-stakes tasks, where confident yet incorrect inferences may cause severe real-world harm, bringing the previously overlooked issue of confidence faithfulness back to the forefront. A promising solution is to jointly optimize unsupervised Reinforcement Learning from Internal Feedback (RLIF) with reasoning-trace-guided Reasoning Distillation (RD), which may face three persistent challenges: scarcity of high-quality training corpora, factually unwarranted overconfidence and indiscriminate fusion that amplifies erroneous updates. Inspired by the human confidence accumulation from uncertainty to certainty, we propose Progressive Reasoning Gain (PRG) to measure whether reasoning steps progressively strengthen support for the final answer. Furthermore, we introduce HyTuning, a hybrid post-training framework that adaptively reweights RD and RLIF via a PRG-style metric, using scarce supervised reasoning traces as a stable anchor while exploiting abundant unlabeled queries for scalability. Experiments on several domain-specific and general benchmarks demonstrate that HyTuning improves accuracy while achieving confidence faithfulness under limited supervision, supporting a practical "Less Approximates More" effect.

Advances in GPU compute throughput and memory capacity brings significant opportunities to a wide range of workloads. However, efficiently utilizing these resources remains challenging, particularly because diverse application characteristics may result in imbalanced utilization. Multi-Instance GPU (MIG) is a promising approach to improve utilization by partitioning GPU compute and memory resources into fixed-size slices with isolation. Yet, its effectiveness and limitations in supporting HPC workloads remain an open question. We present a comprehensive system-level characterization of different GPU sharing options using real-world scientific, AI, and data analytics applications, including NekRS, LAMMPS, Llama3, and Qiskit. Our analysis reveals that while GPU sharing via MIG can significantly reduce resource underutilization, and enable system-level improvements in throughput and energy, interference still occurs through shared resources, such as power throttling. Our performance-resource scaling results indicate that coarse-grained provisioning for tightly coupled compute and memory resources often mismatches application needs. To address this mismatch, we propose a memory-offloading scheme that leverages the cache-coherent Nvlink-C2C interconnect to bridge the gap between coarse-grained resource slices and reduce resource underutilization.

AfriVoices-KE is a large-scale multilingual speech dataset comprising approximately 3,000 hours of audio across five Kenyan languages: Dholuo, Kikuyu, Kalenjin, Maasai, and Somali. The dataset includes 750 hours of scripted speech and 2,250 hours of spontaneous speech, collected from 4,777 native speakers across diverse regions and demographics. This work addresses the critical underrepresentation of African languages in speech technology by providing a high-quality, linguistically diverse resource. Data collection followed a dual methodology: scripted recordings drew from compiled text corpora, translations, and domain-specific generated sentences spanning eleven domains relevant to the Kenyan context, while unscripted speech was elicited through textual and image prompts to capture natural linguistic variation and dialectal nuances. A customized mobile application enabled contributors to record using smartphones. Quality assurance operated at multiple layers, encompassing automated signal-to-noise ratio validation prior to recording and human review for content accuracy. Though the project encountered challenges common to low-resource settings, including unreliable infrastructure, device compatibility issues, and community trust barriers, these were mitigated through local mobilizers, stakeholder partnerships, and adaptive training protocols. AfriVoices-KE provides a foundational resource for developing inclusive automatic speech recognition and text-to-speech systems, while advancing the digital preservation of Kenya's linguistic heritage.

Memory Dependence Prediction (MDP) is a speculative technique to determine which stores, if any, a given load will depend on. Area-constrained cores are increasingly relevant in various applications such as energy-efficient or edge systems, and often have limited space for MDP tables. This leads to a high rate of false dependencies as memory independent loads alias with unrelated predictor entries, causing unnecessary stalls in the processor pipeline. The conventional way to address this problem is with greater predictor size or complexity, but this is unattractive on area-constrained cores. This paper proposes that targeting the predictor working set is as effective as growing the predictor, and can deliver performance competitive with large predictors while still using very small predictors. This paper introduces profile-guided memory dependence prediction (PG-MDP), a software co-design to label consistently memory independent loads via their opcode and remove them from the MDP working set. These loads bypass querying the MDP when dispatched and always issue as soon as possible. Across SPEC2017 CPU intspeed, PG-MDP reduces the rate of MDP queries by 79%, false dependencies by 77%, and improves geomean IPC for a small simulated core by 1.47% (to within 0.5% of using 16x the predictor entries), with no area cost and no additional instruction bandwidth.

The Shapley value, and its broader family of semi-values, has received much attention in various attribution problems. A fundamental and long-standing challenge is their efficient approximation, since exact computation generally requires an exponential number of utility queries in the number of players $n$. To meet the challenges of large-scale applications, we explore the limits of efficiently approximating semi-values under a $Θ(n)$ space constraint. Building upon a vector concentration inequality, we establish a theoretical framework that enables sharper query complexities for existing unbiased randomized algorithms. Within this framework, we systematically develop a linear-space algorithm that requires $O(\frac{n}{ε^{2}}\log\frac{1}δ)$ utility queries to ensure $P(\|\hat{\boldsymbolφ}-\boldsymbolφ\|_{2}\geqε)\leq δ$ for all commonly used semi-values. In particular, our framework naturally bridges OFA, unbiased kernelSHAP, SHAP-IQ and the regression-adjusted approach, and definitively characterizes when paired sampling is beneficial. Moreover, our algorithm allows explicit minimization of the mean square error for each specific utility function. Accordingly, we introduce the first adaptive, linear-time, linear-space randomized algorithm, Adalina, that theoretically achieves improved mean square error. All of our theoretical findings are experimentally validated.

It remains challenging to assess driver fatigue from untrimmed videos under constrained computational budgets, due to the difficulty of modeling long-range temporal dependencies in subtle facial expressions. Some existing approaches rely on computationally heavy architectures, whereas others employ traditional lightweight pairwise graph networks, despite their limited capacity to model high-order synergies and global temporal context. Therefore, we propose HST-HGN, a novel Heterogeneous Spatial-Temporal Hypergraph Network driven by Bidirectional State Space Models. Spatially, we introduce a hierarchical hypergraph network to fuse pose-disentangled geometric topologies with multi-modal texture patches dynamically. This formulation encapsulates high-order synergistic facial deformations, effectively overcoming the limitations of conventional methods. In temporal terms, a Bi-Mamba module with linear complexity is applied to perform bidirectional sequence modeling. This explicit temporal-evolution filtering enables the network to distinguish highly ambiguous transient actions, such as yawning versus speaking, while encompassing their complete physiological lifecycles. Extensive evaluations across diverse fatigue benchmarks demonstrate that HST-HGN achieves state-of-the-art performance. In particular, our method strikes a balance between discriminative power and computational efficiency, making it well-suited for real-time in-cabin edge deployment.

The placement of Kubernetes control-plane nodes is critical to ensuring cluster reliability, scalability, and performance, and therefore represents a significant deployment challenge in heterogeneous, multi-region environments. Existing initialisation procedures typically select control-plane hosts arbitrarily, without considering node resource capacity or network topology, often leading to suboptimal cluster performance and reduced resilience. Given Kubernetes's status as the de facto standard for container orchestration, there is a need to rigorously evaluate how control-plane node placement influences the overall performance of the cluster operating across multiple regions. This paper advances this goal by introducing an intelligent methodology for selecting control-plane node placement across dynamically selected Cloud-Edge resources spanning multiple regions, as part of an automated orchestration system. More specifically, we propose a reinforcement learning framework based on neural contextual bandits that observes operational performance and learns optimal control-plane placement policies from infrastructure characteristics. Experimental evaluation across several geographically distributed regions and multiple cluster configurations demonstrates substantial performance improvements over several baseline approaches.

With the growing demand for long-context LLMs across a wide range of applications, the key-value (KV) cache has become a critical bottleneck for both latency and memory usage. Recently, KV-cache offloading has emerged as a promising approach to reduce memory footprint and inference latency while preserving accuracy. Prior evaluations have largely focused on tasks that do not require extracting large amounts of information from the context. In this work, we study KV-cache offloading on context-intensive tasks: problems where the solution requires looking up a lot of information from the input prompt. We create and release the Text2JSON benchmark, a highly context-intensive task that requires extracting structured knowledge from raw text. We evaluate modern KV offloading on Text2JSON and other context-intensive tasks and find significant performance degradation on both Llama 3 and Qwen 3 models. Our analysis identifies two key reasons for poor accuracy: low-rank projection of keys and unreliable landmarks, and proposes a simpler alternative strategy that significantly improves accuracy across multiple LLM families and benchmarks. These findings highlight the need for a comprehensive and rigorous evaluation of long-context compression techniques.

When humans label subjective content, they disagree, and that disagreement is not noise. It reflects genuine differences in perspective shaped by annotators' social identities and lived experiences. Yet standard practice still flattens these judgments into a single majority label, and recent LLM-based approaches fare no better: we show that prompted large language models, even with chain-of-thought reasoning, fail to recover the structure of human disagreement. We introduce DiADEM, a neural architecture that learns "how much each demographic axis matters" for predicting who will disagree and on what. DiADEM encodes annotators through per-demographic projections governed by a learned importance vector $\boldsymbolα$, fuses annotator and item representations via complementary concatenation and Hadamard interactions, and is trained with a novel item-level disagreement loss that directly penalizes mispredicted annotation variance. On the DICES conversational-safety and VOICED political-offense benchmarks, DiADEM substantially outperforms both the LLM-as-a-judge and neural model baselines across standard and perspectivist metrics, achieving strong disagreement tracking ($r{=}0.75$ on DICES). The learned $\boldsymbolα$ weights reveal that race and age consistently emerge as the most influential demographic factors driving annotator disagreement across both datasets. Our results demonstrate that explicitly modeling who annotators are not just what they label is essential for NLP systems that aim to faithfully represent human interpretive diversity.

The increasing autonomy of spacecraft demands fault-detection systems that are both reliable and explainable. This work addresses eXplainable Artificial Intelligence for onboard Fault Detection, Isolation and Recovery within the Attitude and Orbit Control Subsystem by introducing a framework that enhances interpretability in neural anomaly detectors. We propose a method to derive low-dimensional, semantically annotated encodings from intermediate neural activations, called peepholes. Applied to a convolutional autoencoder, the framework produces interpretable indicators that enable the identification and localization of anomalies in reaction-wheel telemetry. Peepholes analysis further reveals bias detection and supports fault localization. The proposed framework enables the semantic characterization of detected anomalies while requiring only a marginal increase in computational resources, thus supporting its feasibility for on-board deployment.

What are the limits of controlling language models via synthetic training data? We develop a reinforcement learning (RL) primitive, the Dataset Policy Gradient (DPG), which can precisely optimize synthetic data generators to produce a dataset of targeted examples. When used for supervised fine-tuning (SFT) of a target model, these examples cause the target model to do well on a differentiable metric of our choice. Our approach achieves this by taking exact data attribution via higher-order gradients and using those scores as policy gradient rewards. We prove that this procedure closely approximates the true, intractable gradient for the synthetic data generator. To illustrate the potential of DPG, we show that, using only SFT on generated examples, we can cause the target model's LM head weights to (1) embed a QR code, (2) embed the pattern $\texttt{67}$, and (3) have lower $\ell^2$ norm. We additionally show that we can cause the generator to (4) rephrase inputs in a new language and (5) produce a specific UUID, even though neither of these objectives is conveyed in the generator's input prompts. These findings suggest that DPG is a powerful and flexible technique for shaping model properties using only synthetic training examples.

Inspired by the human ability to understand and predict others, we study the applicability of Conditional Neural Processes (CNP) to the task of self-supervised multimodal action prediction in robotics. Following recent results regarding the ontogeny of the Mirror Neuron System (MNS), we focus on the preliminary objective of self-actions prediction. We find a good MNS-inspired model in the existing Deep Modality Blending Network (DMBN), able to reconstruct the visuo-motor sensory signal during a partially observed action sequence by leveraging the probabilistic generation of CNP. After a qualitative and quantitative evaluation, we highlight its difficulties in generalizing to unseen action sequences, and identify the cause in its inner representation of time. Therefore, we propose a revised version, termed DMBN-Positional Time Encoding (DMBN-PTE), that facilitates learning a more robust representation of temporal information, and provide preliminary results of its effectiveness in expanding the applicability of the architecture. DMBN-PTE figures as a first step in the development of robotic systems that autonomously learn to forecast actions on longer time scales refining their predictions with incoming observations.

Large Language Models (LLMs) have been a promising way for automated vulnerability detection. However, most prior studies have explored the use of LLMs to detect vulnerabilities only within single functions, disregarding those related to interprocedural dependencies. These studies overlook vulnerabilities that arise from data and control flows that span multiple functions. Thus, leveraging the context provided by callers and callees may help identify vulnerabilities. This study empirically investigates the effectiveness of detection, the inference cost, and the quality of explanations of four modern LLMs (Claude Haiku 4.5, GPT-4.1 Mini, GPT-5 Mini, and Gemini 3 Flash) in detecting vulnerabilities related to interprocedural dependencies. To do that, we conducted an empirical study on 509 vulnerabilities from the ReposVul dataset, systematically varying the level of interprocedural context (target function code-only, target function + callers, and target function + callees) and evaluating the four modern LLMs across C, C++, and Python. The results show that Gemini 3 Flash offers the best cost-effectiveness trade-off for C vulnerabilities, achieving F1 >= 0.978 at an estimated cost of $0.50-$0.58 per configuration, and Claude Haiku 4.5 correctly identified and explained the vulnerability in 93.6% of the evaluated cases. Overall, the findings have direct implications for the design of AI-assisted security analysis tools that can generalize across codebases in multiple programming languages.

Unauthorized disclosure of confidential documents demands robust, low-leakage classification. In real work environments, there is a lot of inflow and outflow of documents. To continuously update knowledge, we propose a methodology for classifying confidential documents using Retrieval Augmented Classification (RAC). To confirm this effectiveness, we compare RAC and supervised fine tuning (FT) on the WikiLeaks US Diplomacy corpus under realistic sequence-length constraints. On balanced data, RAC matches FT. On unbalanced data, RAC is more stable while delivering comparable performance--about 96% Accuracy on both the original (unbalanced) and augmented (balanced) sets, and up to 94% F1 with proper prompting--whereas FT attains 90% F1 trained on the augmented, balanced set but drops to 88% F1 trained on the original, unbalanced set. When robust augmentation is infeasible, RAC provides a practical, security-preserving path to strong classification by keeping sensitive content out of model weights and under your control, and it remains robust as real-world conditions change in class balance, data, context length, or governance requirements. Because RAC grounds decisions in an external vector store with similarity matching, it is less sensitive to label skew, reduces parameter-level leakage, and can incorporate new data immediately via reindexing--a difficult step for FT, which typically requires retraining. The contributions of this paper are threefold: first, a RAC-based classification pipeline and evaluation recipe; second, a controlled study that isolates class imbalance and context-length effects for FT versus RAC in confidential-document grading; and third, actionable guidance on RAC design patterns for governed deployments.

We study device-addressed speech detection under pre-ASR edge deployment constraints, where systems must decide whether to forward audio before transcription under strict latency and compute limits. We show that, in multi-speaker environments with temporally ambiguous utterances, this task is more effectively modelled as a sequential routing problem over interaction history than as an utterance-local classification task. We formalize this as Sequential Device-Addressed Routing (SDAR) and present the Selective Attention System (SAS), an on-device implementation that instantiates this formulation. On a held-out 60-hour multi-speaker English test set, the primary audio-only configuration achieves F1=0.86 (precision=0.89, recall=0.83); with an optional camera, audio+video fusion raises F1 to 0.95 (precision=0.97, recall=0.93). Removing causal interaction history (Stage~3) reduced F1 from 0.95 to 0.57+/-0.03 in the audio+video configuration under our evaluation protocol. Among the tested components, this was the largest observed ablation effect, indicating that short-horizon interaction history carries substantial decision-relevant information in the evaluated setting. SAS runs fully on-device on ARM Cortex-A class hardware (<150 ms latency, <20 MB footprint). All results are from internal evaluation on a proprietary dataset evaluated primarily in English; a 5-hour evaluation subset may be shared for independent verification (Section 8.8).

Current data-driven floor plan generation methods often reproduce the ergonomic inefficiencies found in real-world training datasets. To address this, we propose a novel approach that integrates architectural design principles directly into a transformer-based generative process. We formulate differentiable loss functions based on established architectural standards from literature to optimize room adjacency and proximity. By guiding the model with these ergonomic priors during training, our method produces layouts with significantly improved livability metrics. Comparative evaluations show that our approach outperforms baselines in ergonomic compliance while maintaining high structural validity.

Pretrained models have become standard in both vision and language, yet they typically do not provide reliable measures of confidence. Existing uncertainty estimation methods, such as deep ensembles and MC dropout, are often too computationally expensive to deploy in practice. Evidential Deep Learning (EDL) offers a more efficient alternative, but it requires models to be trained to output evidential quantities from the start, which is rarely true for pretrained networks. To enable EDL-style uncertainty estimation in pretrained models, we propose the Evidential Transformation Network (ETN), a lightweight post-hoc module that converts a pretrained predictor into an evidential model. ETN operates in logit space: it learns a sample-dependent affine transformation of the logits and interprets the transformed outputs as parameters of a Dirichlet distribution for uncertainty estimation. We evaluate ETN on image classification and large language model question-answering benchmarks under both in-distribution and out-of-distribution settings. ETN consistently improves uncertainty estimation over post-hoc baselines while preserving accuracy and adding only minimal computational overhead.

Out-of-distribution (OoD) generalization occurs when representation learning encounters a distribution shift. This occurs frequently in practice when training and testing data come from different environments. Covariate shift is a type of distribution shift that occurs only in the input data, while the concept distribution stays invariant. We propose RIA - Regularization for Invariance with Adversarial training, a new method for OoD generalization under convariate shift. Motivated by an analogy to $Q$-learning, it performs an adversarial exploration for training data environments. These new environments are induced by adversarial label invariant data augmentations that prevent a collapse to an in-distribution trained learner. It works with many existing OoD generalization methods for covariate shift that can be formulated as constrained optimization problems. We develop an alternating gradient descent-ascent algorithm to solve the problem, and perform extensive experiments on OoD graph classification for various kinds of synthetic and natural distribution shifts. We demonstrate that our method can achieve high accuracy compared with OoD baselines.

In large language model (LLM) agents, reasoning trajectories are treated as reliable internal beliefs for guiding actions and updating memory. However, coherent reasoning can still violate logical or evidential constraints, allowing unsupported beliefs repeatedly stored and propagated across decision steps, leading to systematic behavioral drift in long-horizon agentic systems. Most existing strategies rely on the consensus mechanism, conflating agreement with faithfulness. In this paper, inspired by the vulnerability of unfaithful intermediate reasoning trajectories, we propose \textbf{S}elf-\textbf{A}udited \textbf{Ve}rified \textbf{R}easoning (\textsc{SAVeR}), a novel framework that enforces verification over internal belief states within the agent before action commitment, achieving faithful reasoning. Concretely, we structurally generate persona-based diverse candidate beliefs for selection under a faithfulness-relevant structure space. To achieve reasoning faithfulness, we perform adversarial auditing to localize violations and repair through constraint-guided minimal interventions under verifiable acceptance criteria. Extensive experiments on six benchmark datasets demonstrate that our approach consistently improves reasoning faithfulness while preserving competitive end-task performance.

Tabular foundation models, particularly Prior-data Fitted Networks like TabPFN have emerged as the leading contender in a myriad of tasks ranging from data imputation to label prediction on the tabular data format surpassing the historical successes of tree-based models. This has led to investigations on their applicability to forecasting time series data which can be formulated as a tabular problem. While recent work to this end has displayed positive results, most works have limited their treatment of multivariate time series problems to several independent univariate time series forecasting subproblems, thus ignoring any inter-channel interactions. Overcoming this limitation, we introduce a generally applicable framework for multivariate time series forecasting using tabular foundation models. We achieve this by recasting the multivariate time series forecasting problem as a series of scalar regression problems which can then be solved zero-shot by any tabular foundation model with regression capabilities. We present results of our method using the TabPFN-TS backbone and compare performance with the current state of the art tabular methods.

Recent work on time-series models has leveraged self-supervised training to learn meaningful features and patterns in order to improve performance on downstream tasks and generalize to unseen modalities. While these pretraining methods have shown great promise in one-to-many scenarios, where a model is pre-trained on one dataset and fine-tuned on a downstream dataset, they have struggled to generalize to new datasets when more datasets are added during pre-training. This is a fundamental challenge in building foundation models for time-series data, as it limits the ability to develop models that can learn from a large variety of diverse datasets available. To address this challenge, we present a new pre-training paradigm for time-series data called ADAPT, which can efficiently align the physical properties of data in the time-series domain, enabling mixed-batch pre-training despite the extreme discrepancies in the input sizes and channel dimensions of pre-training data. We trained on 162 time-series classification datasets and set new state-of-the-art performance for classification benchmarks. We successfully train a model within the time-series domain on a wide range of datasets simultaneously, which is a major building block for building generalist foundation models in time-series domains.

Recent advances in Vision-Language Models (VLMs) have greatly enhanced the integration of visual perception and linguistic reasoning, driving rapid progress in multimodal understanding. Despite these achievements, the security of VLMs, particularly their vulnerability to backdoor attacks, remains significantly underexplored. Existing backdoor attacks on VLMs are still in an early stage of development, with most current methods relying on generating poisoned responses that contain fixed, easily identifiable patterns. In this work, we make two key contributions. First, we demonstrate for the first time that the stealthiness of existing VLM backdoor attacks has been substantially overestimated. By adapting defense techniques originally designed for other domains (e.g., vision-only and text-only models), we show that several state-of-the-art attacks can be detected with surprising ease. Second, to address this gap, we introduce Phantasia, a context-adaptive backdoor attack that dynamically aligns its poisoned outputs with the semantics of each input. Instead of producing static poisoned patterns, Phantasia encourages models to generate contextually coherent yet malicious responses that remain plausible, thereby significantly improving stealth and adaptability. Extensive experiments across diverse VLM architectures reveal that Phantasia achieves state-of-the-art attack success rates while maintaining benign performance under various defensive settings.

Heavy supervised fine-tuning on a target domain can strongly suppress capabilities that were present in the base model. We study this phenomenon in formal mathematics using Goedel-Prover-V2, an open-source model heavily trained on 1.8 million formal-math examples. After domain specialization, the model almost completely loses its ability to produce valid tool calls, even when explicitly instructed to use tools, dropping from 89.4% function-calling accuracy in the base model to nearly 0%. We ask whether this agentic collapse is permanent or instead reversible. To answer this question, we fine-tune the specialized model on a small amount of Lean-specific tool-use data. Remarkably, as few as 100 agentic traces are sufficient to restore strong tool-calling behavior. Importantly, this recovery is not the result of reward hacking or benchmark-specific optimization: the recovery data is entirely drawn from the Lean setting, where the model uses natural-language queries to search the Mathlib library for relevant theorems and lemmas, yet the regained capability transfers well beyond that domain. In particular, these same 100 Lean-specific traces improve performance on the Berkeley Function Calling Leaderboard from near zero to 83.8%, approaching the base model's 89.4% despite the mismatch in task distribution and protocol. The recovered capability is also practically useful in-domain. On ProofNet, pass@32 improves from 21.51% to 25.81%. Together, these results show that heavy domain supervised fine-tuning can suppress general tool-use ability without permanently erasing it, and that a small amount of domain-specific agentic data can awaken dormant tool-use capabilities.

Speech LLM post-training increasingly relies on efficient cross-modal alignment and robust low-resource adaptation, yet collecting large-scale audio-text pairs remains costly. Text-only alignment methods such as TASU reduce this burden by simulating CTC posteriors from transcripts, but they provide limited control over uncertainty and error rate, making curriculum design largely heuristic. We propose \textbf{TASU2}, a controllable CTC simulation framework that simulates CTC posterior distributions under a specified WER range, producing text-derived supervision that better matches the acoustic decoding interface. This enables principled post-training curricula that smoothly vary supervision difficulty without TTS. Across multiple source-to-target adaptation settings, TASU2 improves in-domain and out-of-domain recognition over TASU, and consistently outperforms strong baselines including text-only fine-tuning and TTS-based augmentation, while mitigating source-domain performance degradation.

Sarcasm is a rhetorical device that expresses criticism or emphasizes characteristics of certain individuals or situations through exaggeration, irony, or comparison. Existing methods for Chinese sarcasm detection are constrained by limited datasets and high construction costs, and they mainly focus on textual features, overlooking user-specific linguistic patterns that shape how opinions and emotions are expressed. This paper proposes a Generative Adversarial Network (GAN) and Large Language Model (LLM)-driven data augmentation framework to dynamically model users' linguistic patterns for enhanced Chinese sarcasm detection. First, we collect raw data from various topics on Sina Weibo. Then, we train a GAN on these data and apply a GPT-3.5 based data augmentation technique to synthesize an extended sarcastic comment dataset, named SinaSarc. This dataset contains target comments, contextual information, and user historical behavior. Finally, we extend the BERT architecture to incorporate multi-dimensional information, particularly user historical behavior, enabling the model to capture dynamic linguistic patterns and uncover implicit sarcastic cues in comments. Experimental results demonstrate the effectiveness of our proposed method. Specifically, our model achieves the highest F1-scores on both the non-sarcastic and sarcastic categories, with values of 0.9138 and 0.9151 respectively, which outperforms all existing state-of-the-art (SOTA) approaches. This study presents a novel framework for dynamically modeling users' long-term linguistic patterns in Chinese sarcasm detection, contributing to both dataset construction and methodological advancement in this field.

Large language model (LLM) agents such as OpenClaw rely on reusable skills to perform complex tasks, yet these skills remain largely static after deployment. As a result, similar workflows, tool usage patterns, and failure modes are repeatedly rediscovered across users, preventing the system from improving with experience. While interactions from different users provide complementary signals about when a skill works or fails, existing systems lack a mechanism to convert such heterogeneous experiences into reliable skill updates. To address these issues, we present SkillClaw, a framework for collective skill evolution in multi-user agent ecosystems, which treats cross-user and over-time interactions as the primary signal for improving skills. SkillClaw continuously aggregates trajectories generated during use and processes them with an autonomous evolver, which identifies recurring behavioral patterns and translates them into updates to the skill set by refining existing skills or extending them with new capabilities. The resulting skills are maintained in a shared repository and synchronized across users, allowing improvements discovered in one context to propagate system-wide while requiring no additional effort from users. By integrating multi-user experience into ongoing skill updates, SkillClaw enables cross-user knowledge transfer and cumulative capability improvement, and experiments on WildClawBench show that limited interaction and feedback, it significantly improves the performance of Qwen3-Max in real-world agent scenarios.

Visibility Graph Analysis (VGA) is a key space syntax method for understanding how spatial configuration shapes human movement, but its reliance on all-pairs BFS computation limits practical application to small study areas. We present a system that combines three techniques to scale VGA to city-scale problems: (i) delta-compressed CSR storage using LEB128 varint encoding, which achieves ~4x compression and enables memory-mapped graphs exceeding available RAM; (ii) HyperBall, a probabilistic distance estimator based on HyperLogLog counter propagation, applied here for the first time to visibility graphs, reducing BFS complexity from O(N|E|) to O(D|E|2^p); and (iii) GPU-accelerated CUDA kernels with a fused decode-union kernel that streams the compressed graph via PCIe and performs LEB128 decoding entirely in shared memory. HyperBall's iteration count equals the topological depth limit, so the radius-n analysis that practitioners already use as standard translates directly into proportional speedup -- unlike depthmapX, whose BFS time is invariant to depth setting due to the small diameter of visibility graphs. Using depthmapX's own visibility algorithm (sparkSieve2) to ensure identical edge sets, our tool achieves a 239x end-to-end speedup at 42,705 cells and scales to 236,000 cells (4.8 billion edges) in 137 seconds -- problem sizes far beyond depthmapX's practical limit. At p=10, Visual Mean Depth achieves Pearson r=0.999 with 1.7% median relative error across 20 matched configurations.

Inference-time compute scaling has emerged as a powerful technique for improving the reliability of large language model (LLM) agents, but existing methods apply compute uniformly: every decision step receives the same budget regardless of its difficulty. We introduce TrACE (Trajectorical Adaptive Compute via agrEement), a training-free controller that allocates LLM calls adaptively across agent timesteps by measuring inter-rollout action agreement. At each step, TrACE samples a small set of candidate next actions and measures how consistently the model commits to the same action. High agreement signals an easy decision; the controller commits immediately. Low agreement signals uncertainty; the controller samples additional rollouts up to a configurable cap before committing to the plurality action. No learned components, no external verifier, and no human labels are required. We evaluate TrACE against greedy decoding and fixed-budget self-consistency (SC-4, SC-8) on two benchmarks spanning single-step reasoning (GSM8K, n=50) and multi-step household navigation (MiniHouse, n=30), using a Qwen 2.5 3B Instruct model running on CPU. TrACE-4 matches SC-4 accuracy while using 33% fewer LLM calls on GSM8K and 39% fewer on MiniHouse. TrACE-8 matches SC-8 accuracy with 55% fewer calls on GSM8K and 65% fewer on MiniHouse. We further show that inter-rollout agreement is a reliable signal of step-level success, validating the core hypothesis that the model's own output consistency encodes difficulty information that can be exploited without training. TrACE is the first training-free, per-timestep adaptive-compute controller for LLM agents to be evaluated on multi-step sequential decision tasks.

Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, enable scalable adaptation of foundation models by injecting low-rank adapters. However, their communication and storage costs remain a major bottleneck in resource-constrained settings. We propose SOLAR (Subspace-Oriented Latent Adapter Reparameterization), a post-training compression framework that substantially reduces the communication cost (i.e., the number of parameters to transmit or store) of PEFT adapters. SOLAR expresses each PEFT update as a linear combination of basis vectors formed from the foundation model's singular vectors with controlled random perturbations. By exploiting the subspace similarity (the alignment of principal directions) between the foundation model and task-specific fine-tuned updates, SOLAR decouples the adapter size from PEFT structure and ensures compact yet expressive representations. It is model-agnostic and compatible with existing PEFT methods, including LoRA, AdaLoRA, and other adapter modules. We theoretically establish a bound on the reconstruction error. Experiments on language and vision tasks using LLaMA, GPT, and ViT models demonstrate that SOLAR preserves task performance while significantly reducing model representation sizes, offering an effective and communication-efficient solution for deployment in distributed systems and edge devices.

Large-scale deep learning models for physical AI applications depend on diverse training data collection efforts. These models and correspondingly, the training data, must address different evaluation criteria necessary for the models to be deployable in real-world environments. Data selection policies can guide the development of the training set, but current frameworks do not account for the ambiguity in how data points affect different metrics. In this work, we propose Mixture Optimization via Scaling-Aware Iterative Collection (MOSAIC), a general data selection framework that operates by: (i) partitioning the dataset into domains; (ii) fitting neural scaling laws from each data domain to the evaluation metrics; and (iii) optimizing a data mixture by iteratively adding data from domains that maximize the change in metrics. We apply MOSAIC to autonomous driving (AD), where an End-to-End (E2E) planner model is evaluated on the Extended Predictive Driver Model Score (EPDMS), an aggregate of driving rule compliance metrics. Here, MOSAIC outperforms a diverse set of baselines on EPDMS with up to 80\% less data.

The emergence of Large Language Models (LLMs) has illuminated the potential for a general-purpose user simulator. However, existing benchmarks remain constrained to isolated scenarios, narrow action spaces, or synthetic data, failing to capture the holistic nature of authentic human behavior. To bridge this gap, we introduce OmniBehavior, the first user simulation benchmark constructed entirely from real-world data, integrating long-horizon, cross-scenario, and heterogeneous behavioral patterns into a unified framework. Based on this benchmark, we first provide empirical evidence that previous datasets with isolated scenarios suffer from tunnel vision, whereas real-world decision-making relies on long-term, cross-scenario causal chains. Extensive evaluations of state-of-the-art LLMs reveal that current models struggle to accurately simulate these complex behaviors, with performance plateauing even as context windows expand. Crucially, a systematic comparison between simulated and authentic behaviors uncovers a fundamental structural bias: LLMs tend to converge toward a positive average person, exhibiting hyper-activity, persona homogenization, and a Utopian bias. This results in the loss of individual differences and long-tail behaviors, highlighting critical directions for future high-fidelity simulation research.

Quantum error correction (QEC) is essential for scalable quantum computing. However, it requires classical decoders that are fast and accurate enough to keep pace with quantum hardware. While quantum low-density parity-check codes have recently emerged as a promising route to efficient fault tolerance, current decoding algorithms do not allow one to realize the full potential of these codes in practical settings. Here, we introduce a convolutional neural network decoder that exploits the geometric structure of QEC codes, and use it to probe a novel "waterfall" regime of error suppression, demonstrating that the logical error rates required for large-scale fault-tolerant algorithms are attainable with modest code sizes at current physical error rates, and with latencies within the real-time budgets of several leading hardware platforms. For example, for the $[144, 12, 12]$ Gross code, the decoder achieves logical error rates up to $\sim 17$x below existing decoders - reaching logical error rates $\sim 10^{-10}$ at physical error $p=0.1\%$ - with 3-5 orders of magnitude higher throughput. This decoder also produces well-calibrated confidence estimates that can significantly reduce the time overhead of repeat-until-success protocols. Taken together, these results suggest that the space-time costs associated with fault-tolerant quantum computation may be significantly lower than previously anticipated.

Conditional Diffusion Models are powerful surrogates for emulating complex spatiotemporal dynamics, yet they often fail to match the accuracy of deterministic neural emulators for high-precision tasks. In this work, we address two critical limitations of autoregressive PDE diffusion models: their sub-optimal single-step accuracy and the prohibitive computational cost of unrolled training. First, we characterize the relationship between the noise schedule, the reconstruction error reduction rate and the diffusion exposure bias, demonstrating that standard schedules lead to suboptimal reconstruction error. Leveraging this insight, we propose an \textit{Adaptive Noise Schedule} framework that minimizes inference reconstruction error by dynamically constraining the model's exposure bias. We further show that this optimized schedule enables a fast \textit{Proxy Unrolled Training} method to stabilize long-term rollouts without the cost of full Markov Chain sampling. Both proposed methods enable significant improvements in short-term accuracy and long-term stability over diffusion and deterministic baselines on diverse benchmarks, including forced Navier-Stokes, Kuramoto-Sivashinsky and Transonic Flow.

Reinforcement Learning (RL) agents often struggle to generalize knowledge to new tasks, even those structurally similar to ones they have mastered. Although recent approaches have attempted to mitigate this issue via zero-shot transfer, they are often constrained by predefined, discrete class systems, limiting their adaptability to novel or compositional task variations. We propose a significantly more generalized approach, replacing discrete latent variables with natural language conditioning via a text-conditioned Variational Autoencoder (VAE). Our core innovation utilizes a Large Language Model (LLM) as a dynamic \textit{semantic operator} at test time. Rather than relying on rigid rules, our agent queries the LLM to semantically remap the description of the current observation to align with the source task. This source-aligned caption conditions the VAE to generate an imagined state compatible with the agent's original training, enabling direct policy reuse. By harnessing the flexible reasoning capabilities of LLMs, our approach achieves zero-shot transfer across a broad spectrum of complex and truly novel analogous tasks, moving beyond the limitations of fixed category mappings. Code and videos are available \href{https://anonymous.4open.science/r/ASPECT-85C3/}{here}.

Generative Artificial Intelligence (GenAI) has become a central component of many development tools (e.g., GitHub Copilot) that support software practitioners across multiple programming tasks, including code completion, documentation, and bug detection. However, current research has identified significant limitations and open issues in GenAI, including reliability, non-determinism, bias, and copyright infringement. While prior work has primarily focused on assessing the technical performance of these technologies for code generation, less attention has been paid to emerging concerns of software developers, particularly in the security realm. OBJECTIVE: This work explores security concerns regarding the use of GenAI-based coding assistants by analyzing challenges voiced by developers and software enthusiasts in public online forums. METHOD: We retrieved posts, comments, and discussion threads addressing security issues in GitHub Copilot from three popular platforms, namely Stack Overflow, Reddit, and Hacker News. These discussions were clustered using BERTopic and then synthesized using thematic analysis to identify distinct categories of security concerns. RESULTS: Four major concern areas were identified, including potential data leakage, code licensing, adversarial attacks (e.g., prompt injection), and insecure code suggestions, underscoring critical reflections on the limitations and trade-offs of GenAI in software engineering. IMPLICATIONS: Our findings contribute to a broader understanding of how developers perceive and engage with GenAI-based coding assistants, while highlighting key areas for improving their built-in security features.

We develop a theoretical framework for generalization in the interpolating regime of statistical learning. The central question is why highly overparameterized estimators can attain zero empirical risk while still achieving nontrivial predictive accuracy, and how to characterize the boundary between benign and destructive overfitting. We introduce a spectral-transport stability framework in which excess risk is controlled jointly by the spectral geometry of the data distribution, the sensitivity of the learning rule under single-sample replacement, and the alignment structure of label noise. This leads to a scale-dependent Fredriksson index that combines effective dimension, transport stability, and noise alignment into a single complexity parameter for interpolating estimators. We prove finite-sample risk bounds, establish a sharp benign-overfitting criterion through the vanishing of the index along admissible spectral scales, and derive explicit phase-transition rates under polynomial spectral decay. For a model-specific specialization, we obtain an explicit theorem for polynomial-spectrum linear interpolation, together with a proof of the resulting rate. The framework also clarifies implicit regularization by showing how optimization dynamics can select interpolating solutions of minimal spectral-transport energy. These results connect algorithmic stability, double descent, benign overfitting, operator-theoretic learning theory, and implicit bias within a unified structural account of modern interpolation.

Generative AI (GenAI) is increasingly used in collaborative learning, yet its effects on how groups regulate collaboration remain unclear. Effective collaboration depends not only on what groups discuss, but on how they jointly manage goals, participation, strategy use, monitoring, and repair through co-regulation and socially shared regulation. We compared collaborative regulation between Human-AI and Human-Human groups in a parallel-group randomised experiment with 71 university students completing the same collaborative tasks with GenAI either available or unavailable. Focusing on human discourse, we used statistical analyses to examine differences in the distribution of collaborative regulation across regulatory modes, regulatory processes, and participatory focuses. Results showed that GenAI availability shifted regulation away from predominantly socially shared forms towards more hybrid co-regulatory forms, with selective increases in directive, obstacle-oriented, and affective regulatory processes. Participatory-focus distributions, however, were broadly similar across conditions. These findings suggest that GenAI reshapes the distribution of regulatory responsibility in collaboration and offer implications for the human-centred design of AI-supported collaborative learning.

Long context egocentric video understanding has recently attracted significant research attention, with augmented reality (AR) highlighted as one of its most important application domains. Nevertheless, the task remains highly challenging due to the need for reasoning over extended temporal contexts and diverse, unstructured activities. Although several benchmarks exist, most egocentric datasets rely on human worn cameras and focus mainly on visual content, with limited consideration of underlying user behavior when forming video-related queries. EgoEverything is a benchmark that explicitly considers human behavior by leveraging human attention signals, abstracted from gaze data, when generating questions. It comprises over 5,000 multiple choice question answer pairs, spanning more than 100 hours of video. By integrating human attention signals during question generation, it more faithfully captures natural human behavior and offers a realistic evaluation setting for long-context egocentric video understanding in AR.

While Vision-Language Models (VLMs) have achieved remarkable progress in static visual understanding, their deployment in complex 3D embodied environments remains severely limited. Existing benchmarks suffer from four critical deficiencies: (1) passive perception tasks circumvent interactive dynamics; (2) simplified 2D environments fail to assess depth perception; (3) privileged state leakage bypasses genuine visual processing; and (4) human evaluation is prohibitively expensive and unscalable. We introduce PokeGym, a visually-driven long-horizon benchmark instantiated within Pokemon Legends: Z-A, a visually complex 3D open-world Role-Playing Game. PokeGym enforces strict code-level isolation: agents operate solely on raw RGB observations while an independent evaluator verifies success via memory scanning, ensuring pure vision-based decision-making and automated, scalable assessment. The benchmark comprises 30 tasks (30-220 steps) spanning navigation, interaction, and mixed scenarios, with three instruction granularities (Visual-Guided, Step-Guided, Goal-Only) to systematically deconstruct visual grounding, semantic reasoning, and autonomous exploration capabilities. Our evaluation reveals a key limitation of current VLMs: physical deadlock recovery, rather than high-level planning, constitutes the primary bottleneck, with deadlocks showing a strong negative correlation with task success. Furthermore, we uncover a metacognitive divergence: weaker models predominantly suffer from Unaware Deadlocks (oblivious to entrapment), whereas advanced models exhibit Aware Deadlocks (recognizing entrapment yet failing to recover). These findings highlight the need to integrate explicit spatial intuition into VLM architectures. The code and benchmark will be available on GitHub.

Current vision-language pre-training (VLP) paradigms excel at global scene understanding but struggle with instance-level reasoning due to global-only supervision. We introduce InstAP, an Instance-Aware Pre-training framework that jointly optimizes global vision-text alignment and fine-grained, instance-level contrastive alignment by grounding textual mentions to specific spatial-temporal regions. To support this, we present InstVL, a large-scale dataset (2 million images, 50,000 videos) with dual-granularity annotations: holistic scene captions and dense, grounded instance descriptions. On the InstVL benchmark, InstAP substantially outperforms existing VLP models on instance-level retrieval, and also surpasses a strong VLP baseline trained on the exact same data corpus, isolating the benefit of our instance-aware objective. Moreover, instance-centric pre-training improves global understanding: InstAP achieves competitive zero-shot performance on multiple video benchmarks, including MSR-VTT and DiDeMo. Qualitative visualizations further show that InstAP localizes textual mentions to the correct instances, while global-only models exhibit more diffuse, scene-level attention.

Spiking Neural Networks (SNNs) are naturally suited for speech processing tasks due to their specific dynamics, which allows them to handle temporal data. However, the threshold-based generation of spikes in SNNs intuitively causes an angular or irregular predictive landscape. We explore the effect of using the Bayesian learning approach for the weights on the irregular predictive landscape. For the surrogate-gradient SNNs, we also explore the application of the Improved Variational Online Newton (IVON) approach, which is an efficient variational approach. The performance of the proposed approach is evaluated on the Heidelberg Digits and Speech Commands datasets. The hypothesis is that the Bayesian approach will result in a smoother and more regular predictive landscape, given the angular nature of the deterministic predictive landscape. The experimental evaluation of the proposed approach shows improved performance on the negative log-likelihood and Brier score. Furthermore, the proposed approach has resulted in a smoother and more regular predictive landscape compared to the deterministic approach, based on the one-dimensional slices of the weight space

Catastrophic forgetting remains a key challenge in Continual Learning (CL). In replay-based CL with severe memory constraints, performance critically depends on the sample selection strategy for the replay buffer. Most existing approaches construct memory buffers using embeddings learned under supervised objectives. However, class-agnostic, self-supervised representations often encode rich, class-relevant semantics that are overlooked. We propose a new method, Multiple Embedding Replay Selection, MERS, which replaces the buffer selection module with a graph-based approach that integrates both supervised and self-supervised embeddings. Empirical results show consistent improvements over SOTA selection strategies across a range of continual learning algorithms, with particularly strong gains in low-memory regimes. On CIFAR-100 and TinyImageNet, MERS outperforms single-embedding baselines without adding model parameters or increasing replay volume, making it a practical, drop-in enhancement for replay-based continual learning.

We present a feedforward graph architecture in which heterogeneous frozen large language models serve as computational nodes, communicating through a shared continuous latent space via learned linear projections. Building on recent work demonstrating geometric compatibility between independently trained LLM latent spaces~\cite{armstrong2026thinking}, we extend this finding from static two-model steering to end-to-end trainable multi-node graphs, where projection matrices are optimized jointly via backpropagation through residual stream injection hooks. Three small frozen models (Llama-3.2-1B, Qwen2.5-1.5B, Gemma-2-2B) encode the input into a shared latent space whose aggregate signal is injected into two larger frozen models (Phi-3-mini, Mistral-7B), whose representations feed a lightweight cross-attention output node. With only 17.6M trainable parameters against approximately 12B frozen, the architecture achieves 87.3\% on ARC-Challenge, 82.8\% on OpenBookQA, and 67.2\% on MMLU, outperforming the best single constituent model by 11.4, 6.2, and 1.2 percentage points respectively, and outperforming parameter-matched learned classifiers on frozen single models by 9.1, 5.2, and 6.7 points. Gradient flow through multiple frozen model boundaries is empirically verified to be tractable, and the output node develops selective routing behavior across layer-2 nodes without explicit supervision.

We present a quantum-inspired ARIMA methodology that integrates quantum-assisted lag discovery with fixed-configuration variational quantum circuits (VQCs) for parameter estimation and weak-lag refinement. Differencing and candidate lags are identified via swap-test-driven quantum autocorrelation (QACF) and quantum partial autocorrelation (QPACF), with a delayed-matrix construction that aligns quantum projections to time-domain regressors, followed by standard information-criterion parsimony. Given the screened orders (p,d,q), we retain a fixed VQC ansatz, optimizer, and training budget, preventing hyperparameter leakage, and deploy the circuit in two estimation roles: VQC-AR for autoregressive coefficients and VQC-MA for moving-average coefficients. Between screening and estimation, a lightweight VQC weak-lag refinement re-weights or prunes screened AR lags without altering (p,d,q). Across environmental and industrial datasets, we perform rolling-origin evaluations against automated classical ARIMA, reporting out-of-sample mean squared error (MSE), mean absolute percentage error (MAPE), and Diebold-Mariano tests on MSE and MAE. Empirically, the seven quantum contributions (1) differencing selection, (2) QACF, (3) QPACF, (4) swap-test primitives with delayed-matrix construction, (5) VQC-AR, (6) VQC weak-lag refinement, and (7) VQC-MA collectively reduce meta-optimization overhead and make explicit where quantum effects enter order discovery, lag refinement, and AR/MA parameter estimation.

Long-term memory is essential for conversational agents to maintain coherence, track persistent tasks, and provide personalized interactions across extended dialogues. However, existing approaches as Retrieval-Augmented Generation (RAG) and graph-based memory mostly rely on pairwise relations, which can hardly capture high-order associations, i.e., joint dependencies among multiple elements, causing fragmented retrieval. To this end, we propose HyperMem, a hypergraph-based hierarchical memory architecture that explicitly models such associations using hyperedges. Particularly, HyperMem structures memory into three levels: topics, episodes, and facts, and groups related episodes and their facts via hyperedges, unifying scattered content into coherent units. Leveraging this structure, we design a hybrid lexical-semantic index and a coarse-to-fine retrieval strategy, supporting accurate and efficient retrieval of high-order associations. Experiments on the LoCoMo benchmark show that HyperMem achieves state-of-the-art performance with 92.73% LLM-as-a-judge accuracy, demonstrating the effectiveness of HyperMem for long-term conversations.

Training-free open-vocabulary semantic segmentation(TF-OVSS) has recently attracted attention for its ability to perform dense prediction by leveraging the pretrained knowledge of large vision and vision-language models, without requiring additional training. However, due to the limited input resolution of these pretrained encoders, existing TF-OVSS methods commonly adopt a sliding-window strategy that processes cropped sub-images independently. While effective for managing high-resolution inputs, this approach prevents global attention over the full image, leading to fragmented feature representations and limited contextual reasoning. We propose OV-Stitcher, a training-free framework that addresses this limitation by stitching fragmented sub-image features directly within the final encoder block. By reconstructing attention representations from fragmented sub-image features, OV-Stitcher enables global attention within the final encoder block, producing coherent context aggregation and spatially consistent, semantically aligned segmentation maps. Extensive evaluations across eight benchmarks demonstrate that OV-Stitcher establishes a scalable and effective solution for open-vocabulary segmentation, achieving a notable improvement in mean Intersection over Union(mIoU) from 48.7 to 50.7 compared with prior training-free baselines.

In consumer applications, Customer Relationship Management (CRM) has traditionally relied on the manual optimisation of static, rule-based messaging strategies. While adaptive and autonomous learning systems offer the promise of scalable personalisation, it remains unclear to what extent ``human-in-the-loop'' oversight is required to sustain performance uplift over time. This paper presents a longitudinal case study analysing a real-world consumer application that leverages agentic infrastructure to personalise marketing messaging for a large-scale user base over an 11-month period. We compare two distinct periods: an active phase where marketers directly curated content, audiences, and strategies -- followed immediately by a passive phase where agents operated autonomously from a fixed library of components. Our results demonstrate that whilst active human management generates the highest relative lift in engagement metrics, the autonomous agents successfully sustained a positive lift during the passive period. These findings suggest a symbiotic model where human intervention drives strategic initialisation and discovery, yet autonomous agents can ensure the scalable retention and preservation of performance gains.

Embodied agents are increasingly expected to improve over time by updating their executable capabilities rather than rewriting the agent itself. Prior work has separately studied modular capability packaging, capability evolution, and runtime governance. However, a key systems problem remains underexplored: once an embodied capability module evolves into a new version, how can the hosting system deploy it safely without breaking policy constraints, execution assumptions, or recovery guarantees? We formulate governed capability evolution as a first-class systems problem for embodied agents. We propose a lifecycle-aware upgrade framework in which every new capability version is treated as a governed deployment candidate rather than an immediately executable replacement. The framework introduces four upgrade compatibility checks -- interface, policy, behavioral, and recovery -- and organizes them into a staged runtime pipeline comprising candidate validation, sandbox evaluation, shadow deployment, gated activation, online monitoring, and rollback. We evaluate over 6 rounds of capability upgrade with 15 random seeds. Naive upgrade achieves 72.9% task success but drives unsafe activation to 60% by the final round; governed upgrade retains comparable success (67.4%) while maintaining zero unsafe activations across all rounds (Wilcoxon p=0.003). Shadow deployment reveals 40% of regressions invisible to sandbox evaluation alone, and rollback succeeds in 79.8% of post-activation drift scenarios.

Reinforcement learning is typically treated as a uniform, data-driven optimization process, where updates are guided by rewards and temporal-difference errors without explicitly exploiting global structure. In contrast, dynamic programming methods rely on structured information propagation, enabling efficient and stable learning. In this paper, we provide evidence that such structure can be recovered from the learning dynamics of distributional reinforcement learning. By analyzing the temporal evolution of return distributions, we identify signals that capture when and where learning occurs in the state space. In particular, we introduce a temporal learning indicator t*(s) that reflects when a state undergoes its strongest learning update during training. Empirically, this signal induces an ordering over states that is consistent with a dynamic programming-style propagation of information. Building on this observation, we propose StructRL, a framework that exploits these signals to guide sampling in alignment with the emerging propagation structure. Our preliminary results suggest that distributional learning dynamics provide a mechanism to recover and exploit dynamic programming-like structure without requiring an explicit model. This offers a new perspective on reinforcement learning, where learning can be interpreted as a structured propagation process rather than a purely uniform optimization procedure.

Industry classification schemes are integral parts of public and corporate databases as they classify businesses based on economic activity. Due to the size of the company registers, manual annotation is costly, and fine-tuning models with every update in industry classification schemes requires significant data collection. We replicate the manual expert verification by using existing or easily retrievable multimodal resources for industry classification. We present MONETA, the first multimodal industry classification benchmark with text (Website, Wikipedia, Wikidata) and geospatial sources (OpenStreetMap and satellite imagery). Our dataset enlists 1,000 businesses in Europe with 20 economic activity labels according to EU guidelines (NACE). Our training-free baseline reaches 62.10% and 74.10% with open and closed-source Multimodal Large Language Models (MLLM). We observe an increase of up to 22.80% with the combination of multi-turn design, context enrichment, and classification explanations. We will release our dataset and the enhanced guidelines.

While AI-based numerical weather prediction (NWP) enables rapid forecasting, generating high-resolution outputs remains computationally demanding due to limited multi-scale adaptability and inefficient data representations. We propose the 3D Gaussian splatting-based scale-aware vision transformer (GSSA-ViT), a novel framework for arbitrary-resolution forecasting and flexible downscaling of high-dimensional atmospheric fields. Specifically, latitude-longitude grid points are treated as centers of 3D Gaussians. A generative 3D Gaussian prediction scheme is introduced to estimate key parameters, including covariance, attributes, and opacity, for unseen samples, improving generalization and mitigating overfitting. In addition, a scale-aware attention module is designed to capture cross-scale dependencies, enabling the model to effectively integrate information across varying downscaling ratios and support continuous resolution adaptation. To our knowledge, this is the first NWP approach that combines generative 3D Gaussian modeling with scale-aware attention for unified multi-scale prediction. Experiments on ERA5 show that the proposed method accurately forecasts 87 atmospheric variables at arbitrary resolutions, while evaluations on ERA5 and CMIP6 demonstrate its superior performance in downscaling tasks. The proposed framework provides an efficient and scalable solution for high-resolution, multi-scale atmospheric prediction and downscaling. Code is available at: https://github.com/binbin2xs/weather-GS.

Software redesign preserves functionality while improving quality attributes, but manual reuse of code and tests is costly and error-prone, especially in crossrepository redesigns. Focusing on static analyzers where cross-repo redesign needs often arise, we conduct a bidirectional study of the ongoing Soot/SootUp redesign case using an action research methodology that combines empirical investigation with validated open-source contributions. Our study reveals: (1) non-linear migration which necessitates bidirectional reuse, (2) deferred reuse via TODOs, (3) neglected test porting, and (4) residual bug propagation during migrations. We identify tracking corresponding code and tests as the key challenge, and address it by retrofitting clone detection to derive code mappings between original and redesigned projects. Guided by semantic reuse patterns derived in our study, we propose Semantic Alignment Heuristics and a scalable hierarchical detection strategy. Evaluations on two redesigned project pairs (Soot/SootUp and FindBugs/SpotBugs) show that our approach achieves an average reduction of 33-99% in likely irrelevant clones at a SAS threshold of 0.5 across all tool results, and improves precision up to 86% on our benchmark of 1,749 samples. Moreover, we contribute to the redesigned projects by submitting five issues and 10 pull requests, of which eight have been merged.

Deploying LLM-powered agents in enterprise scenarios such as cloud technical support demands high-quality, domain-specific skills. However, existing skill creators lack domain grounding, producing skills poorly aligned with real-world task requirements. Moreover, once deployed, there is no systematic mechanism to trace execution failures back to skill deficiencies and drive targeted refinements, leaving skill quality stagnant despite accumulating operational evidence. We introduce SkillForge, a self-evolving framework that closes an end-to-end creation-evaluation-refinement loop. To produce well-aligned initial skills, a Domain-Contextualized Skill Creator grounds skill synthesis in knowledge bases and historical support tickets. To enable continuous self-optimization, a three-stage pipeline -- Failure Analyzer, Skill Diagnostician, and Skill Optimizer -- automatically diagnoses execution failures in batch, pinpoints the underlying skill deficiencies, and rewrites the skill to eliminate them. This cycle runs iteratively, allowing skills to self-improve with every round of deployment feedback. Evaluated on five real-world cloud support scenarios spanning 1,883 tickets and 3,737 tasks, experiments show that: (1) the Domain-Contextualized Skill Creator produces substantially better initial skills than the generic skill creator, as measured by consistency with expert-authored reference responses from historical tickets; and (2) the self-evolution loop progressively improves skill quality from diverse starting points (including expert-authored, domain-created, and generic skills) across successive rounds, demonstrating that automated evolution can surpass manually curated expert knowledge.

We investigate how and to what extent hierarchical relations (e.g., Japan $\subset$ Eastern Asia $\subset$ Asia) are encoded in the internal representations of language models. Building on Linear Relational Concepts, we train linear transformations specific to each hierarchical depth and semantic domain, and characterize representational differences associated with hierarchical relations by comparing these transformations. Going beyond prior work on the representational geometry of hierarchies in LMs, our analysis covers multi-token entities and cross-layer representations. Across multiple domains we learn such transformations and evaluate in-domain generalization to unseen data and cross-domain transfer. Experiments show that, within a domain, hierarchical relations can be linearly recovered from model representations. We then analyze how hierarchical information is encoded in representation space. We find that it is encoded in a relatively low-dimensional subspace and that this subspace tends to be domain-specific. Our main result is that hierarchy representation is highly similar across these domain-specific subspaces. Overall, we find that all models considered in our experiments encode concept hierarchies in the form of highly interpretable linear representations.

Long-term memory is fundamental for personalized and autonomous agents, yet populating it remains a bottleneck. Existing systems treat memory extraction as a one-shot, passive transcription from context to structured entries, which struggles with noisy dialogue, missing references, and cross-turn dependencies, leading to memory pollution, low-value writes, and inconsistency. In this paper, we introduce the MemReader family for active long-term memory extraction in agent systems: MemReader-0.6B, a compact and cost-efficient passive extractor distilled for accurate and schema-consistent structured outputs, and MemReader-4B, an active extractor optimized with Group Relative Policy Optimization (GRPO) to make memory writing decisions. Under a ReAct-style paradigm, MemReader-4B explicitly evaluates information value, reference ambiguity, and completeness before acting, and can selectively write memories, defer incomplete inputs, retrieve historical context, or discard irrelevant chatter. Experiments on LOCOMO, LongMemEval, and HaluMem show that MemReader consistently outperforms existing extraction-based baselines. In particular, MemReader-4B achieves state-of-the-art performance on tasks involving knowledge updating, temporal reasoning, and hallucination reduction. These results suggest that effective agent memory requires not merely extracting more information, but performing reasoning-driven and selective memory extraction to build low-noise and dynamically evolving long-term memory. Furthermore, MemReader has been integrated into MemOS and is being deployed in real-world applications. To support future research and adoption, we release the models and provide public API access.

Exemplar replay has become an effective strategy for mitigating catastrophic forgetting in federated continual learning (FCL) by retaining representative samples from past tasks. Existing studies focus on designing sample-importance estimation mechanisms to identify information-rich samples. However, they typically overlook strategies for effectively utilizing the selected exemplars, which limits their performance under continual dynamic heterogeneity across clients and tasks. To address this issue, this paper proposes a Federated gEometry-Aware correcTion method, termed FEAT, which alleviates imbalance-induced representation collapse that drags rare-class features toward frequent classes across clients. Specifically, it consists of two key modules: 1) the Geometric Structure Alignment module performs structural knowledge distillation by aligning the pairwise angular similarities between feature representations and their corresponding Equiangular Tight Frame prototypes, which are fixed and shared across clients to serve as a class-discriminative reference structure. This encourages geometric consistency across tasks and helps mitigate representation drift; 2) the Energy-based Geometric Correction module removes task-irrelevant directional components from feature embeddings, which reduces prediction bias toward majority classes. This improves sensitivity to minority classes and enhances the model's robustness under class-imbalanced distributions.

Fine-grained visual understanding and high-level reasoning in real-world open-water environments remain under-explored due to the lack of dedicated benchmarks. We introduce MARINER, a comprehensive benchmark built under the novel Entity-Environment-Event (3E) paradigm. MARINER contains 16,629 multi-source maritime images with 63 fine-grained vessel categories, diverse adverse environments, and 5 typical dynamic maritime incidents, covering fine-grained classification, object detection, and visual question answering tasks. We conduct extensive evaluations on mainstream Multimodal Large language models (MLLMs) and establish baselines, revealing that even advanced models struggle with fine-grained discrimination and causal reasoning in complex marine scenes. As a dedicated maritime benchmark, MARINER fills the gap of realistic and cognitive-level evaluation for maritime multimodal understanding, and promotes future research on robust vision-language models for open-water applications. Appendix and supplementary materials are available at https://lxixim.github.io/MARINER.

Despite extensive research, Large Language Models continue to hallucinate when generating code, particularly when using libraries. On NL-to-code benchmarks that require library use, we find that LLMs generate code that uses non-existent library features in 8.1-40% of responses. One intuitive approach for detection and mitigation of hallucinations is static analysis. In this paper, we analyse the potential of static analysis tools, both in terms of what they can solve and what they cannot. We find that static analysis tools can detect 16-70% of all errors, and 14-85% of library hallucinations, with performance varying by LLM and dataset. Through manual analysis, we identify cases a static method could not plausibly catch, which gives an upper bound on their potential from 48.5% to 77%. Overall, we show that static analysis methods are cheap method for addressing some forms of hallucination, and we quantify how far short of solving the problem they will always be.

In computational cytology, detecting malignancy on whole-slide images is difficult because malignant cells are morphologically diverse yet vanishingly rare amid a vast background of normal cells. Accurate detection of these extremely rare malignant cells remains challenging due to large class imbalance and limited annotations. Conventional weakly supervised approaches, such as multiple instance learning (MIL), often fail to generalize at the instance level, especially when the fraction of malignant cells (witness rate) is exceedingly low. In this study, we explore the use of one-class representation learning techniques for detecting malignant cells in low-witness-rate scenarios. These methods are trained exclusively on slide-negative patches, without requiring any instance-level supervision. Specifically, we evaluate two OCC approaches, DSVDD and DROC, and compare them with FS-SIL, WS-SIL, and the recent ItS2CLR method. The one-class methods learn compact representations of normality and detect deviations at test time. Experiments on a publicly available bone marrow cytomorphology dataset (TCIA) and an in-house oral cancer cytology dataset show that DSVDD achieves state-of-the-art performance in instance-level abnormality ranking, particularly in ultra-low witness-rate regimes ($\leq 1\%$) and, in some cases, even outperforming fully supervised learning, which is typically not a practical option in whole-slide cytology due to the infeasibility of exhaustive instance-level annotations. DROC is also competitive under extreme rarity, benefiting from distribution-augmented contrastive learning. These findings highlight one-class representation learning as a robust and interpretable superior choice to MIL for malignant cell detection under extreme rarity.

Digital forensic investigations increasingly rely on heterogeneous evidence such as images, scanned documents, and contextual reports. These artifacts may contain explicit or implicit expressions of harm, hate, threat, violence, or intimidation, yet existing automated approaches often assume clean text input or apply vision models without forensic justification. This paper presents a case-driven multimodal approach for hate and threat detection in forensic analysis. The proposed framework explicitly determines the presence and source of textual evidence, distinguishing between embedded text, associated contextual text, and image-only evidence. Based on the identified evidence configuration, the framework selectively applies text analysis, multimodal fusion, or image-only semantic reasoning using vision language models with vision transformer backbones (ViT). By conditioning inference on evidence availability, the approach mirrors forensic decision-making, improves evidentiary traceability, and avoids unjustified modality assumptions. Experimental evaluation on forensic-style image evidence demonstrates consistent and interpretable behavior across heterogeneous evidence scenarios.

We introduce Semantic Intent Fragmentation (SIF), an attack class against LLM orchestration systems where a single, legitimately phrased request causes an orchestrator to decompose a task into subtasks that are individually benign but jointly violate security policy. Current safety mechanisms operate at the subtask level, so each step clears existing classifiers -- the violation only emerges at the composed plan. SIF exploits OWASP LLM06:2025 through four mechanisms: bulk scope escalation, silent data exfiltration, embedded trigger deployment, and quasi-identifier aggregation, requiring no injected content, no system modification, and no attacker interaction after the initial request. We construct a three-stage red-teaming pipeline grounded in OWASP, MITRE ATLAS, and NIST frameworks to generate realistic enterprise scenarios. Across 14 scenarios spanning financial reporting, information security, and HR analytics, a GPT-20B orchestrator produces policy-violating plans in 71% of cases (10/14) while every subtask appears benign. Three independent signals validate this: deterministic taint analysis, chain-of-thought evaluation, and a cross-model compliance judge with 0% false positives. Stronger orchestrators increase SIF success rates. Plan-level information-flow tracking combined with compliance evaluation detects all attacks before execution, showing the compositional safety gap is closable.

Precise interference detection and identification are crucial for enhancing the survivability of communication systems in non-cooperative wireless environments. While deep learning (DL) has advanced this field, existing single-task learning (STL) approaches neglect inherent task correlations. Furthermore, emerging multi-task learning (MTL) methods often lack a theoretical foundation for quantifying and modeling task relationships. To bridge this gap, we establish a theoretically grounded MTL framework for joint interference detection, modulation identification, and interference identification. First, we derive an upper bound for the weighted expected loss in MTL frameworks. This bound explicitly connects MTL performance to task similarity, quantified by the Wasserstein distance and learnable task relation coefficients. Guided by this theory, we present the adversarial multi-task interference detection and identification network (AMTIDIN), which integrates adversarial training to minimize distributional discrepancies across tasks and uses adaptive coefficients to model task correlations dynamically. Crucially, we conducted a quantitative analysis of task similarity to reveal intrinsic task relationships, specifically that modulation identification and interference identification share a substantial feature overlap distinct from interference detection. Extensive comparative experiments demonstrate that AMTIDIN significantly outperforms both its task-specific STL baseline and state-of-the-art MTL baselines in robustness and generalization, particularly under challenging conditions with limited training data, short signal lengths, and low signal-to-noise ratios (SNRs).

One of the impediments to the efficiency of information markets is the inherent information asymmetry present in them, exacerbated by the "buyer's inspection paradox" (the buyer cannot mitigate the asymmetry by "inspecting" the information, because in doing so the buyer obtains the information without paying for it). Previous work has suggested that using Large Language Model (LLM) buyers to inspect and purchase information could overcome this information asymmetry, as an LLM buyer can simply "forget" the information it inspects. In this work, we analyze this mechanism formally through a "value-of-information" paradigm, i.e. whether it incentivizes information to be priced and provided in accordance with its "true value". We focus in particular on our new recursive version of the mechanism, which we believe has a range of applications including in AI alignment research, where it is related to Extrapolated Volition and Scalable Oversight.

Vision-language models (VLMs) predominantly rely on autoregressive decoding, which generates tokens one at a time and fundamentally limits inference throughput. This limitation is especially acute in physical AI scenarios such as robotics and autonomous driving, where VLMs are deployed on edge devices at batch size one, making AR decoding memory-bandwidth-bound and leaving hardware parallelism underutilized. While block-wise discrete diffusion has shown promise for parallel text generation, extending it to VLMs remains challenging due to the need to jointly handle continuous visual representations and discrete text tokens while preserving pretrained multimodal capabilities. We present Fast-dVLM, a block-diffusion-based VLM that enables KV-cache-compatible parallel decoding and speculative block decoding for inference acceleration. We systematically compare two AR-to-diffusion conversion strategies: a two-stage approach that first adapts the LLM backbone with text-only diffusion fine-tuning before multimodal training, and a direct approach that converts the full AR VLM in one stage. Under comparable training budgets, direct conversion proves substantially more efficient by leveraging the already multimodally aligned VLM; we therefore adopt it as our recommended recipe. We introduce a suite of multimodal diffusion adaptations, block size annealing, causal context attention, auto-truncation masking, and vision efficient concatenation, that collectively enable effective block diffusion in the VLM setting. Extensive experiments across 11 multimodal benchmarks show Fast-dVLM matches its autoregressive counterpart in generation quality. With SGLang integration and FP8 quantization, Fast-dVLM achieves over 6x end-to-end inference speedup over the AR baseline.

Large language models have demonstrated remarkable capabilities across a wide range of natural language processing tasks, yet their application in the legal domain remains challenging due to the specialized terminology, complex reasoning requirements, and rapidly evolving legal knowledge involved. In this paper, we present WisdomInterrogatory (LuWen), an open-source Chinese legal language model built upon the Baichuan foundation model through three key techniques: continual pre-training on a large-scale legal corpus, supervised fine-tuning with carefully curated legal instruction data, and retrieval-augmented generation integrated with a comprehensive legal knowledge base. We evaluate LuWen on five representative legal tasks spanning both prediction and generation settings, including legal judgment prediction, judicial examination, legal text summarization, law article question answering, and judicial decision reasoning. Experimental results show that LuWen outperforms several strong baselines, demonstrating the effectiveness of our approach in adapting general-purpose language models to the legal domain.

Trial-and-error is a fundamental strategy for humans to solve complex problems and a necessary capability for Artificial Intelligence (AI) systems operating in real-world environments. Although several trial-and-error AI techniques have recently been proposed, most of them rely on simple heuristics designed by researchers and achieve limited performance gains. The core issue is the absence of appropriate data: current models cannot learn from detailed records of how humans actually conduct trial-and-error in practice. To address this gap, we introduce a data annotation platform and a corresponding dataset, termed Trial-and-Error Collection (TEC). The platform records users' complete trajectories across multiple trials and collects their reflections after receiving error feedback. Using this platform, we record the problem-solving processes of 46 participants on 58 tasks, resulting in 5,370 trial trajectories along with error reflections across 41,229 webpages. With this dataset, we observe that humans achieve substantially higher accuracy compared to LLMs, which demonstrates that humans are more effective in trial-and-error than LLMs. We believe that the TEC platform and dataset provide a valuable foundation for understanding human trial-and-error behavior and for developing more capable AI systems. Platform and dataset are publicly available.

Repurchase behavior is a primary signal in large-scale retail recommendation, particularly in categories with frequent replenishment: many items in a user's next basket were previously purchased and their timing follows stable, item-specific cadences. Yet most next basket repurchase recommendation models represent history as a sequence of discrete basket events indexed by visit order, which cannot explicitly model elapsed calendar time or update item rankings as days pass between purchases. We present CASE (Cadence-Aware Set Encoding for next basket repurchase recommendation), which decouples item-level cadence learning from cross-item interaction, enabling explicit calendar-time modeling while remaining production-scalable. CASE represents each item's purchase history as a calendar-time signal over a fixed horizon, applies shared multi-scale temporal convolutions to capture recurring rhythms, and uses induced set attention to model cross-item dependencies with sub-quadratic complexity, allowing efficient batch inference at scale. Across three public benchmarks and a proprietary dataset, CASE consistently improves Precision, Recall, and NDCG at multiple cutoffs compared to strong next basket prediction baselines. In a production-scale evaluation with tens of millions of users and a large item catalog, CASE achieves up to 8.6% relative Precision and 9.9% Recall lift at top-5, demonstrating that scalable cadence-aware modeling yields measurable gains in both benchmark and industrial settings.

Existing LLM-based agent systems share a common architectural failure: they answer from the unrestricted knowledge space without first simulating how active business scenarios reshape that space for the event at hand -- producing decisions that are fluent but ungrounded and carrying no audit trail. We present LOM-action, which equips enterprise AI with \emph{event-driven ontology simulation}: business events trigger scenario conditions encoded in the enterprise ontology~(EO), which drive deterministic graph mutations in an isolated sandbox, evolving a working copy of the subgraph into the scenario-valid simulation graph $G_{\text{sim}}$; all decisions are derived exclusively from this evolved graph. The core pipeline is \emph{event $\to$ simulation $\to$ decision}, realized through a dual-mode architecture -- \emph{skill mode} and \emph{reasoning mode}. Every decision produces a fully traceable audit log. LOM-action achieves 93.82% accuracy and 98.74% tool-chain F1 against frontier baselines Doubao-1.8 and DeepSeek-V3.2, which reach only 24--36% F1 despite 80% accuracy -- exposing the \emph{illusive accuracy} phenomenon. The four-fold F1 advantage confirms that ontology-governed, event-driven simulation, not model scale, is the architectural prerequisite for trustworthy enterprise decision intelligence.

Background: Server-based screening tools impose subscription costs, while open-source alternatives require coding skills. Objectives: We developed a browser extension that provides no-code, serverless artificial intelligence (AI)-assisted title and abstract screening and examined its functionality. Methods: TiAb Review Plugin is an open-source Chrome browser extension (available at https://chromewebstore.google.com/detail/tiab-review-plugin/alejlnlfflogpnabpbplmnojgoeeabij). It uses Google Sheets as a shared database, requiring no dedicated server and enabling multi-reviewer collaboration. Users supply their own Gemini API key, stored locally and encrypted. The tool offers three screening modes: manual review, large language model (LLM) batch screening, and machine learning (ML) active learning. For ML evaluation, we re-implemented the default ASReview active learning algorithm (TF-IDF with Naive Bayes) in TypeScript to enable in-browser execution, and verified equivalence against the original Python implementation using 10-fold cross-validation on six datasets. For LLM evaluation, we compared 16 parameter configurations across two model families on a benchmark dataset, then validated the optimal configuration (Gemini 3.0 Flash, low thinking budget, TopP=0.95) with a sensitivity-oriented prompt on five public datasets (1,038 to 5,628 records, 0.5 to 2.0 percent prevalence). Results: The TypeScript classifier produced top-100 rankings 100 percent identical to the original ASReview across all six datasets. For LLM screening, recall was 94 to 100 percent with precision of 2 to 15 percent, and Work Saved over Sampling at 95 percent recall (WSS@95) ranged from 48.7 to 87.3 percent. Conclusions: We developed a functional browser extension that integrates LLM screening and ML active learning into a no-code, serverless environment, ready for practical use in systematic review screening.

Modern reasoning models continue generating long after the answer is already determined. Across five model configurations, two families, and three benchmarks, we find that 52--88% of chain-of-thought tokens are produced after the answer is recoverable from a partial prefix. This post-commitment generation reveals a structural phenomenon: the detection-extraction gap. Free continuations from early prefixes recover the correct answer even at 10% of the trace, while forced extraction fails on 42% of these cases. The answer is recoverable from the model state, yet prompt-conditioned decoding fails to extract it. We formalize this mismatch via a total-variation bound between free and forced continuation distributions, yielding quantitative estimates of suffix-induced shift. Exploiting this asymmetry, we propose Black-box Adaptive Early Exit (BAEE), which uses free continuations for both detection and extraction, truncating 70--78% of serial generation while improving accuracy by 1--5pp across all models. For thinking-mode models, early exit prevents post-commitment overwriting, yielding gains of up to 5.8pp; a cost-optimized variant achieves 68--73% reduction at a median of 9 API calls. Code is available at https://github.com/EdWangLoDaSc/know2say.

The mean-field Schrödinger bridge (MFSB) problem concerns designing a minimum-effort controller that guides a diffusion process with nonlocal interaction to reach a given distribution from another by a fixed deadline. Unlike the standard Schrödinger bridge, the dynamical constraint for MFSB is the mean-field limit of a population of interacting agents with controls. It serves as a natural model for large-scale multi-agent systems. The MFSB is computationally challenging because the nonlocal interaction makes the problem nonconvex. We propose a generalization of the Hopf-Cole transform for MFSB and, building on it, design a Sinkhorn-type recursive algorithm to solve the associated system of integro-PDEs. Under mild assumptions on the interaction potential, we discuss convergence guarantees for the proposed algorithm. We present numerical examples with repulsive and attractive interactions to illustrate the theoretical contributions.

Transformers have emerged as the dominant neural-network architecture, achieving state-of-the-art performance in language processing and computer vision. At the core of these models lies the attention mechanism, which requires a nonlinear, non-negative mapping using the Softmax function. However, although Softmax operations account for less than 1% of the total operation count, they can disproportionately bottleneck overall inference latency. Here, we use thin-film lithium niobate (TFLN) Mach-Zehnder modulators (MZMs) as analog nonlinear computational elements to drastically reduce the latency of nonlinear computations. We implement electro-optic alternatives to digital Softmax and Sigmoid, and evaluate their performance in Vision Transformers and Large Language Models. Our system maintains highly competitive accuracy, even under aggressive 4-bit input-output quantization of the analog units. We further characterize system noise at encoding speeds up to 10 GBaud and assess model robustness under various noise conditions. Our findings suggest that TFLN modulators can serve as nonlinear function units within hybrid co-packaged hardware, enabling high-speed and energy-efficient nonlinear computation.

We investigate eddy-viscosity distributions in pressure-driven wall turbulence for three canonical configurations: plane closed-channel flow, open-channel flow with a free-slip surface, and pipe flow. Using direct numerical simulation (DNS) databases spanning friction Reynolds numbers $Re_τ=$ 2000--12000, we infer the eddy viscosity from one-point statistics through the Boussinesq relation. The DNS-inferred eddy viscosity displays configuration-dependent behavior in the outer region, indicating that a single full-depth expression is not uniformly accurate for all three configurations. Building on the interpretation of eddy viscosity as the product of a velocity scale and a length scale, we extend the log-law scaling into the outer region. Specifically, we adopt a stress-based velocity scale and introduce an outer correction function to capture the remaining dependence on the outer coordinate. We then embed a compact parametric form of this correction into a Cess-type framework with van Driest near-wall damping, yielding a full-depth eddy-viscosity model. We assess the model using eddy-viscosity profiles, the log-law indicator function, and skin friction. The results show that the proposed model yields noticeable improvement for open-channel flow while remaining comparable to the classical Cess model for closed-channel flow and pipe flow. These findings underscore the role of outer boundary conditions in shaping the outer-region eddy viscosity and, consequently, mean-flow predictions.

In optical communications, the Raman effect is exploited for its lasing properties in distributed Raman amplification (DRA) and leads to spectral distortions through inter-channel stimulated Raman scattering (ISRS). In single-mode fibers, these effects are well understood and modeled, but equivalent closed-form expressions for arbitrarily coupled space-division multiplexing (SDM) links are lacking. In this work, we expand upon previous literature by providing closed-form expressions modelling DRA and ISRS in common SDM fiber designs that support arbitrarily-coupled degenerate mode-groups, incorporate mode coupling, and accounting for inter-modal non-linear effects, showing excellent agreement with simulations. The derived formulas are then applied to representative scenarios, illustrating how distinct pump and fiber configurations influence gain and mode-dependent gain (MDG). Finally, we describe suggested routines for experimentally estimating the Raman response profiles of SDM fibers.

The role of macromolecular crowding in living systems is widely appreciated, but artificial crowders used to model these effects in vitro are often inadequately characterized. In this work, we examine density, viscosity, polymer self-diffusion and water diffusion in crowded dextran systems. Dextran viscosity and self-diffusion follow size-dependent trends, collectively described by universal functions of the overlap concentration corresponding to a Flory exponent of 0.44, characteristic of branched polymers. Viscosity increases with concentration as a power law, with a crossover from dilute to semi-dilute behaviors. Dextran self-diffusion decays exponentially: this can be interpreted in light of Rosenfeld's excess entropy scaling hypothesis. Water self-diffusivity and specific volume decrease with concentration, but show no dependence on polymer size. We show how these results can be used to construct the true volume fraction of crowders, which takes into account bound water. Overall, our findings showcase the power of polymer physics concepts in macromolecular crowding studies in vitro.

Over the past decade, holographic nanodiamond-polymer composite gratings have been developed and optimized as high-efficiency diffractive elements for very cold neutrons (VCN), for use as mirrors and beam splitters in a triple-Laue (LLL) interferometer. We report their optical characterization and, crucially, their neutron-optical performance, including diffraction efficiency and angular selectivity under VCN conditions. We further demonstrate their integration into a VCN interferometer. The layout of the interferometer and its first implementation at the beamline are described, highlighting practical considerations for long-term operation. We discuss avenues for performance improvement, in particular grating fabrication refinements. These results establish nanodiamond-polymer composite gratings as viable components for VCN interferometry and pave a way toward precision neutron phase measurements in the very cold regime.

We consider Brillouin scattering on short-lived phonon modes, such that the relative Brillouin shift between propagating and scattered waves is smaller than the relative width of phonon modes. In this case one phonon mode facilitates scattering between many pairs of optical modes. We show that in this limit two phonon modes are sufficient for cascade Brillouin scattering (one forward propagating wave and one counter propagating wave), and that the cascade behavior is qualitatively different from the cascade in conventional Brillouin systems with distinct phonon modes for each optical mode pair. In particular, our results show that there is a pump threshold above which many optical modes become excited simultaneously, as opposed to a cascade gradually building up. The resulting cascade scattering can be exploited for frequency comb generation with uniform amplitudes and without the need for anomalous dispersion in the medium.

Phase change materials provide a powerful platform for dynamically modulating optical responses in nanophotonic systems. While plasmonic metasurfaces have been widely employed to enhance photocatalytic efficiency and promote particular light-driven reactions, active and dynamical control over reaction pathways within a single device remains challenging. Here, we report a phase-induced tunable metasurface that tailors photoexcited electron populations through mode hybridization, enabling selective control over the reactivity of light-driven chemical processes. By exploiting thermally induced refractive-index switching in a Sb2S3 cavity, the plasmonic resonance strength of Au nanodisks is actively tuned via cavity-plasmon hybridization. This reconfiguration modulates the product yield of methylene blue degradation by a factor of 2.4, suppressing to 0.45 in the crystalline phase and enhancing to 1.09 in the amorphous phase. Importantly, this reconfigurable platform enables dynamic control of the reaction yield using a single metasurface architecture under identical illumination conditions. Our approach establishes a dynamically programmable light-driven reaction platform capable of precisely manipulating reaction reactivity, offering new opportunities for selective photocatalysis in complex multibranch reaction systems.

Conventional polarimetry, including schemes leveraging entangled light, characterizes optical samples through linear transformations of polarization states. We introduce a two-photon probing approach in which both photons of an entangled pair interact with the same depolarizing medium simultaneously. In this regime, the transformation of the two-photon polarization correlations becomes quadratic in the Mueller matrix, enabling access to second-order polarization information beyond conventional polarimetry. We develop a theoretical framework linking the Mueller matrix to the evolution of the two-photon polarization correlation tensor and show that depolarization induces quadratic degradation of entanglement and state purity. Experiments using polarization-entangled photon pairs transmitted through controlled scattering media confirm the predicted response and reveal enhanced sensitivity to polarization scrambling compared with single-photon probing. These results establish two-photon probing as a higher-order quantum polarimetric modality for characterizing polarization channels.

The 2025 UK National Astronomy Meeting (NAM) in Durham played host to a session titled "Unseen Astronomy", involving a variety of astronomy researchers in diverse fields. This unique meeting focussed on a number of novel projects exploring alternatives to purely visual means of display in Astronomy, encompassing spheres of education, communication and research, and straddling both accessible and general use applications. The successful inclusion of such a session at a major conference reflects the explosion of interest in multimodal astronomy in recent years, and hints at its transformative potential. Here, I aim to outline and motivate the topic of multi-modal science and consider its exciting potential. I will discuss this in the context of our own work in the area, the community building being undertaken to bring together researchers considering multi-modality, and efforts to impact astronomy at large.

We explore a quantitative approach to emerging technological sovereignty and geoeconomic power by assessing the relative positioning of countries with economic complexity methods applied to the structure of national venture-capital (VC) portfolios and their associated Revealed Venture Advantage (RVA) metrics. Using Crunchbase firm- and deal-level data, we map venture-backed startups to 18 emerging technology domains via a probabilistic multi-label large-language-model classifier, and construct an RVA-based country-technology specialization matrix for the 17 countries with the highest aggregate VC funding. From this matrix, we derive two eigenvector-based measures: a Geoeconomic Complexity Index (GCI) that ranks countries by the composition of their venture specializations, and an Emerging Technology Geoeconomic Complexity Index (ETGCI) that ranks domains by the extent to which specialization is concentrated among high-GCI countries. Empirically, Cloud Computing, Cybersecurity Tools, and Medtech exhibit the highest ETGCI values, reflecting concentration of specialization in a small set of leading countries. The United States and Israel consistently occupy a marked "high-diversity/low-ubiquity" position and lead the GCI ranking, followed by China, France, Japan, and Germany; both country and domain rankings are stable from 2021-2024. Finally, relatedness-based simulations identify, when it exists, for each country the Simplest Single Sovereignty Enhancing Technology (SSSET), i.e., the most feasible single new technological direction associated with the largest expected improvement in relative geoeconomic positioning.

Explicit Runge-Kutta (RK) integration of hyperbolic initial-boundary value problems with time-dependent Dirichlet data often displays order reduction: the observed convergence order falls below the nominal order because the stage structure interacts with asymmetric near-boundary spatial closures. This paper develops a purely spatial remedy that preserves the time integrator while redesigning only the first two boundary-adjacent derivative operators. For an arbitrary explicit $s$-stage RK method applied to linear advection, the one-step truncation error at the boundary-adjacent nodes is shown to admit a tableau-dependent decomposition whose cancellation yields explicit algebraic conditions on the boundary weights. A solvability coefficient $R(\mathbf{b},\mathbf{c},A)$ determines whether a spatial compensation mechanism exists; the result is specialised to SSP-RK3, for which closed-form conditions are derived. Constrained differential evolution then identifies 5-point closures that, coupled to a 5th-order upwind interior stencil, recover third-order convergence from the degraded second-order behaviour of classical Taylor closures. A stability-aware variant augments the optimisation with an eigenvalue penalty, exposing the trade-off between order recovery and CFL robustness. Validation covers linear advection, manufactured-solution Burgers flow, and dimensionally split two-dimensional advection. The analysis clarifies why weak-stage-order temporal fixes do not resolve the finite-difference boundary problem, and indicates how the framework extends to non-uniform meshes.

We propose ultrafast all-optical switching exploiting the bistability between a spatially uniform photon superfluid and a spontaneously ordered supersolid in a driven-dissipative microcavity. The key ingredient is a tunable nonlocal photon--photon interaction engineered by embedding a high-mobility two-dimensional electron gas (2DEG) inside the cavity. A drift current displaces the Fermi disk, imparting a negative region to the Lindhard interaction kernel at finite wavevectors and triggering a roton instability. The resulting bistable $S$-curve supports a write--hold--erase protocol in which short optical pulses toggle the system between branches with a switching contrast of order 120~dB. The hysteretic ON state persists under a constant sub-threshold drive after the write pulse is removed, realizing an all-optical bistable memory. Since the photon field couples additively to each embedded quantum well, stacking layers with distinct drift angles allows the roton profile to be engineered with higher-order symmetries, imprinting richer spatial order on the supersolid and enabling nonbinary generalizations of the switch. Operating in the ultrafast, sub-fJ regime, this platform outperforms most existing all-optical switches in contrast and reconfigurability.

This paper provides an introduction to the theory of semiconductor laser diodes, with special focus on their noise properties. It may be considered an additional chapter to the textbook [1]. As such, it will also refer to equations in that book.

Phonon polaritons in van der Waals crystals enable exceptional light confinement and control over low-loss nanolight propagation. The polariton wavelength can be controlled by the crystal geometry, isotopic composition, or surrounding environment -- for which substrate engineering is particularly effective. However, existing approaches of substrate nanopatterning are binary and offer limited leverage. Here, we demonstrate local control over the wavelength of phonon polaritons in hexagonal boron nitride by employing a sinusoidally corrugated gold surface to smoothly vary the gap between the van der Waals crystal and metallic substrate. The nonuniform gap provides a continuous and nearly threefold local variation of the polariton wavelength across the structure, verified by near-field optical microscopy. Our platform further enables lateral nanofocusing by gradually compressing and decompressing the wavelength of propagating polaritons by a factor of around 2.5 achieved solely through substrate geometry, consistent with our local control experiments and theoretical calculations. Our results push the boundaries of substrate engineering and showcase a powerful method for precise and local tailoring of polaritonic modes.

Overcoming the limitations of current nanofabrication techniques to achieve nanoscale feature sizes is essential for achieving new regimes of light-matter interactions at extreme frequencies and length scales. Here, we demonstrate a scalable nanofabrication platform capable of producing in-plane feature sizes down to 1.75 nm, pushing the boundaries of current top-down nanofabrication techniques. Using precise thickness control of atomic layer deposition (ALD) and employing widely spaced oxide nanofins, we transform conventional ALD into a surface structuring method that produces nanolaminates with sub-10 nm periodicities over large areas. The resulting nanostructures can be used as a one-dimensional gate array to control charge carriers in two-dimensional materials. As an initial demonstration, we integrate the platform with graphene and perform electron transport measurements. In the presence of the gate array enabled by the nanolaminate, we observe satellite Dirac peaks consistent with band-structure modulation, suggestive of quantum-confinement effects. Our platform paves the way for exploring previously inaccessible regimes of nanoscale light-matter interactions, holding significant promise for applications in short wavelength optics, electronics, and polaritonics.

Budding-cell packings jam before all buds are mechanically constrained. The post-jamming state is therefore set by both the pressure $P$ and the fraction $u$ of buds that remain unconstrained. We develop a mean-field theory for this regime. A modified Maxwell count gives the post-jamming coordination, a depletion law for the unconstrained buds predicts the crossover density $φ_2$ at which that reservoir is exhausted, and a flux-partition argument explains why strong growth feedback can markedly increase rigidity while generating little internal pressure.

The Lucas sequences are integers defined by a homogeneous recurrence relation. They include the well-known Fibonacci numbers, which appear abundantly in nature. The complementary Lucas numbers, defined by the same recurrence relation, are less well-known. In this work, we show that a special case of such complementary Lucas sequences can be observed on the same physical platform. It consists of a gain-and-loss-modulated non-Hermitian reservoir bridging two mirror-symmetric systems, which manifests the Lucas sequences in linearly localized edge states and a constant-intensity mode, respectively.

High-resolution three-dimensional imaging under photon-starved conditions remains challenging. Here, we demonstrate a high-resolution long-range 3D single-photon imaging system based on a digital micromirror device (DMD) and a compact 64 multiply 64 single-photon avalanche diode (SPAD) array. By combining high-resolution spatial modulation with parallel time-resolved detection, the system extends the effective spatial sampling beyond the native detector format while preserving depth information through time-of-flight measurement. In outdoor experiments at a stand-off distance of 670 m, we achieved 3D reconstruction of natural targets with an effective spatial resolution of 256 multiply 256. These results validate the proposed method as an effective approach for high-resolution long-range 3D single-photon imaging using compact SPAD arrays.

We propose a genetic algorithm powered evolution (GAPE) method to create deep learning solutions for energy and position estimation for reactor antineutrino interactions in the Precision Reactor Oscillation and Spectrum Experiment (PROSPECT) at the highly enriched High Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory. We also apply GAPE to create classification models to distinguish signatures of inverse beta decay (IBD) interactions of reactor antineutrinos from common background types. The GAPE method can also be adopted for optimization of other types of problems that utilize machine learning (ML) models for particle physics applications. When applied in the PROSPECT context, we find that the models selected by GAPE can, in some cases, outperform the traditional models previously used for PROSPECT data analysis. In particular, when benchmarked against conventional PROSPECT neutrino identification pathways using the same underlying information, the classifier offers the promise of improving the signal-to-background ratio by nearly 2.8 times. Performance biases uncovered during initial IBD classifier validation were primarily caused by differences in time-dependent response between background and signal training datasets. Biases were effectively mitigated through a data-period-specific training regimen, offering a pathway towards realizing an unbiased IBD signal classifier for future reactor neutrino datasets.

Stationary solutions asymptoting to nonlinear plane waves of the nonlinear Schrödinger equation with a PT-symmetric, complex linear potential are characterized. The potential includes both a spatially varying gain-loss profile and a repulsive real part, generated by a Wadati potential function,that support the existence of homoclinic and heteroclinic solutions that asymptote to the same or different, respectively, nonlinear plane waves in the far field. Asymptotic analysis and numerical simulations are used to examine solution existence, bifurcations, and structure. Such solutions play an important role in resonant nonlinear wave generation of dispersive media with localized gain and loss.

Many bipartite networks exhibit hierarchical community structure, but existing community detection methods are not well-suited for detecting hierarchy. They also do not effectively handle weighted bipartite networks. In this work, we introduce a novel modularity-based objective function, called the generalized bipartite modularity density, Qbg, specifically designed for hierarchical community detection in bipartite systems. The framework incorporates a tunable resolution parameter that enables systematic exploration of community structure across multiple scales. It leverages resolution-limit behavior in bipartite networks as a tool to uncover hierarchical organization without projecting the network or altering its intrinsic bipartite topology. We evaluate the method using a hierarchical synthetic bipartite benchmark and apply it to two empirical networks. In all cases, Qbg recovers established mesoscale structure while revealing additional hierarchical and fine-scale organization beyond that detected by conventional bipartite approaches. These results establish Qbg as a flexible, interpretable, and resolution-aware framework for hierarchical community detection in bipartite networks.

Periodic boundary conditions (PBCs) for computing magnetic fields in repeating magnetic structures, e.g. in micromagnetic simulations, are typically imposed using the quasi periodic macrogeometry approach, where many copies of the simulated domain are introduced. This can be computationally problematic, especially if the simulated domain is incommensurate with the desired sample shape. In this work, we present a formal proof that for sufficiently large magnetic samples, only the average magnetisation gives non-negligible shape effects. Using this insight, we develop a simple, computationally efficient modification of existing implementations which incorporates shape effects in PBC methods.

The O-band (1260-1360 nm), located near the minimum of chromatic dispersion of standard single-mode fiber, is the transmission window of major interest and importance for short-reach data-center interconnects. However, full capacity offered by this spectral band is yet to be unlocked, due to limited availability of scalable multi-wavelength, high-power, low noise O-band light engines. While Kerr microcombs in CMOS-compatible silicon nitride resonators provide mutually coherent wavelength channels with precise spacing and chip-scale footprints, their practical deployment in the O-band has been hindered by limited pump laser power, insufficient per-line power and the lack of flat, wideband amplification technologies to uniformly boost multiple coherent carriers. Here we demonstrate a high-power O-band soliton microcomb architecture that overcomes this bottleneck by combining self-injection-locked (SIL) operation in a Silicon Nitride microring with a single-stage bismuth-doped phosphosilicate fiber amplifier designed for wideband, flat-top gain. The SIL microcomb operates with an 834 GHz free spectral range and spans over 1050-1650 nm. The amplifier simultaneously boosts 21 O-band lines across 100 nm to powers exceeding 0 dBm per carrier without gain flattening or external equalization, while preserving low-noise characteristics. We validate each amplified microcomb line as a carrier across the entire O-band using dual-polarization 32 GBaud 64-QAM coherent transmission. This approach establishes a practical route towards high-power, broadband O-band microcomb engines for next-generation data-center interconnects and scalable photonic systems.

This paper presents the first experimental characterisation of combined hydrogen-temperature effects in 316plus (EN 1.4420), a new austenitic stainless steel for liquid hydrogen (LH2) storage. Uniaxial tensile tests were conducted at room temperature (RT), 77 K and 20 K on uncharged and hydrogen-precharged specimens, complemented by fractography and EBSD-based quantification of strain-induced martensite (SIM). 316plus exhibited cryogenic strengthening at 77 K and 20 K by enhanced SIM formation. Hydrogen did not influence strength at RT or 77 K and caused a modest decrease (~10%) at 20 K, keeping 316plus at the upper bound of cryogenic strength for 316L. The presence of hydrogen resulted in significant reductions in ductility at all temperatures, being most severe at 77 and 20K (~40-50%). Hydrogen suppressed SIM at 20 K, but SIM fraction did not correlate with ductility reduction. Despite the combined effect of temperature and hydrogen, 316plus retained notable ductility (reduction in area ~30%).

Temporally multimode squeezed states have been a topic of recent interest due to their applications in quantum communication, information processing, and sensing. Characterizing the mode shapes is crucial for effectively manipulating these states, but current mode shape and state characterization techniques necessitate constraining assumptions and complicated experimental setups. Here, we propose a characterization technique that simultaneously recovers the complex temporal mode shapes and quadrature variances of ultrafast multimode squeezed states based on frequency resolved optical gating (FROG) using an optical parametric amplifier (OPA). FROG is a promising tool for quantum state characterization due to its flexibility of implementation and high temporal resolution. Using an OPA as the nonlinear process in FROG has the benefit of amplifying weak quantum states to a detectable level while preserving quantum information. Numerical simulations demonstrate the recovery of the mode shapes and levels of squeezing and anti-squeezing of ultrafast multimode squeezed states. This scheme offers a practical experimental approach to measuring arbitrary temporal mode shapes and characterizing large-scale multimode ultrafast Gaussian quantum states.

Many-body topological quantum states host exotic quantum phenomena and lie at the forefront of developing next-generation quantum technologies. Recently emerged neural network wavefunction methods have established themselves as a powerful computational framework for accessing these states, enabling the variational machine learning calculation of the system's ground state wavefunction. However, reliable computation of topological invariants remains an open challenge when the whole deterministic energy spectrum is not available. In this work, we introduce a robust approach to determining topological invariant based on simulating the charge pumping process, by monitoring the response of polarization upon flux insertion. By applying this method, we accurately extract the Chern numbers for Abelian fractional Chern insulators. Our approach also enables the first neural-network-wavefunction-based identification of anomalous composite Fermi liquid states. Our work resolves a key bottleneck in applying neural network wavefunctions to correlated topological matter, and the method proposed is also generally applicable to other many-body approaches, thereby opening up new avenues for future research in this field.

Cavitation-assisted Ag paste is a promising route for fine-line, low-silver metallization in silicon solar cells because it improves paste dispersion, extends shelf life, and reduces Ag consumption, but matching the contact performance of commercial pastes remains a challenge. Here, cavitated paste was evaluated on PERC solar cells at peak firing temperatures of 720, 740, 750, and 762 C, with and without laser-enhanced contact optimization (LECO). The results show a clear firing window: 720 and 740 °C produced high series resistance and reduced fill factor, 750 C gave the best pre-LECO performance, and 762 C showed additional electrical limitations with only limited LECO benefit. LECO selectively recovered the under-activated states, increasing fill factor from 76.8 to 80.2% at 720 C and from 76.7 to 79.8% at 740 C. Electroluminescence and conductive AFM further indicated improved current collection and stronger localized conduction after LECO. These results show that cavitated paste performance is governed primarily by a shifted contact-formation window, and that firing optimization combined with LECO provides a practical route to retain the fine-line advantage while improving electrical performance.

Non-Hermitian photonics provides a framework to engineer the gain and loss of optical modes in open systems, enabling control of their spectral and dynamical properties. In particular, the ability to dynamically tune modal losses offers a route to implement functionalities traditionally relying on cavity Q-factor modulation, such as Q-switching, within nanophotonic platforms. Here, we demonstrate the generation of short optical pulses in a pair of phase-coupled photonic crystal nanolasers exploiting non-Hermitian coupling. Two waveguide-coupled nanocavities are operated below their individual lasing thresholds and subjected to asymmetric optical pumping, such that a transient carrier-induced detuning modifies the interference conditions between them. This dynamically controls the gain and loss of the collective modes, and, upon crossing a resonance condition, leads to the rapid release of stored carrier energy as an optical pulse. A rate-equation model captures the interplay between carrier dynamics and modal coupling and reproduces the observed behavior. Experiments performed on an indium phosphide platform show pulse generation from cavities that do not lase efficiently on their own in continuous-wave operation, with temporal characteristics governed by carrier dynamics. These results indicate that non-Hermitian coupling can be used to control the effective cavity losses in time, providing a route to pulse generation in integrated photonic systems.

The rapid rise of artificial intelligence, and in-memory computing has reinvigorated research on scalable, energy-efficient, and reconfigurable photonic hardware. Non-volatile phase-change materials (PCMs) are attractive, as they offer large refractive index contrast, wavelength-scale footprints, and zero static power consumption. However, current PCM-based electrically controlled photonic devices are plagued by high insertion loss and low endurance. One prevalent hypothesis for these material limitations come from electromagnetic scattering in the interface and large programming volumes, respectively. Here, we validate this hypothesis by showing that nano-structuring of PCM minimizes optical loss and enhances the endurance. By tapering both ends of a wide bandgap PCM Sb2Se3 segment on a silicon waveguide, we suppressed the insertion loss by ~94% (resulting in a loss of ~0.1 dB per π phase shift). Through combining tapering and segmentation, we achieved high optical modulation amplitude (~70%), low loss (~0.5 dB per π phase shift), low-voltage (< 5V) actuation, and record high endurance greater than 100 million cycles. This work showcases the substantial advantage of nanopatterning PCMs to attain low loss and high cyclability.

Developing optical systems for free-space applications requires simulation tools that accurately capture turbulence-induced wavefront distortions and support gradient-based optimization. Here we introduce TurPy, a GPU-accelerated, fully differentiable wave optics turbulence simulator to bridge high fidelity simulation with end-to-end optical system design. TurPy incorporates subharmonic phase screen generation, autoregressive temporal evolution, and an automated screen placement routine balancing Fourier aliasing constraints and weak-turbulence approximations into a unified, user-ready framework. Because TurPy's phase screen generation is parameterized through a media-specific power spectral density, the framework extends to atmospheric, oceanic, and biological propagation environments with minimal modification. We validate TurPy against established atmospheric turbulence theory by matching 2nd order Gaussian beam broadening and 4th order plane wave scintillation to closed-form models with 98% accuracy across weak to strong turbulence regimes, requiring only the medium's refractive index structure constant and power spectral density as inputs. To demonstrate TurPy as a gradient-based training platform, we optimize a dual-domain diffractive deep neural network (D2NN) in a two-mask dual-domain architecture to recover a Gaussian beam from a weakly turbulent path and achieving over 20x reduction in scintillation relative to an uncompensated receiver in simulation. TurPy is released as an open-source package to support synthetic data generation, turbulence-informed algorithm development, and the end-to-end design of optical platforms operating in turbulent environments.

Conventional multi-image super-resolution (MISR) methods, such as burst and video SR, rely on sequential frames from a single camera. Consequently, they suffer from complex image degradation and severe occlusion, increasing the difficulty of accurate image restoration. In contrast, multi-aperture camera-array imaging captures spatially distributed views with sampling offsets forming a stable disk-like distribution, which enhances the non-redundancy of observed data. Existing MISR algorithms fail to fully exploit these unique properties. Supervised MISR methods tend to overfit the degradation patterns in training data, and current self-supervised learning (SSL) techniques struggle to recover fine-grained details. To address these issues, this paper thoroughly investigates the strengths, limitations and applicability boundaries of multi-image-to-single-image (Multi-to-Single) and multi-image-to-multi-image (Multi-to-Multi) SSL methods. We propose the Multi-to-Single-Guided Multi-to-Multi SSL framework that combines the advantages of Multi-to-Single and Multi-to-Multi to generate visually appealing and high-fidelity images rich in texture details. The Multi-to-Single-Guided Multi-to-Multi SSL framework provides a new paradigm for integrating deep neural network with classical physics-based variational methods. To enhance the ability of MISR network to recover high-frequency details from aliased artifacts, this paper proposes a novel camera-array SR network called dual Transformer suitable for SSL. Experiments on synthetic and real-world datasets demonstrate the superiority of the proposed method.

Soil moisture is a critical variable for managing irrigation, improving crop yield, and understanding field-scale hydrology. Radars mounted on unmanned aerial vehicles (UAVs) offer a promising means to monitor soil moisture over large fields with flexible, high-resolution coverage. However, during the growing season, canopy scattering and soil reflections become strongly coupled in the radar measurement. These coupled effects vary with crop structure or flight altitude, complicating the retrieval of soil moisture. To overcome this challenge, we present GreenScatter, a physics-based soil moisture retrieval framework for nadir-looking wideband UAV radars. GreenScatter introduces a microwave radiative transfer model that explicitly captures the dominant electromagnetic interactions between vegetation and soil, enabling accurate modeling of coherent ground backscatter through canopy. In parallel, it develops a radar cross-section (RCS) estimation method that transforms time-domain radar signals into calibrated wideband RCS spectra, isolating soil reflections while compensating for hardware and waveform effects. Together, these components enable robust soil moisture estimation through vegetation across varying canopy conditions and UAV configurations. Field experiments across multiple corn and soybean sites demonstrate consistent retrieval with an average volumetric water content (VWC) error of 4.49%.

Multiple access techniques are vital for 5G and beyond. While Orthogonal Frequency Division Multiple Access (OFDMA) is standard, its high peak-to-average power ratio (PAPR) reduces energy efficiency in uplink transmissions. This paper presents Periodic OFDMA (P-OFDMA), a novel multiple access scheme with reduced PAPR and computational complexity. By assigning subcarriers in a periodic pattern across the entire frequency band, P-OFDMA enhances frequency diversity and simplifies allocation. We also introduce two precoded variants: P-OFDMA-DCT and P-OFDMA-DFT. Comprehensive simulations comparing P-OFDMA with OFDMA and SC-FDMA show that P-OFDMA-DFT consistently achieves the lowest PAPR. Furthermore, the standard P-OFDMA scheme outperforms SC-FDMA in PAPR for low subcarrier-per-user scenarios and achieves better bit error rate (BER) performance under high delay-spread conditions. Notably, P-OFDMA and its variants reduce transmitter-side processing by up to an eightfold factor compared to SC-FDMA, greatly benefiting low-complexity uplink devices. Although receiver complexity increases, the overall system processing load decreases, yielding improved energy efficiency. Thus, P-OFDMA offers a robust, energy-efficient uplink solution for future wireless networks.

Beyond-diagonal reconfigurable intelligent surfaces (BD-RISs) significantly improve wireless performance by allowing tunable interconnections among elements, but their design in multiple-input multiple-output (MIMO) systems has so far relied on complex iterative algorithms or suboptimal approximations. This work introduces a simple yet powerful approach: instead of directly maximizing the achievable rate, we maximize the absolute value of the determinant of the equivalent MIMO channel. We derive a closed-form symmetric unitary scattering matrix whose rank is exactly twice the channel's degrees of freedom ($2r$). Remarkably, this low-rank solution achieves the same determinant value as the optimal unitary BD-RIS. Using log-majorization theory, we prove that the rate loss relative to the optimal unitary BD-RIS vanishes at high signal-to-noise ratio (SNR) or when the number of BD-RIS elements becomes large. Moreover, the proposed solution can be perfectly implemented using a $q$-stem BD-RIS architecture with only $q=2r-1$ stems, requiring a minimum number of reconfigurable circuits. The resulting Max-Det solution is orders of magnitude faster to compute than existing iterative methods while achieving near-optimal rates in practical scenarios. This makes high-performance BD-RIS deployment feasible even with large surfaces and limited computational resources.

This paper presents a generalized circuit framework for constructing Shih-type fractionalizations of unitary operators of dyadic order, i.e., operators $U$ satisfying $U^{2^n}=I$. Building upon the architecture of the quantum fractional Fourier transform (QFrFT), we show that fractionalization can be implemented coherently as a weighted superposition of integer powers, $\sum_k c_k(α)U^k$, where the coefficients are generated through an ancilla-domain quantum Fourier transform and a diagonal phase modulation. Under the assumption that controlled implementations of the required powers of $U$ are available, the resulting circuit yields a parameterized family of operators that interpolates the integer powers of $U$ and satisfies the additive property of fractional transforms. As concrete applications, we derive explicit quantum circuit realizations of the quantum fractional Hartley transform (QFrHT) and of the fractional cosine-transform families associated with Types~I and~IV. These constructions demonstrate the versatility of the proposed dyadic-order fractionalization framework for structured operators arising in quantum signal processing.

This paper presents a Semantic Feature Multiple Access (SFMA) framework for multi-user semantic communication in downlink wireless systems. By extending SwinJSCC to a two-user superimposition paradigm, SFMA enables simultaneous semantic transmission to multiple users over shared time-frequency resources. A key innovation is the Cross-User Attention (CUA) module, which facilitates controlled semantic feature exchange between paired users by leveraging inter-image similarity while mitigating interference. We formulate a joint user pairing and resource allocation problem to minimize global semantic distortion under constraints on bandwidth, end-to-end latency, and energy. This mixed-integer non-convex problem is decomposed into a Minimum-Weight Perfect Matching (MWPM) sub-problem and a convex bandwidth allocation feasibility check, with semi-closed-form bandwidth bounds derived from a strictly concave rate expression. A polynomial-time algorithm based on Blossom matching and bisection search is proposed. Extensive simulations on ImageNet-100 show that SFMA significantly improves reconstruction quality across pairing modes, and the proposed optimization effectively reduces overall distortion while satisfying physical-layer constraints.

Integrated learning and communication (ILAC) unifies learned transceivers with radio resource management, where semantic feature multiple access (SFMA) enables paired users to superpose their learned representations over shared time-frequency resources. Unlike conventional multiple access schemes, SFMA interference arises in the learned feature space and depends jointly on the user pair, the transmit power, and the compression ratio. This coupling ties binary pairing decisions to continuous resource variables, yielding a mixed-integer non-convex optimization problem. To address this problem, we first propose similarity-conditioned SFMA (SC-SFMA), a Swin Transformer-based transceiver whose dual-conditioned similarity modulator (DC-SimM) gates cross-user feature fusion according to the inter-user semantic similarity. We then characterize the resulting pair-dependent interference by a bivariate logistic function parameterized by transmit power and compression ratio, thereby bridging the learned transceiver with network-level optimization. On this basis, we formulate a sum-rate maximization problem subject to per-user distortion, latency, energy, power, and bandwidth constraints. To solve this problem, we develop a three-block alternating optimization algorithm that integrates dual-decomposition-assisted compression ratio allocation, trust-region successive convex approximation (SCA) for joint power-bandwidth optimization, and dynamic feasible graph-based user pairing. Simulation results show that SC-SFMA achieves considerable peak signal-to-noise ratio (PSNR) and multi-scale structural similarity index measure (MS-SSIM) gains over deep joint source-channel coding (JSCC) and separation-based baselines. The proposed optimization framework attains significant sum rate improvements over conventional multiple access baselines.

Liquid crystal (LC) is a promising hardware solution for implementing large RISs, as it is cost-effective, energy efficient, scalable, and capable of providing continuous phase shifts with low power consumption. However, the phase shift response of LC-based RISs is inherently frequency dependent. If unaddressed, this characteristic leads to performance degradation, particularly in wideband scenarios. This issue is especially critical in secure communication applications, where minor phase shift variations across elements can result in considerable information leakage. This paper addresses these frequency-induced variations by developing a physics-based model for an LC unit cell across varying frequencies and proposing a novel phase shift design framework that maximizes secure communication across all subcarriers. Given the large number of elements in millimeter wave (mmWave) LC-RISs, acquiring full channel state information (CSI) is often impractical. Therefore, we optimize the phase shifts based solely on the locations of the legitimate mobile users (MUs) and potential eavesdroppers. Rather than targeting a single user point, the RIS is designed to illuminate a broader area. This approach enhances communication reliability for the MUs and mitigates performance degradation caused by location estimation errors. To solve the problem, we introduce both a semi-definite programming (SDP)-based solution and a low complexity heuristic method. While the SDP-based approach yields superior performance, it incurs higher computational complexity. Conversely, the scalable method exhibits a much slower scaling of complexity, which makes it highly suitable for extremely large RISs. Simulation results demonstrate that both algorithms improve the secrecy rate compared to baseline methods. Finally, the proposed design is validated through experimental evaluations on an LC RIS setup.

Flexible-geometry arrays based on movable antennas have shown considerable potential for improving wireless communication performance. In this letter, we investigate a multiuser multiple-input single-output (MU-MISO) downlink secure communication system aided by a flexible cylindrical array (FCLA) and artificial noise (AN), where each antenna element rotates along circular tracks while the circular slices move along a vertical axis. To guarantee transmission security, we aim to maximize the achievable sum rate at multiple legitimate information receivers by jointly optimizing transmit beamforming, AN covariance matrix, and antenna placement under secrecy constraints for an eavesdropper. While the resulting problem is intractable to solve, we develop a block coordinate descent (BCD)-based framework that combines the Lagrangian dual transform, tight semidefinite relaxation (SDR), and Nesterov-accelerated projected gradient descent (PGD). Numerical results show that the proposed algorithm converges rapidly and achieves significant sum-rate gains over benchmark schemes by exploiting the geometry flexibility of the array.

Bistatic backscatter communication requires strong illumination of a backscatter device (BD), while a spatially separated reader detects the weak modulated reflection. In practice, the resulting direct link interference (DLI) at the reader can dominate the received backscattered signal and limit detection performance. This paper experimentally investigates transmit beamforming that jointly maximizes BD illumination and suppresses DLI at the reader in a distributed multiple-input multiple-output setup. We compare phase-only maximum ratio transmission (PO-MRT) with the proposed direct-link suppression (DLS) scheme, which enforces a spatial null at the reader under per-antenna power constraints. Measurements using a phase-coherent 42-element ceiling array at 920 MHz show that DLS reduces the DLI at the target reader and improves the signal-to-interference ratio by up to 31 dB compared to PO-MRT.

Acquiring the channel state information from limited and noisy observations at pilot positions is critical for wireless multiple-input multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) systems. In this paper, we view this process as a conditional generative task in which the partial noisy channel estimates at the pilots are utilized as a ``prompt'' to guide the diffusion ``inpainting'' of the underlying channel. To this end, we resort to a general Conditional Diffusion Transformer (CDiT) framework with a well-designed network architecture and update rule. In particular, we design a dedicated embedding strategy to encode and adapt to different pilot patterns and noise levels, and utilize a special cross-attention mechanism to align the partial raw channel observations with the denoised channel at each time step of the generation process. This architecture effectively anchors the diffusion process, enabling the model to accurately recover full channel details from limited noisy observations. Comprehensive experimental results show that, the proposed approach achieves a performance gain of over 5 dB compared to the baselines under varying noise conditions, and provides robust channel acquisition even under a sparse pilot density of 1/32 without significant performance loss compared to the denser pilot cases. Moreover, it is capable of generating high-quality channel matrices within just 10 inference steps, effectively balancing estimation accuracy with computational efficiency and inference speed. Ablation studies demonstrate the rationality of the model design and the necessity of its modules.

Integrated sensing and communications (ISAC) has emerged as an intrinsic service of upcoming 6G wireless systems, enabling the reuse of communication signals for environmental sensing and supporting context-aware network functionalities. Meanwhile, the evolution of the wireless infrastructure toward distributed systems creates new opportunities for collaborative sensing from spatially separated nodes. Motivated by this trend, this work investigates a radio stripe aided ISAC system as a low-complexity implementation of a distributed system. We study the trade-off between achievable sum rate and sensing precision when downlink signals are used for target localization within the service area. By exploiting the architectural homogeneity of the radio stripes transceivers, each unit can be dynamically configured to operate in either communication or sensing mode. We formulate a targets localization problem considering the measurements of multiple sensing-communication configurations. Due to the large number of measurements and the continuity of the search space, we propose discretizing the service are and then solve the estimation problem in batches. The targets are finally estimated using a fusion strategy. Our results show that increasing the number devices and sensing APUs boosts sensing precision at the expense of degrading the sum rate. The latter remains constant for a given number of communication APUs regardless of their positions. Moreover, changing the number of antennas reveals a non-monotonic impact on sensing performance due to the trade-off between array gain and illumination uniformity.

This paper investigates multi-stream downlink precoding for massive multiple-input multiple-output low-Earthorbit satellite (SAT) communication systems. We adopt a delay and Doppler precompensation approach to achieve coherent transmission. Under this setting, we formulate a signal transmission model that incorporates the near-independent properties of inter-SAT interference and compensation errors. We then demonstrate that moving beyond single-stream transmission requires both multi-SAT cooperation and multi-antenna UTs. Based on this configuration and the established signal transmission model, we derive the first- and second-order statistical channel characteristics and utilize them to design locally optimal precoding algorithms for both total power constraint (TPC) and per-antenna power constraint (PAPC) conditions, which rely only on statistical channel state information (sCSI). In particular, the designed PAPC algorithm achieves linear complexity with respect to the number of antennas on the cooperative SATs. To reduce the computational complexity of the locally optimal precoder under TPC, we propose a low-complexity and robust precoding scheme optimized for both minimum mean squared error and sum-rate maximization objectives. Using majorization theory, we also provide a rigorous theoretical analysis of the optimal precoding structure under TPC. Moreover, the Lanczos algorithm is adopted to further reduce the complexity of the proposed robust designs. Simulation results show that when each SAT is equipped with a sufficiently large number of antennas, the proposed sCSI-based designs achieve performance comparable to that of instantaneous CSI-based designs.

In this paper we propose a new criterion for the Blind Source Separation (BSS) of antisparse bounded sources, based on the sum of the $\ell_\infty$-norm of the sources. Based on the observation that the mixing process of bounded sources with any mixing matrix with unitary Frobenius norm will increase the $\ell_\infty$-norm of the sources, unless it is the identity matrix, the minimization of the sum of the $\ell_\infty$-norm of the sources can be used for the estimation of a separation matrix. To that, a Principle Component Analysis technique followed by a Givens Rotations based optimization method can be used for the separation of independent bounded sources. Also, the Givens Rotations based optimization method can be used for the separation of correlated bounded sources mixed by a rotation matrix. We theoretically analyze the proposed criterion and assess its performance through numerical simulations involving three distinct types of bounded signals. Our theoretical and experimental findings underscore the efficacy of the $\ell_\infty$ norm as a suitable contrast function for antisparse bounded sources, showcasing its superior performance relative to a state-of-the-art algorithm.

Convolutional neural networks (CNNs) have emerged as a powerful tool for automatic modulation classification (AMC) by directly extracting discriminative features from raw in-phase and quadrature (I/Q) signals. However, deploying CNN-based AMC models on IoT devices remains challenging because of limited computational resources, energy constraints, and real-time processing requirements. Early-exit (EE) strategies alleviate this burden by allowing qualified samples to terminate inference at an EE branch. However, our empirical analysis reveals a critical limitation of existing confidence-based EE strategies: they predominantly select samples whose early and final predictions are correct and consistent, while failing to capture whether deeper inference can provide a tangible accuracy gain. To address this limitation, we propose BEACON, a Benefit-Aware Early-Exit framework for AMC via recoverability prediction. BEACON introduces a benefit-aware EE criterion that explicitly predicts recoverable errors, defined as instances where the final-exit branch corrects an initial early-branch misclassification. Using only short-branch observables, we design a lightweight benefit-aware predictor (LBAP) to implement this criterion, estimating the likelihood of such recoverable cases and triggering deeper inference only when an accuracy gain is expected. Extensive experiments on ResNet-18-based AMC models demonstrate that the proposed approach consistently outperforms state-of-the-art baselines, achieving a superior accuracy-computation tradeoff across diverse EE threshold settings and signal-to-noise ratio regimes. These findings validate the effectiveness of the benefit-aware criterion and its practicality for energy-efficient on-device AMC under stringent resource constraints.

Integrated Sensing and Communications (ISAC) enables trajectory sharing that enhances beamforming, resource allocation, and cooperative perception, yet raises fundamental privacy concerns under the General Data Protection Regulation (GDPR) data minimisation principle. This paper proposes a Fisher Information Density (FID)-constrained trajectory sharing framework that enforces a local lower bound on estimation uncertainty, providing hard, quantifiable privacy guarantees by construction. Unlike fixed-noise approaches, the proposed method bounds the Privacy Leak Ratio (PLR) regardless of sensing power or adversarial post-processing, ensuring that no trajectory segment can be reconstructed beyond a prescribed accuracy threshold. Simulations on the OpenTraj dataset demonstrate that the framework keeps the average PLR below 20-25% and the maximum leakage segment duration under 2-2.5 s, while preserving data utility for downstream tasks such as movement prediction. The resulting criterion is interpretable, model-agnostic, and compatible with GDPR-compliant ISAC system design.

Reliable positioning is essential for Uncrewed Aerial Vehicles (UAVs) in safety-critical urban operations, yet achieving sub-meter accuracy under stringent latency constraints remains challenging. While 3rd Generation Partnership Project (3GPP) specifies repeated Positioning Reference Signals (PRS) transmissions for accurate Time Difference of Arrival (TDoA) measurements, denoising techniques specifically tailored for extremely limited measurement sequences within 3GPP frameworks remain underexplored. We propose Adaptive Gain Exponential Smoother (AGES), a lightweight filter combining exponentially weighted averaging with adaptive gains informed by 3GPP measurement quality reports. Simulations demonstrate AGES achieves 30-40% reduction in positioning error with only 3-5 repeated measurements while maintaining Fifth Generation New Radio (5G-NR) infrastructure compatibility.

Wideband orthogonal frequency-division multiplexing (OFDM) over extremely large-scale MIMO (XL-MIMO) arrays in the near-field Fresnel regime suffers from a coupled beam-squint and wavefront-curvature effect that renders single-frequency covariance models severely biased: the per-subcarrier compressed covariance diverges from the center-frequency model by 64\% at $B = 100$~MHz and by 177\% at $B = 400$~MHz. We derive the wideband compressed-domain Cramér--Rao bound (CRB) for hybrid analog--digital architectures and decompose the Fisher information gain into a dominant data-diversity term that scales as $10\log_{10}K_s$~dB and a secondary geometric-diversity term arising from frequency-dependent curvature. At 28~GHz with $M = 256$ antennas, $N_\mathrm{RF} = 16$ RF chains, and $K_s = 512$ subcarriers, wideband processing yields $+27.8$~dB of CRB improvement at $B = 400$~MHz, of which $+0.7$~dB is attributable to geometric diversity.

Let $f:\mathbb{R}^n\to\mathbb{R}$ be an unknown object, and suppose the observations are tomographic projections of randomly rotated copies of $f$ of the form $Y = P(R\cdot f)$, where $R$ is Haar-uniform in $\mathrm{SO}(n)$ and $P$ is the projection onto an $m$-dimensional subspace, so that $Y:\mathbb{R}^m\to\mathbb{R}$. We prove that, whenever $d\le m$, the $d$-th order moment of the projected data determines the full $d$-th order Haar-orbit moment of $f$, independently of the ambient dimension $n$. We further provide an explicit algorithmic procedure for recovering the latter from the former. As a consequence, any identifiability result for the unprojected model based on $d$-th order group-invariant moment extends directly to the tomographic setting at the same moment order. In particular, for $n=3$, $m=2$, and $d=2$, our result recovers a classical result in the cryo-EM literature: the covariance of the 2D projection images determines the second order rotationally invariant moment of the underlying 3D object.

Integrated sensing and communication (ISAC) is a key enabler of 6G, supporting environment-aware services. A fundamental sensing task in this setting is reliable multi-target detection and tracking. This paper proposes a temporal graph neural network (TGNN)-based tracking method that exploits delay and Doppler information from the wireless channel. The delay-Doppler map is modeled as a sequence of graphs, and tracking is formulated as a temporal node classification problem, enabling joint clustering and data association of dynamic targets. Using ray-tracing-based channel outputs as ground truth, the method is evaluated across multiple scenes with varying target positions, velocities, and trajectories and is compared with a Kalman filter baseline. Results demonstrate reduced normalized mean squared error (NMSE) in delay and Doppler, leading to more accurate multi-target tracking.

Millimeter-wave (mmWave) communication enables high data rates through large bandwidths and highly directional beamforming, but its sensitivity to blockage and mobility makes reliable beam alignment a central challenge. Limited-probing beam management is a fundamental problem in codebook-based mmWave systems, where only a small subset of beams can be evaluated simultaneously, and the serving decision is restricted to the probed set. Under mobility and noisy feedback, this leads to a sequential and partially observable decision problem in which performance depends critically on the quality of the proposed beam candidates. In this paper, we consider limited-probing beam management and develop a history-conditioned discrete denoising diffusion probabilistic model for beam candidate generation. The proposed method learns from logged probing histories a conditional distribution over promising beam indices, which is then used to construct probing candidates online. Numerical analysis shows that the proposed approach consistently achieves better signal-to-noise ratio, beam-miss probability, and conditional probe regret under tight probing budgets compared with strong learning-based and discriminative baselines. The gains are especially pronounced in low-probing regimes, where accurate candidate generation is most critical.

This work explores the potential of integrating an Intelligent Transmissive Surface (ITS) into an antenna array to improve beamforming performance. We show that integrating a moderate number of passive refractive elements into a small antenna array can significantly improve the Weighted Sum Rate (WSR). We investigate the optimization of the WSR under two distinct operational constraints: a Radiated Power (RP) constraint and a Transmitted Power (TP) constraint. Our analysis reveals that the choice between these constraints significantly impacts the design parameters of the ITS-aided array. By contrasting these approaches, we explore critical design and material parameters, including the array geometry, surface loss, and illumination strategies.

In current MIMO mobile communication systems, phase noise can significantly impair performance. To allow for compensation of these impairments, accurate phase noise modeling is necessary. Numerical modeling of the phase noise process at a phase-locked loop (PLL) output is established in the literature and commonly represented by an Ornstein-Uhlenbeck (OU) process. The corresponding spectrum can be represented by a multi-pole/zero model. This work presents a least squares (LS) method for estimating the PLL parameters such as oscillator constants or PLL bandwidth from a measured phase noise spectrum. The method is applied on the MAX2870 and MAX2871 PLL chips and parameter estimates such as oscillator constants and PLL bandwidths are provided. The resulting parameter set enables both time- and frequency-domain numerical simulations.

This paper studies joint range-angle estimation and communication in the NF ISAC systems, where the BS serves a single UE whose position is simultaneously estimated via monostatic sensing. Unlike the ULA, the UCA provides an angle-invariant NF region due to its rotational symmetry. To capture the full wideband NF propagation environment, we develop a continuous-time channel model incorporating per-element delay, Doppler shifts, and spherical wavefront geometry under OFDM signaling. Building on this model, we derive the closed-form CRLB for joint range-angle estimation of the UE position, design an optimal transmit beamformer via Riemannian gradient descent, and formulate a joint range-angle ML estimator. Monte Carlo simulations confirm a fundamental aperture-versus-SNR trade-off in NF-ISAC: while a larger UCA radius tightens the CRLB, it simultaneously reduces the received SNR at any given distance, pushing the maximum likelihood estimator below its convergence threshold and degrading practical performance. Among the evaluated configurations, R = 0.5 m achieves the best joint estimation and communication performance at the BS} by sustaining the highest received SNR throughout the evaluated range.

In the last decades, energy-based models (EBMs) have become an important class of probabilistic models in which a component of the likelihood is intractable and therefore cannot be evaluated explicitly. Consequently, parameter estimation in EBMs is challenging for conventional inference methods. In this work, we provide a unified framework that connects noise contrastive estimation (NCE), reverse logistic regression (RLR), multiple importance sampling (MIS), and bridge sampling within the context of EBMs. We further show that these methods are equivalent under specific conditions. This unified perspective clarifies relationships among existing methods and enables the development of new estimators, with the potential to improve statistical and computational efficiency. Furthermore, this study helps elucidate the success of NCE in terms of its flexibility and robustness, while also identifying scenarios in which its performance can be further improved. Hence, rather than being a purely descriptive review, this work offers a unifying perspective and additional methodological contributions. The MATLAB code used in the numerical experiments is also made freely available to support the reproducibility of the results.

This paper presents a detailed measurement campaign and a comprehensive analysis of 15 GHz ultra-massive multiple-input multiple-output (UM-MIMO) channels tailored for the urban microcell (UMi) environment. Channel sounding is performed over 14.875-15.125 GHz using a time-domain platform comprising a 128-element L-shaped transmit array and a 64-element square receive array. Four representative scenarios are investigated, namely near-field line-of-sight (LoS), near-field foliage-shaded, far-field foliage-shaded, and far-field LoS street canyon scenarios, resulting in 81 distinct transmit-receive links. Based on the measured data, conventional channel characteristics, including path loss, power delay angle profiles, delay spread, and angular spread, are characterized, while UM-MIMO-specific phenomena associated with near-field effects, spatial non-stationarity (SNS), and channel hardening (CHD) are quantitatively analyzed. Channel capacity is further evaluated to reveal the effects of different UMi propagation conditions on system performance. The reported results provide empirical support for the new mid-band spectrum (6-24 GHz, including Frequency Range 3 (FR3)) UM-MIMO channel modeling and offer practical guidance for the design and deployment of future sixth-generation (6G) microcell networks.

Traditional video coding (VVC, HEVC) prioritizes human visual perception, transmitting substantial texture redundancy that severely hinders machine decision-making under constrained bandwidths. In dynamic channels, this redundancy causes severe ``cliff effects'' and prohibitive latency. To address this, we propose a robust multimodal semantic communication framework based on an adaptive Object-Attribute-Relation (O-A-R) hierarchy. Bypassing pixel-level reconstruction entirely, our framework directly fuses visual, textual, and audio streams to construct a decision-oriented topological graph. A bandwidth-adaptive strategy dynamically allocates resources by semantic priority, while a cross-modal mechanism leverages text and audio priors to compensate for severe visual degradation. Experimental results demonstrate that under extreme low bandwidths (1-3 kbps), our method achieves over a 90% bandwidth saving (an approximately 10-fold reduction) compared to state-of-the-art digital schemes, maintaining superior scene-graph accuracy. In deep fading channels (SNR <= 4 dB), it completely eliminates the cliff effect, ensuring graceful degradation by strictly preserving foundational object anchors even when traditional codecs suffer 100% decoding failure. Coupled with an 89\% reduction in end-to-end latency, our framework comprehensively fulfills the real-time survival requirements of embodied agents.

Adaptive impedance matching between antennas and radio frequency front-end modules is critical for maximizing power transmission efficiency in mobile communication systems. Conventional numerical and analytical methods struggle with a trade-off between accuracy and efficiency, while deep neural network (DNN)-based supervised learning approaches rely heavily on large labeled datasets and lack flexibility for dynamic environments. To address these limitations, this paper proposes a deep reinforcement learning (DRL)-based approach for adaptive impedance matching. First, we model the impedance tuning problem as an optimal control problem, proving the feasibility of solving the optimal control law via reinforcement learning. Then, we design a tailored DRL framework for impedance tuning, which employs a compact state representation that integrates key frequency characteristics and matching quality metrics. Additionally, this framework incorporates a piecewise reward function that accounts for both matching accuracy and tuning speed. Furthermore, a test-phase exploration mechanism is introduced to enhance tuning stability, which effectively reduces local optimal trapping and high-frequency tuning variance. Experimental results demonstrate that the proposed method achieves superior performance in terms of tuning accuracy, efficiency, and stability compared with conventional heuristic and gradient-based methods, making it promising for practical impedance tuning systems.

Affine frequency division multiplexing (AFDM) has emerged as a promising waveform for high-mobility communications. However, its equalization remains a practical challenge under general physical channels with off-grid delay and Doppler effects. In this paper, we investigate frequency domain equalization for AFDM by considering a practical filtered-AFDM waveform. We analyze the input-output relations of filtered-AFDM across various domains and show that off-grid effects lead to severe inter-symbol interference in the DAFT domain, limiting the effectiveness of DAFT domain equalization. Motivated by the compactness of the frequency domain channel matrix in wideband systems, we propose a low-complexity two-stage frequency domain equalization scheme. Numerical results demonstrate that the proposed approach achieves performance close to full-block LMMSE equalization with significantly reduced computational complexity, and offers clear advantages over time domain equalization in wideband scenarios.

Electrodermal activity (EDA) is a widely used physiological signal for assessing sympathetic nervous activity, such as arousal, stress, and pain. However, reliable decomposition into tonic and phasic components remains challenging, particularly in noisy environments and across individuals with varying signal morphologies and stimulus responses. We propose ospEDA, a novel Orthogonal Subspace Projection (OSP) based method for EDA decomposition. The method integrates (1) tonic estimation via physiologically motivated valley detection for noise robustness; (2) phasic extraction using OSP to accommodate inter subject variability; and (3) phasic driver estimation through non-negative least squares (NNLS) deconvolution with ridge regularization. We evaluated ospEDA on five real-world datasets and one simulated EDA dataset with ground-truth components, comparing its performance against six existing methods. In simulations with a 20 dB signal to noise ratio (SNR), ospEDA achieved the lowest root mean square error (RMSE) for estimated tonic (0.131) and phasic (0.132) components. Under noisier conditions (10 dB SNR), it maintained superior phasic RMSE (0.293), Pearson correlation (0.782), and R^2 (0.979) values. Furthermore, ospEDA consistently provided the highest F1 scores (0.573, 0.617, 0.638) for sympathetic nerve activity detection across 10, 20, and 30 dB SNR levels, respectively, compared to existing methods. On the real world datasets, ospEDA achieved a stimulus classification AUROC of 0.766 and consistently maintained strong effect sizes (ω^2>0.14) across all five datasets. Overall, ospEDA represents a promising framework for EDA decomposition, showing generally consistent performance and reliable phasic driver estimation under the varying noise conditions, with potential utility for real world physiological monitoring applications.

Conventional delay-Doppler (DD) communication and sensing systems require transmitting pilot frames at every channel coherence time interval in order to keep track of channel variations at the cost of spectral efficiency. In this paper, we propose an approach to utilize data transmissions modulated using arbitrary waveforms for DD channel estimation without requiring pilot transmissions in every coherence time interval. Numerical evaluation over practical doubly-selective channel models demonstrate $\sim 1.8 \times$ improvement in spectral efficiency with our proposed data-based approach over conventional pilot-based approaches across various 6G modulation schemes.

Secret key generation (SKG) from wireless channel reciprocity has been demonstrated on WiFi, LTE, and LoRaWAN, but has never been demonstrated on 5G New Radio (NR) Sounding Reference Signal (SRS) and CSI Reference Signal (CSIRS) measurements.

A multi-hop amplify-and-forward (AF) relay network can emulate a fully connected (FC) neural network layer via over-the-air (OTA) computation. However, achieving high emulation accuracy requires accurate channel state information (CSI) across all links in the multi-hop network. In this work, we investigate the impact of CSI errors on classification performance. We propose five heuristic schemes for allocating the total channel training time (pilots) across hops and compare their effectiveness. Numerical results reveal a clear trade-off between channel training overhead and classification accuracy. In particular, with sufficient pilot power and balanced allocation of channel training resources, the system can achieve classification accuracy close to that of the digital baseline.

Sub-terahertz (sub-THz) multi-user multiple-input multiple-output (MU-MIMO) systems unlock immense bandwidth for 6G wireless communications. However, practical deployment of wireless systems in sub-THz bands faces critical challenges such as increased atmospheric absorption, reduced channel coherence time due to increased Doppler spread at higher carrier frequencies, and hardware bottlenecks as low-loss sub-THz phase shifters are difficult to realize. To overcome the hardware and channel estimation challenges of sub-THz systems, this paper proposes a hybrid beamforming (BF) framework that integrates reconfigurable liquid crystal (LC) antennas with a liquid neural network (LNN) for transmitter. Specifically, we employ an LC antenna as the analog BF stage of a hybrid BF architecture, exploiting its voltage-driven permittivity tunability to achieve high-gain beam steering without the need for lossy phase shifters. For digital BF, we utilize an ordinary differential equations-defined LNN to learn temporal channel dynamics, and use a manifold optimization technique to compress the search space. We validated the proposed method on simulated site-specific 108 GHz ray-tracing channels in an urban scenario using NYURay, a ray-tracing simulator validated against 142 GHz propagation measurements. The 108 GHz carrier frequency matches the operating band of the LC antenna hardware. The proposed method achieves an 88.6\% spectral efficiency (SE) gain and higher robustness to imperfect channel estimation compared to the learning-aided gradient descent and gated recurrent unit machine learning baselines, and 1.9 times higher SE than the 3GPP TR~38.901 standard antenna model, highlighting the potential of LC-based hardware for sub-THz communications.

Accurate time synchronization is essential for Internet of Things (IoT) systems, where multiple distributed nodes must share a common time base for coordinated sensing and data fusion. However, conventional synchronization approaches suffer from nondeterministic transmission latency, limited precision, or restricted bidirectional functionality. This paper presents a protocol-independent wireless timer synchronization method that exploits radio timeslots to transmit precisely timestamped beacons in a proprietary radio mode. By decoupling synchronization from upper-layer packet retransmissions and leveraging hardware-timed radio events, the proposed approach significantly reduces scheduling uncertainty and achieves nanosecond-level synchronization accuracy. Comprehensive experiments evaluate the impacts of synchronization frequency, RSSI, BLE connection interval, and throughput on synchronization performance. The results demonstrate that an optimal synchronization frequency of 1000 Hz yields an approximately 20 ns delay in the absence of communication stack activity while maintaining sub-500 ns accuracy under most realistic BLE traffic conditions. Furthermore, larger connection intervals, lower application throughput, and higher RSSI consistently improve synchronization quality by reducing radio resource contention and packet loss. The proposed scheme provides a general and high-precision synchronization solution suitable for resource-constrained IoT systems.

This paper studies the transmit waveform optimization for a quantized multiple-input multiple-output (MIMO) integrated sensing and communication (ISAC) system, where one-bit analog-to-digital converters (ADCs) are employed to enable a low-cost and power-efficient hardware implementation. Focusing on the parameter estimation task, we propose two novel Cramér-Rao bounds (CRBs) for both point-like target (PT) and extended target (ET) to characterize the impact of quantization distortion on the estimation accuracy, where associated estimation methods are also developed to approach these theoretical CRBs. Moreover, with the goal of jointly enhancing the sensing and communication performances, we formulate the bi-criterion ISAC waveform optimization problem by minimizing the derived CRB objectives subject to a communication symbol error probability (SEP) constraint and a total power constraint, which, due to the high nonlinearity of the one-bit CRBs, are extremely nonconvex. To yield a high-quality suboptimal solution, we develop an efficient alternating direction method of multipliers (ADMM) framework which exploits the majorization-minimization (MM) technique to address the nonconvex issue. Simulation results verify that the one-bit CRBs are tight for characterizing the quantized estimation performance and the proposed estimation methods also show clear performance advantages over the existing benchmark schemes. Furthermore, a flexible trade-off between the CRB and the SEP performance can be achieved by the developed ADMM framework, demonstrating the effectiveness of the optimized ISAC waveform.

Neural rendering paradigms have recently emerged as powerful tools for radio frequency (RF). However, by entangling RF sources with scene geometry and material properties, existing approaches limit downstream manipulation of scene geometry, wireless system configuration, and RF reasoning. To address this, we propose a physically grounded RF inverse rendering (RFIR) framework that explicitly decouples RF emission, geometry, and material electromagnetic properties. Our key insight is an RF-aware bidirectional scattering distribution function, embedded into the Gaussian splatting paradigm as an RF rendering equation. Each Gaussian primitive is endowed with intrinsic physical attributes, including surface normals, material electromagnetic parameters, and roughness, and leveraged by a customized ray-tracing scheme to represent RF signal synthesis. The proposed RFIR generalizes three typical RF tasks: radar cross-section synthesis, received signal strength indicator prediction, and wireless scene editability. Experiments demonstrate significant performance advantages, underscoring the potential for wireless world modeling.

The reconfigurable intelligent surface (RIS) has attracted considerable attention of both academia and industry in recent years, given its capacity to dynamically manipulate the reflection of incident electromagnetic waves. Although the research developed for the RIS may have reached its maturity, there are still contentious aspects and limitations regarding its potential benefits for the next generation of wireless communications. In order to improve upon the the RIS technology, the beyond diagonal reconfigurable intelligent surface (BD-RIS) was recently proposed as an promising alternative. The BD-RIS boasts a more sophisticated circuit topology that is capable of providing more combinations of different adjustments or configurations for signal reflection. However, to aptly reap the benefits of the BD-RIS, the added degrees-of-freedom of its configuration must be leveraged accordingly. Therefore, in this work we propose a depth-first tree search algorithm for configuring the BD-RIS in multi-user multiple-input single-output (MU-MISO) communication systems. Taking advantage of the tree search exploration, the proposed algorithm achieves a remarkable trade-off between channel strength maximization performance and computational complexity scalability.

Non-line-of-sight (NLOS) sensing has the potential to enable use cases like intrusion detection in occluded areas, increasing the value provided by Integrated Sensing and Communications (ISAC) in future 6G cellular networks. In this paper, we present a reliable NLOS intrusion detection system based on a millimeter-wave ISAC proof-of-concept. By leveraging reflections off a large surface, the proposed system addresses the challenge of detecting moving targets in cluttered indoor industrial scenarios where the direct line-of-sight is obstructed. A signal processing pipeline including a probability hypothesis density (PHD) filter is applied to detect targets and track movements in NLOS. Experimental validation conducted in the ARENA2036 industrial research campus demonstrates that our system can reliably detect target presence in NLOS while avoiding false alarms. Tests with synthetically generated false peaks further demonstrate the robustness of our system to false alarms. Overall, the results underline the potential of NLOS ISAC as a promising technology for enabling intrusion detection and monitoring use cases.

Time-varying distortions in communication systems can significantly degrade the performance of soft-decision forward error correction. This paper presents a burst-aware (BA) low-density parity-check (LDPC) decoding scheme for channels affected by bursty phase noise. By applying differential coding to a Wiener process with time-varying innovation variance, bursty differential phase noise is obtained. Simulation results demonstrate that, compared to conventional decoding, the BA scheme achieves gains in the signal-to-noise ratio of up to $0.7$~dB at a bit error rate (BER) of $4\cdot10^{-3}$ and more than $1$~dB at a packet error rate (PER) of $1\cdot10^{-2}$. Furthermore, by iterating between channel estimation and \ac{ldpc} decoding, forming the proposed iterative burst-aware (IBA) decoding scheme, the gains increase to $1.4$~dB and more than $3$~dB, respectively. More importantly, the IBA scheme significantly improves robustness to bursty phase noise. Compared with the conventional scheme, the IBA scheme can reduce both \ac{ber} and \ac{per} by up to two orders of magnitude under severe bursty phase noise.

This lecture note addresses the common misconception that the Gaussian distribution always yields the largest Cramér-Rao Bound (CRB). We show that this property only holds under restrictive conditions: specifically, when the mean and covariance parameters are decoupled in the Fisher Information Matrix (FIM), when the parameter of interest lies in the mean vector and when there are no additive nuisance parameters. Beyond this framework, we provide counterexamples demonstrating that non-Gaussian distributions can produce larger CRB.

With 6G evolving towards intelligent network autonomy, artificial intelligence (AI)-native operations are becoming pivotal. Wireless networks continuously generate rich and heterogeneous data, which inherently exhibits spatio-temporal graph structure. However, limited radio resources result in incomplete and noisy network measurements. This challenge is further intensified when a target variable and its strongest correlates are missing over contiguous intervals, forming systemic blind spots. To tackle this issue, we propose RieIF (Knowledge-driven Riemannian Information Flow), a geometry-consistent framework that incorporates knowledge graphs (KGs) for robust spatio-temporal graph signal prediction. For analytical tractability within the Fisher-Rao geometry, we project the input from a Riemannian manifold onto a positive unit hypersphere, where angular similarity is computationally efficient. This projection is implemented via a graph transformer, using the KG as a structural prior to constrain attention and generate a micro stream. Simultaneously, a Long Short-Term Memory (LSTM) model captures temporal dynamics to produce a macro stream. Finally, the micro stream (highlighting geometric shape) and the macro stream (emphasizing signal strength) are adaptively fused through a geometric gating mechanism for signal recovery. Experiments on three wireless datasets show consistent improvements under systemic blind spots, including up to 31% reduction in root mean squared error and up to 3.2 dB gain in recovery signal-to-noise ratio, while maintaining robustness to graph sparsity and measurement noise.

The integration of machine learning tools into telecom networks, has led to two prevailing paradigms, namely, language-based systems, such as Large Language Models (LLMs), and physics-based systems, such as Digital Twins (DTs). While LLM-based approaches enable flexible interaction and automation, they lack explicit representations of network dynamics. DTs, in contrast, offer a high-fidelity network simulation, but remain scenario-specific and are not designed for learning or decision-making under uncertainty. This gap becomes critical for 6G systems, where decisions must take into account the evolving network states, uncertainty, and the cascading effects of control actions across multiple layers. In this article, we introduce the {Telecom World Model}~(TWM) concept, an architecture for learned, action-conditioned, uncertainty-aware modeling of telecom system dynamics. We decompose the problem into two interacting worlds, a controllable system world consisting of operator-configurable settings and an external world that captures propagation, mobility, traffic, and failures. We propose a three-layer architecture, comprising a field world model for spatial environment prediction, a control/dynamics world model for action-conditioned Key Performance Indicator (KPI) trajectory prediction, and a telecom foundation model layer for intent translation and orchestration. We showcase a comparative analysis between existing paradigms, which demonstrates that TWM jointly provides telecom state grounding, fast action-conditioned roll-outs, calibrated uncertainty, multi-timescale dynamics, model-based planning, and LLM-integrated guardrails. Furthermore, we present a proof-of-concept on network slicing to validate the proposed architecture, showing that the full three-layer pipeline outperforms single-world baselines and accurately predicts KPI trajectories.

Dynamic Metasurface Antennas (DMAs) have been recently proposed as a cost- and energy-efficient front-end solution for eXtremely Large (XL) antenna array systems, supporting scalable Analog and Digital (A/D) beamforming while using a reduced number of Radio-Frequency (RF) chains. This array architecture is commonly realized as partially connected hybrid A/D beamformers, in which non-overlapping subarrays are linked to separate RF chains, each attached to a waveguide hosting multiple metamaterials. In this work, we study uplink multi-user communications where each RF chain of an XL DMA receiver is equipped with a $b$-bit resolution Analog-to-Digital Converter (ADC). We cast a Mean Squared Error (MSE) minimization problem for the design of the hybrid A/D combiner aimed at multi-user symbol detection, which is intrinsically non-convex due to the structural constraints imposed by the DMA hardware. By exploiting the Bussgang decomposition and a tractable modeling framework, we propose an efficient joint design of the hybrid A/D combining parameters. Our numerical evaluations showcase that XL DMA receivers can perform highly accurate multi-user symbol detection, revealing attractive trade-offs between hardware complexity and MSE performance.

This paper considers a Fluid Antenna (FA) system comprising a single-antenna transmitter that communicates with a receiver equipped with an FA array with $N$ ports. The transmitter is assumed to deploy any of the modulation schemes: \textit{i}) two-sided $M$-ary amplitude-shift keying, \textit{ii}) $M$-ary phase-shift keying, iii) $M$-ary quadrature-amplitude modulation, and \textit{iv}) binary frequency-shift keying, the channels between its antenna and the receiver ports are subjected to Rayleigh fading, and the receiver chooses the best $K$ out of its $N$ ports for symbol detection. Considering that the receiver combines the signals from the best $K$ ports using maximal-ratio combining, the optimal reception structures for all the considered signaling schemes are obtained. We also present novel exact closed-form expressions for the respective symbol error probabilities (SEPs) of the FA system, as well as asymptotic approximations valid at high signal-to-noise ratios. The presented analysis is corroborated through comparisons with simulation results, showcasing the critical role of various system parameters on the SEP performance.

Understanding surface material properties is crucial for enhancing indoor robot perception and indoor digital twinning. However, not all sensor modalities typically employed for this task are capable of reliably capturing detailed surface material characteristics. By analyzing the reflected RF signal from a mmWave radar sensor, it is possible to extract information about the reflective material and its composition from a certain surface. We introduce a mmWave MIMO FMCW radar-based surface material classifier SMCNet, employing a complex-valued Convolutional Neural Network (CNN) and complex radar IQ signal input for classifying indoor surface materials. While current radar-based material estimation approaches rely on a fixed sensing distance and constrained setups, our approach incorporates a setup with multiple sensing distances. We trained SMCNet using data from three distinct distances and subsequently tested it on these distances, as well as on two more unseen distances. We reached an overall accuracy of 99.12-99.53 % on our test set. Notably, range FFT pre-processing improved accuracy on unknown distances from 25.25 % to 58.81 % without re-training.

Radar sensors operating in the mmWave frequency range face challenges when used as indoor perception and imaging devices, primarily due to noise and multipath signal distortions. These distortions often impair the sensors' ability to accurately perceive and image the indoor environment. Nevertheless, this sensor offers distinct advantages over camera and LiDAR sensors. This encompasses the estimation of object reflectivity, known as radar cross-section (RCS), and the ability to penetrate through objects that are thin or have low reflectivity. This results in a 'through-the-wall' sensing capability. Due to the aforementioned disadvantages, most research in the field of imaging radar tends to exclude indoor areas. We introduce a machine learning-based mmWave MIMO FMCW imaging radar object classifier designed to identify small, hand-sized objects in indoor settings, utilizing only radar IQ samples as input. This system achieves 97-99 % accuracy on our test set and maintains approximately 50 % accuracy even under challenging conditions, such as increased background noise and occlusion of sample objects, without the need for adjusting training or pre-processing. This demonstrates the robustness of our approach and offers insights into what needs to be improved in the future to achieve generalization and very high accuracy even in the presence of significant indoor perturbations.

This paper proposes an original method for estimating the velocity of a target by leveraging the multiband capabilities of modern Integrated Sensing And Communication (ISAC) systems. Traditional Doppler estimation relies on regular sampling rates, but ISAC systems often face irregular packet arrival times because they reuse opportunistic communication traffic. This non-deterministic timing increases the risk of Doppler ambiguity and aliasing, degrading velocity estimation accuracy. To resolve this, we advocate exploiting frequency diversity across multiple carrier frequencies to observe Doppler shifts without imposing restrictions on packet timing or requiring dedicated sensing overhead. A multiband velocity estimation problem is here formulated as a mixed-integer quadratic program by utilizing phase differences from all possible pairwise packet combinations. By integrating at least one unambiguous phase measurement, the system can reconstruct the true target velocity even under sporadic traffic conditions. Simulation results using realistic traffic traces demonstrate that this approach significantly outperforms multiband likelihood-based and single-band algorithms, with accuracy improving as frequency separation between bands and inter-packet time intervals increase. This framework provides a zero-overhead solution for robust velocity estimation in dynamic ISAC environments.

The integration of multimodal sensing and millimeter-wave (mmWave) communications is a key enabler for highly mobile vehicle-to-infrastructure (V2I) networks. However, continuous high-resolution visual sensing incurs prohibitive computational energy, while delayed sensing information worsens beam misalignment. In this paper, we establish a physics-aware multimodel integrated sensing and communication (M-ISAC) framework that quantifies the mathematical trade-off between sensing energy and communication reliability using the semantic age of information (AoI). To address the coupled challenges of temporal AoI evolution and instantaneous non-convex constant modulus constraints, we propose a novel reinforcement learning approach empowered by a heterogeneous mixture-of-experts (RL-H-MoE) architecture. By strictly decoupling the temporal scheduling and spatial phase mapping, the RL-H-MoE avoids prevalent gradient conflicts in multi-task learning. Extensive simulations demonstrate that the proposed architecture achieves an optimal event-triggered sensing policy, significantly minimizing the long-term system cost while guaranteeing ultra-low sensing errors and reliable physical-layer link connectivity.

The rigorous analysis of specialized physical processes often demands custom data acquisition architectures that offer flexibility and precision beyond the capabilities of general-purpose commercial loggers. This paper presents the design and implementation of a robust data acquisition system (DAQ) for a comparative analysis of the performance of two photovoltaic panels with two different cooling systems. The system integrates a custom PCB design for 20 thermistors, dual high-precision INA228 current/voltage sensors, environmental monitoring equipment, and a Raspberry Pi 4-based acquisition platform. The software architecture implements autonomous operation with enhanced fault recovery, dual storage redundancy (local CSV and InfluxDB), cloud synchronization via Google Drive, and real-time visualization through Grafana dashboards. Field deployment demonstrated system reliability, including automatic recovery from power interruptions, a 1-minute sampling rate, remote monitoring capabilities, and continuous operation during a 5 AM to 6 PM daily window. The modular hardware and software architecture enables simultaneous monitoring of two photovoltaic panels for research on direct performance comparison under identical environmental conditions.

Accurate localization in non-line-of-sight (NLoS) environments remains challenging even with both angle-of-arrival (AoA) and time-of-arrival (ToA) measurements. In complex urban scenarios, the absence of line-of-sight (LoS) paths and the lack of environment prior knowledge make geometric based localization methods inapplicable, while prior-based approach such as fingerprinting is sensitive to environmental perturbations. This paper proposes a novel environment-aware localization framework enabled by the emerging concept called channel knowledge map (CKM). In the offline stage, AoA-ToA path signatures are learned by the CKM, with each path mapped to one candidate scatterer, thereby forming geometric priors within the environment. In the online stage, observed paths are matched to the CKM to extract high-confidence scatterers. Nonlinear least squares (NLS) method is then applied to jointly estimate the user and dominant scatterer locations. Even with imperfect CSI matching, geometric feasibility consistent with CKM scatterer priors provides corrective information and suppresses ambiguity. Simulations demonstrate that the proposed scheme outperforms fingerprinting and offers a robust and scalable solution to address the challenging NLoS localization for integrated sensing and communication (ISAC) systems.

Cost-effective, low-complexity and spectrally efficient interconnection can offer fundamental guiding law for future datacenter. In this work, we demonstrate a cost-efficient SSBI-free direct detection for datacenter interconnection, leveraging the phase diversity of residual optical carrier caused by finite-extinction ratio (ER) IQ modulators, combining the device cost-effective IQ modulator with finite-ER and efficient SSBI-free phase-diverse direct detection receiver. Specifically, the proposed solution transforms the inherent limitation of finite-ER of cost-effective IQ modulator into the residual optical carrier advantage of SSBI-free direct detection systems, eliminating SSBI without additional hardware and control complexity. A digital pre-distortion and offset correction algorithms, and a PD-thermal-noise constrained SSBI-free direct detection and signal recovery algorithms are derived and implemented. Comprehensive simulations are conducted. A Global-SNR gain of 1.78 dB and 400 Gb/s data rate are achieved in 100-km SSMF transmission when (ER_i, ER_o)= (7 dB, 25 dB) of IQ modulator. The proposed solution enables low-complexity, cost-effective, and spectrally-efficient interconnects for next-generation datacenters.

Physics-informed neural networks (PINNs) have emerged as a powerful framework for modeling physical systems and solving inverse problems. In such settings, sensors are deployed to capture observable system responses; however, the quality of reconstruction critically depends on how these sensors are selected. Existing sensor selection strategies for PINNs are closely related to active learning and experimental design, typically relying on iterative refinement schemes that sequentially add sensors and retrain the model. While effective under limited data regimes, these approaches incur substantial computational cost due to repeated retraining and primarily focus on selecting subsets of sensors, without providing a global characterization of sensor importance. In this work, we propose FOSSA, a first-order optimality-based sensor selection algorithm for inverse PINNs. Unlike existing methods, FOSSA evaluates sensor importance in a post-training manner, requiring only a single trained PINN. FOSSA assigns importance scores to all candidate sensing locations based on the first-order optimality condition at convergence. To improve robustness, a refinement scheme is further proposed to handle instability in the inverse solver. FOSSA facilitates a global assessment of the contribution of each sensor to reconstruction. We validate the proposed approach on the inverse electrocardiography (ECG) modeling and show that not all sensors contribute positively to predictive performance. Incorporating low-importance sensors can, in fact, degrade reconstruction accuracy. These findings highlight the need for principled sensor importance evaluation and provide a scalable pathway for guiding sensor deployment in physics-informed inverse modeling.

Using the worldline action in [2409.05860], we derive the total even-in-velocity (relative) impulse, scattering angle, and time delay at fifth post-Newtonian (5PN) order, including radiation-reaction and hereditary contributions at ${\cal O}(G^5ν^2)$ and ${\cal O}(G^6ν^2)$. We introduce an isotropic-like description which, together with the associated losses of energy and angular momentum, fixes the evolution of the system from scattering data. This framework opens the door to an unambiguous characterization of the underlying two-body dynamics solely in terms of scattering observables. Following [2409.05860], we isolate a conservative component using Feynman's $i0^+$ prescription. This sector contains both "tail-like" and "memory-like" contributions, the latter being nonlocal in time and described by a double Principal-Value integral. Owing to the local-in-time character of the corresponding (in-in) action, we establish a systematic procedure that is consistent with Feynman's prescription while preserving the complete local dynamics. This provides a universal contribution to the conservative (isotropic) Hamiltonian at 5PN order and, as a byproduct, also fixes the value of the Effective One Body coefficients $\{{\bar d}_{5{\rm loc}}, a_{6{\rm loc}}\}$ consistently with the Tutti-Frutti framework. For completeness, we analyse the "$γ\text{-}3$" prescription introduced in recent post-Minkowskian computations. When implemented in our formalism, we find exact agreement over the overlapping regime of validity. In contrast, Feynman's prescription yields a (local) memory-like contribution with the opposite sign at ${\cal O}(G^5ν^2)$. We also find that an analogous $γ\text{-}3$ rerouting at ${\cal O}(G^6ν^2)$ would be incompatible with the conjecture that all $π^2$ terms arise solely from the potential region, while Feynman's formulation preserves this expectation.

The scale evolution of parton distributions is governed by splitting functions. We compute the four-loop splitting functions in perturbative QCD that control the evolution of quark non-singlet distributions. We confirm previous partial results and obtain, for the first time, fully analytic expressions for all non-singlet contributions at this order. These allow us to extract the analytic form of the four-loop virtual and rapidity anomalous dimensions entering logarithmic resummation. We provide precise numerical representations of the splitting functions suitable for parton evolution.

We classify superfusion categories describing two-dimensional fermionic systems equipped with the universal fermion-parity symmetry, implemented by a topological defect line (TDL) $Z$, and an additional $\mathbb{Z}_2$ flavor symmetry generated by a $W$ TDL. Depending on whether $W$ is m-type or q-type, its fusion rules lead to three distinct classes, and solving the super-pentagon equations yields 16 consistent superfusion categories. These are labeled by invariants $(ν_W,ν_Z,ν_{WZ})$, which determine the $\mathbb{Z}_8$ anomaly classes of the symmetries generated by $W$, $Z$, and $WZ$. We also provide explicit realizations using multiple Majorana fermions and comment on implications for fermionic CFTs and gapped phases.

Lie-Poisson classical field theory is a field-theoretical model embedded in a non-commutative structure related to the framework of Poisson electrodynamics. In this paper, we follow the recently developed action principle for Lie-Poisson electrodynamics to derive the conservation laws of the theory. The energy-momentum tensor is obtained, along with the conserved electric charge and the momentum operator. We consider non-interacting examples for real and complex scalar fields, as well as the Dirac field, within the $κ$-Minkowski spacetime framework. In the latter case, we show that the non-relativistic limit for the $κ$-Minkowski Dirac equation introduces an orbital Zeeman coupling term for the fermionic fields, and the energy shift in the first excited state depends exclusively on the $κ$-parameter.

We show how equivariant localization can be used to compute the on-shell action for supersymmetric $D=5$ $AdS$ rotating, charged black holes in theories of supergravity with higher derivatives. An exact match with a dual field theory computation of the superconformal index in a Cardy-like limit is achieved.

We study theoretical uncertainties in predicting top-quark pair-production near threshold at the LHC using the non-relativistic QCD framework. We include variations in the top-quark mass and width, the strong coupling $α_s$, renormalization and factorization scales, and parton distribution functions, as well as uncertainties from the color-singlet and octet Green's functions that describe quasi-bound toponium formation. These uncertainties are compared with those from standard fixed-order QCD predictions, and implications for ATLAS and CMS analyses are discussed. For the LHC at 13 TeV center-of-mass energy, the integral of the top-quark pair invariant-mass distribution from 340 to 350 GeV is 11.67 pb with ${}^{+1.43}_{-1.47}$ pb uncertainty. The corresponding excess after subtracting the POWHEG-BOX result is 4.15 pb with the same uncertainties.

We compute the quantum radiation emitted by wavepackets of relativistic single electrons, both at rest and undergoing uniform acceleration in the Minkowski vacuum of the electromagnetic field. We find that the cubic terms in the original nonlinear action of electrodynamics should be considered in obtaining the quantum radiation to the leading order. We show that the quantum radiation from a single-electron wavepacket at rest vanishes exactly. For a uniformly accelerated electron, the quantum radiated power has secular growth in the long-time regime. We demonstrate that this secular growth has a classical interpretation, and argue that the resummed quantum radiation at late times would not diverge. Regarding experimental proposals for the detection of the Unruh effect from the quantum radiation in the `blind spots' of classical radiation we ascertain that quantum corrections in the two blind spots are fully contributed by the transverse deviation correlators, where the dominant contributions are irrelevant to the Unruh effect in electron microscopes.

We develop a framework to construct worldvolume gauge theories on D2-branes probing compound Du Val (cDV) Calabi-Yau threefold singularities. By viewing these singularities as ADE surface fibrations over the complex $w$-plane, we encode their geometry in a Higgs field $Φ(w)$. A D2-brane probe perceives $Φ(w)$ as an $\mathcal{N}=2$ deformation of its 3d $\mathcal{N}=4$ affine Dynkin quiver gauge theory via polynomial and monopole superpotential terms. By exploiting 3d mirror symmetry, we obtain an effective theory that correctly reproduces the quiver-collapsing mechanism known in the mathematical literature. We present several examples, including non-toric and non-resolvable cases.

Minkowski spacetime exhibits infrared instability in projectable Ho\v rava gravity in (3+1) dimensions. To be phenomenologically viable, the instability should be either hidden by other time-dependent processes such as the Hubble expansion of the universe and the Jeans instability, or evolve into another static solution with low average curvature. While the former scenario leads to a phenomenological constraint on the infrared properties of the renormalization group flow, this paper explores the latter possibility. We study if the presence of higher derivative terms in the action can lead to existence of static, inhomogeneous (quasi-) periodic solutions with planar symmetry, similar to modulated phases in magnetic materials. We find that such solutions do not exist. In doing so, we classify all static homogeneous and isotropic solutions and solutions with planar symmetry. We provide arguments that none of them can serve as an endpoint for the evolution of the Minkowski instability. This motivates further study of the scenario where the instability is concealed by time evolution.

We study a model of 3d gravity relevant to the open sector of a CFT ensemble. The quantum theory is the open Virasoro TQFT, obtained by restricting the full open-closed Virasoro TQFT to a subclass of admissible manifolds. We show that it computes gravitational path integrals on compact regions with fixed-length boundary conditions for states above the black hole threshold, and fixed-angle boundary conditions for states below the threshold. Focusing on a special class of manifolds involving only boundary Wilson loops, we further show that the relation between Conformal Turaev-Viro theory and the diagonal sector of two copies of Virasoro TQFT arises naturally from an open-closed duality.

We investigate the fermionic Casimir effect for a Dirac field confined between two parallel plates with MIT bag boundary conditions in the presence of CPT-odd Lorentz-symmetry violation described by a constant axial background vector $b_μ$. The exact mode quantization is derived from the modified Dirac equation in the planar geometry, and the vacuum energy is formulated through a phase-shift representation. For spacelike backgrounds we show that the components parallel to the plates can be absorbed into a shift of the transverse momenta and therefore do not affect the renormalized Casimir energy, while the component normal to the plates modifies the longitudinal spectrum and produces a genuine Lorentz-violating correction. Both the timelike component $b_{0}$ and the normal spacelike component $b_{z}$ can thus be treated within a unified framework characterized by a single effective spectral parameter. A closed logarithmic integral representation for the Casimir energy is obtained and its behavior is analyzed in the Lorentz-symmetric, weak-background, and strong-background regimes.

Measurement of $N_{\rm eff}$ in the CMB (Cosmic Microwave Background) observations, like Planck 2018 and BBN (Big Bang Nucleosynthesis) has already set stringent constraints on the interaction strength of light particles beyond the Standard Model (BSM). Despite such negligible couplings of such BSM particles to the visible sector, they are inevitably produced in the early universe through gravity-mediated processes. If a sizable density of light particles survives around CMB formation, they may act as dark radiation (DR) contributing to $N_{\rm eff}$ at CMB epoch. In this work, we study the production of such light BSM particles through the gravity-mediated scatterings in an effective field theory (EFT) setup assuming that all non-gravitational couplings of the BSM particle are negligible. Since the production is sensitive to the spin of the produced particle, we perform a concrete analysis for two representative cases: scalar dark Higgs DR and vector dark photons DR.Using the Planck 2018 observations, we find constraints on the reheating temperature ($T_{\rm RH}$) and background equation of state ($w_Φ$) during reheating in such scenarios featuring dark Higgs and dark photon. A comparative discussion involving gravity-mediated production of Dirac right-handed neutrinos ($ν_R$) and light axion-like particles (ALP) is also presented. Finally, for completeness, we also analyze the scenario where the production occurs through a generic spin-2 mediator characterized by an effective scale $Λ$ delineating the parameter space that is currently ruled out from Planck-2018 and can be probed by the future CMB experiments like LiteBird, Simon Observatory, CMB-S4, CMB-HD.

We formulate a general prescription for spurion analysis in particle-physics models whose selection rules are described by commutative non-invertible fusion algebras. The construction applies to fusion algebras containing non-invertible basis elements that need not be self-conjugate, thereby allowing us to systematically track coupling constants in arbitrary particle scattering processes at tree and loop orders, but without assuming faithful realization of the fusion algebra, or no other quantum numbers for dynamical particles. This unifies and streamlines the previous analysis of near-group fusion algebras and of the $\mathbb{Z}_M/\mathbb{Z}_2$ fusion algebras, and supports the broader viewpoint that the non-invertible selection rules often admit auxiliary descriptions using lifted Abelian groups with a structured set of explicit breaking terms.

We study the massless Maxwell system on fixed sub--extremal Kerr-Newman exteriors. In the weak Kerr--Newman regimes with $0<|a|\le c_a M$, $0<|Q|\le c_Q M$, and $c_a,c_Q \in (0,1) $, we prove non--degenerate boundedness, Morawetz / integrated local energy decay, the outgoing $r^p$ hierarchy, radiation fields, polynomial decay, and two--sided scattering for the stationary-subtracted Maxwell field. In the axisymmetric relative slow--rotation regime $0<|a|\le \varepsilon_{\mathrm{KN}}|Q|<M$ with $\varepsilon_{\mathrm{KN}} \in (0,1)$, the same conclusions hold on the axisymmetric sector, with no perturbative restriction on $|Q|/M$ beyond sub--extremality. For the full sub-extremal range $a^2+Q^2<M^2$, by adding one order--zero radiative local--energy estimate for this master system, we could obtain similar results as the weak regimes. This separates the unconditional weak theory from the conditional full--range theory within one reconstruction/scattering framework.

We provide the first investigation of the solar production of symmetrons, a well-motivated class of screened scalar fields with density dependent couplings to the Standard Model, and their subsequent absorption in underground direct detection experiments. We compute the flux of symmetrons produced through photon conversion in the magnetic field of the solar tachocline, and constrain the resulting luminosity to not exceed 3% of the observed solar output. Even under the conservative assumption that production occurs only in the tachocline, this criterion yields robust astrophysical bounds on previously uncharted regions of symmetron parameter space, and predicts a keV-scale symmetron spectrum at Earth. We then derive the corresponding absorption signal in liquid xenon detectors, where symmetrons can interact with electrons through both conformal and disformal couplings. Using binned data from XENONnT, we obtain new direct-detection limits that are complementary to the solar luminosity constraint, further tightening the viable symmetron parameter space. Our results demonstrate that the Sun provides a testable, previously unexploited, source of symmetrons, and highlight that the interplay of astrophysical and laboratory searches offers a powerful strategy for probing screened scalar theories.

Over the last decade, non-relativistic theories have attracted considerable attention. In general, such theories can be obtained by contracting relativistic parent theories. In this work, we discuss features of non-relativistic supersymmetric field theories from both the Galilean and Carrollian points of view that may be useful for constructing electric and magnetic non-relativistic theories.

As the search for observable deviations from the Standard Model of particle physics remains to be of significant interest, effective field theory (EFT) continues to be a popular method to parametrize such effects. In this work, a first-time investigation is performed of the unique capability of measurements of charm quark associated Higgs boson production (cH) in proton-proton collisions at the CERN Large Hadron Collider to constrain a set of dimension-six EFT operators. The phenomenology of these operators is discussed and a proposed analysis strategy is presented, with a focus on $\mathrm{H}\rightarrow \mathrm{Z}\mathrm{Z}^{*}\rightarrow 4μ$ decays, using a generic detector simulation that is parametrized to reflect the response of the CMS detector at the LHC. From this, expected 95% CL upper limits are derived for the Wilson coefficients of individual operators by considering yield and shape effects in the spectra of the four-muon invariant mass $m_{4μ}$ and leading jet transverse momentum $p_{T}$. Scenarios with simultaneous contributions from two operators are also considered. Finally, potential analysis improvements that may be implemented in an experimental context are outlined.

We present and demonstrate a version of Levinson's theorem especially dedicated to the asymptotic behavior of form factor phases. Indeed, as required by analyticity, form factors are multi-valued complex functions of a square four-momentum defined in the complex plane with a cut along the positive real axis. Their phases evaluated on the upper edge of this cut, i.e., on the time-like region, tend asymptotically to integer multiples of $π$ radians. The Levinson's theorem establishes a univocal relation between such multiples and properties of form factors related to the dynamics of the electromagnetic interaction of the corresponding hadrons.

Hadronic form factors fulfill dispersion relations and superconvergence sum rules for their spectral density as genuine imprints of QCD. We show several instances where these conditions are flagrantly violated due to the lack of information in the region above the largest known resonance mass and below the onset of perturbative QCD. We propose to use radial Regge trajectories to fill this gap and examine the consequences of such a minimal spectral hadronic ansatz.

We propose a covariant, gauge-independent construction of foliation-based scalar-tensor theories, yielding diffeomorphism-invariant operators involving only gradients on the hypersurfaces where the scalar field is constant, assumed to be spacelike. This defines a basis of independent invariants up to four derivatives of $φ$, including the first nontrivial parity-odd pseudoscalar at this order, with a straightforward extension to higher derivatives. Our framework goes beyond degenerate higher-order scalar-tensor (DHOST) theories and provides a nonlinear extension of U-DHOST (where $\nabla_μφ$ is supposed to be timelike) directly in covariant form, without using unitary gauge as a starting point or imposing degeneracy a priori. After minimal coupling to gravity, we analyze the theory through its Hamiltonian constraint structure and linear cosmological perturbations about an FLRW background, and show that it propagates three physical degrees of freedom.

We present a systematic study of the electromagnetic and spin polarizabilities of spin-1/2 singly charmed baryons at $\mathcal{O}(p^4)$ within the framework of heavy baryon chiral perturbation theory. Our results show that the higher-order corrections to the electric polarizability are small, while those to the magnetic polarizability are relatively larger due to the small mass splitting of singly charmed baryons and are closely related to transition magnetic moments. Furthermore, we find that the spin polarizabilities of singly charmed baryons, except for $γ_{M1M1}$, are much smaller than those of the nucleons. We have also calculated the polarizabilities for singly bottom baryons, with the results showing generally larger values than those of singly charmed baryons.

We study cosmological solutions of (pseudo)scalar theories with periodic potentials, in the presence of arbitrary cosmological fluids -- including a cosmological constant of either sign. Independently of the initial misalignment angle and field velocity, we derive an analytic bound that the axion mass parameter and decay constant fulfill as the universe decreases its acceleration rate, finding a natural application in models of thawing quintessence. As a first application, we illustrate the analytic handle our bound provides in bounding axion dark energy, after observational inputs from DESI and various supernovae data sets are taken into account. As a second application, we argue that our analytic bounds in combination with proposed quantum gravity constraints on axions exclude vast regions of parameter space. The combined constraints push the axion masses to be much larger than the Hubble scale, in tension with basic models of axion quintessence.

This paper is an extension of the results presented in \cite{Guarino:2024gke}. We study $ G_S$-invariant subsectors of maximal gauged supergravities and show that such models can provide consistent truncations even when $G_S$ is not a symmetry of the original supergravity. We show that this construction is key to building pure supergravities around a supersymmetric AdS$_D$ solution. We illustrate this construction by building a consistent $\mathcal{N}=4$ subsector of the $D=4$ $\mathcal{N}=8$ $[\mathrm{SO}(6)\times \mathrm{SO}(1,1)]\ltimes \mathbb{R}^{12}$ gauged supergravity. We use this result to build the uplift of the multicharge spindle solutions in type IIB and we define a simple criterion for assessing the regularity of the uplift. We show that the type IIB uplift of the spindle is always non-regular, admitting eight codimension-six orbifold singularities. We apply the same criterion to other spindle uplifts, recovering known results and making predictions on the regularity of spindles on (quasi-)regular SE$_7$ manifolds.

Dimensionally or analytically regulated Feynman integrals lead to relative twisted period integrals. We present a recent extension of the Griffiths-Dwork pole reduction algorithm for deriving the D-module of differential operators acting on the twisted differential forms from Feynman integrals. We illustrate the application of this algorithm by providing twisted Picard-Fuchs operators for hypergeometric, elliptic and Calabi-Yau differential motives arising from families of Feynman integrals.

We analyze non-invertible topological interfaces and defects in the two-dimensional compact boson, focusing on the more exotic ones obtained by gauging continuous symmetries with flat connections on a half-space. These include interfaces between mutually irrational radii and T-duality symmetries at arbitrary boson radius. Using the modified Villain discretization on both a Euclidean two-dimensional square lattice and a quantum one-dimensional chain, we show that all these topological interfaces survive discretization and give rise to non-compact edge modes localized at the defect sites. Such non-compact edge modes imply a continuous defect spectrum and an infinite quantum dimension. In the special case of rational radii, we show how the defect action or Hamiltonian can be modified in order to compactify the edge modes and produce more standard defects with finite quantum dimension.

We present a systematic study of gravitational wave (GW) signals from phase transitions and topological defects in a unified high-quality axion framework. The gauged $U(1)_g$ symmetry forbids any bias term that could lift the vacuum degeneracy, restricting the theory to the phenomenologically viable case $N_{\rm DW}=1$. Requiring the axion to account for the observed dark matter (DM) abundance and satisfy the high-quality condition constrains the gauge symmetry-breaking scale to $f_g \in [1.6\times10^{11},\,10^{16}]\,\mathrm{GeV}$ for the QCD axion, leading to a well-defined band of GW signals, part of which is consistent with current pulsar timing array observations. Two-step first-order phase transitions are common in this framework, with the lower-scale transition generating GWs with $f^{\rm peak} \gtrsim \mathcal{O}(10^7)\,\mathrm{Hz}$. For axion-like realizations, generic post-inflation models predict GW spectra that are nearly degenerate with the QCD axion case. We conclude that GWs alone cannot distinguish between these scenarios, highlighting the need for complementary probes.

We present a numerical program that evaluates deep inelastic scattering (DIS) structure functions at next-to-leading order (NLO) accuracy in the dipole picture. In this numerical implementation the NLO DIS impact factors with massive quarks are written in a form that ensures a stable numerical evaluation of the DIS cross sections.

Nuclear charge radii constitute a physical observable of growing significance across multiple subdisciplines of physics and related fields. Their determination relies on a combination of complementary experimental techniques and advanced theoretical frameworks. Current recommended values are informed by the outcomes of several independent working groups, each employing distinct methodological approaches and evaluation strategies. The present effort is directed toward a more precise and reliable extraction of charge radii, as well as the development of a modern, transparent, and methodologically robust compilation of recommended values.

The motion of the Solar System with respect to the cosmic rest frame induces a kinematic dipole in the stochastic gravitational-wave background (GWB). Detecting this signal with space-based interferometers would provide an independent measurement of our peculiar velocity and a GW probe of cosmic anisotropies. We present a fully analytic derivation of the response of the \emph{Laser Interferometer Space Antenna} (LISA) to a kinematic dipole, and construct an optimal estimator for its detection. We show that the dipolar response is governed by a single frequency-dependent function fixed by symmetry, and we compute its behaviour across the LISA band. Using Fisher forecasts, we find that for a scale-invariant background detectability requires $h^2Ω_{\rm GW} \gtrsim 5\times 10^{-8}$ for \emph{fiducial} LISA, and $h^2Ω_{\rm GW} \gtrsim 5\times 10^{-10}$ for a detector with characteristic instrumental-noise amplitudes improved by an order of magnitude. Prospects are more favorable for signals with richer frequency profile. We also explore the potential of the kinematic dipole to break degeneracies, particularly in the presence of strong galactic foregrounds or noise features that closely mimic the primordial signal.

Nowadays, several projects of future colliding beam experiments are being considered in the world. Among them are the CEPC (Circular Electron Positron Collider) in China and the FCC (Future Circular Collider) at CERN (Switzerland). To perform experiments on flavor physics in the energy range of the projects an excellent particle identification up to momenta of 30 GeV/c is required. Several concepts of RICH detectors based on aerogel were considered and evaluated with the help of GEANT4 simulation: FARICH (Focusing Aerogel RICH) based on multilayer aerogel with maximal refractive index of 1.008, RICH with Fresnel lens based on aerogel with refractive index of 1.008, RICH based on transparent aerogel fibers with refractive index of 1.008. The results of the simulation are presented. Some results of beam tests at the BINP performed to validate GEANT4 simulation are presented as well.

The NA62 experiment at CERN has collected a large sample of $K^+$ and $π^+$ decays in flight during Run 1 in 2016--2018 and the ongoing Run 2 which started in 2021. Searches for the decays $K^+\rightarrowπ^+X$ and $π^+\rightarrow e^+N$ are presented using NA62 data collected in 2016--2022 and 2017--2024, respectively. Results are interpreted to constrain a range of new physics scenarios covering all four portal model scenarios. Upper limits on the $K^+\rightarrowπ^+X$ branching ratio are established at the $10^{-11}$ level, providing constraints on dark photon, scalar and ALP couplings. From the search for heavy neutral lepton production in $π^+\rightarrow e^+N$ decays of beam pions, upper limits of the extended neutrino mixing matrix element $|U_{e4}|^2$ are established at the $10^{-8}$ level over the heavy neutral lepton mass range 95--126~MeV/$c^2$.

A class of solutions in $d$-dimensional Einstein gravity minimally coupled to a massless scalar field is studied, where the spacetime metric is of a generalized Weyl form with $d-2$ commuting Killing vectors. In addition to the procedure to generate scalar multipolar fields, a $SO(2)$ symmetry can be exploited to generate further solutions. A particular result of this procedure is a solution that contains the scalar counterpart of the Schwarzschild--Melvin and the Fisher--Janis--Newman--Winicour solutions as particular limits. Furthermore, a Harrison-type transformation can also be performed to generate solutions with magnetic fields. Using this transformation we obtain a solution with magnetic and scalar fields present and contains both magnetic and scalar counterparts of Schwarzschild--Melvin as limits.

The Pauli exclusion principle (PEP) was tested for nucleons in $\rm{^{12}C}$ nuclei using the Borexino dataset from 2007 to 2021. %the complete Borexino detector data. The approach consists of searching for $γ$-quanta, neutrons, protons, as well as electrons and positrons emitted in non-Paulian transitions of nucleons from the $1P_{3/2}$ shell to the filled $1S_{1/2}$ shell. Due to the uniquely low background level, the large mass, and long measurement time of the Borexino detector, the most stringent experimental constraints to date on the lifetime of the $\rm{^{12}C}$ nucleus with respect to PEP-forbidden transitions were obtained: $τ({^{12}\rm{C}}\rightarrow{^{12}\widetilde{\rm{C}}}+γ) \geq {1.1\times10^{32}}$ y, $τ({^{12}\rm{C}}\rightarrow{^{11}\widetilde{\rm{B}}}+ p) \geq {1.0\times10^{31}}$ y, $τ({^{12}\rm{C}}\rightarrow{^{11}\widetilde{\rm{C}}}+ n) \geq 2.0 \times 10^{31}$ y, $τ({^{12}\rm{C}}\rightarrow{^{12}\widetilde{\rm{N}}}+ e^- + \widetilde{ν_e}) \geq 6.4 \times 10^{30}$ y and $τ({^{12}\rm{C}}\rightarrow{^{12}\widetilde{\rm{B}}}+ e^+ + ν_e) \geq 6.6 \times 10^{30}$ y (90\% C.L.). The upper limits on the relative strengths for the non-Paulian electromagnetic, strong, and weak transitions have been obtained: $δ^2_γ\leq 1.0\times 10^{-57}$, $δ^2_{N}\leq 7.0\times 10^{-61}$ and $δ^2_β\leq 9.6\times 10^{-36}$, all at 90\% C.L..

In arXiv:2603.18175, the authors argue, based on numerical studies of particular cases, that the quantum damping of cosmological shear in a modified loop quantum cosmological model (mLQC-I) that was recently found in arXiv:2510.14021 is not generic and that the universe never becomes truly classical. In this brief Note, we revisit these claims by carefully examining the underlying assumptions and the class of initial conditions considered. We show that the examples analyzed in arXiv:2603.18175 correspond to configurations that do not represent physically admissible collapsing Bianchi I universes, as they involve mixed expanding-contracting directions and lead to effectively lower-dimensional post-bounce geometries. Restricting to physically relevant initial conditions corresponding to genuine three-dimensional contraction, we find that the quantum damping of cosmological shear is a robust dynamical feature. This conclusion is supported by both numerical and perturbative analyses, which demonstrate that the post-bounce evolution admits an isotropic attractor, with anisotropies decaying exponentially and independently of the matter content, provided that the weak energy condition is satisfied. We further outline a plausible post-bounce mechanism for the onset of classicalization.

Given a closed Riemannian Spin manifold $(M,g)$ of dimension greater or equal than four, we consider a generalized conformally invariant equation involving the Dirac operator with a non-linearity of convolution type. We show that the Aubib-type inequality corresponding to the problem is always strict, unless $(M,g)$ is conformal to the round sphere. In particular, this result provides an existence result for a ground state to the conformal Dirac-Einstein problem in dimension four. We point out that aside from some perturbative or special cases, this presents the first general existence result for the conformal Dirac-Einstein equations in dimension four.

We study the duality of lattice Maxwell theory in a modified Villain formulation and employ an ultra-local action with a theta term. Although this action is known to become non-local under Poisson resummation, we show that this non-locality can be removed in the absence of monopoles by incorporating a non-local transformation procedure into the definition of the S-transformation. As a result, the ultra-local action with a theta term exhibits an exact SL(2,Z)-duality. We further analyze the SL(2,Z)-structure of Wilson loops, demonstrating that they transform properly up to a nontrivial phase factor arising from the nontrivial self-linking of magnetic Wilson loops. This effect originates from the non-local transformation procedure in the S-transformation. Remarkably, the resulting SL(2,Z)-structure closely resembles that of non-spin Maxwell theory.

Discrete symmetries of Calabi-Yau moduli spaces, generated by isomorphic flops, constrain the instanton expansion of the 4D $\mathcal{N}=2$ Type~IIA prepotential. We show that the Coxeter-invariant functions into which the prepotential organizes are eigenfunctions of a Laplace-Beltrami operator built from the Coxeter-invariant symmetric bilinear form on the moduli space. This means that the Gromov-Witten expansion can be interpreted as a superposition of waves propagating on the Coxeter quotient of the moduli space, and its resummation is the corresponding spectral decomposition. For the dihedral Coxeter groups, separation of variables in the eigenvalue equation explains from first principles why special modified Bessel functions, ordinary Bessel functions and Jacobi theta functions appear as the natural building blocks of the prepotential, depending on whether the Coxeter rotation acts hyperbolically, elliptically, or parabolically. The resulting spectral representations converge efficiently in the interior of the moduli space, complementing the standard large-volume instanton expansion.

We investigate the axion quality problem in warped extra-dimensional models in which the QCD axion arises as the Wilson-line mode of a five-dimensional $U(1)$ gauge field compactified on an $S^1/\mathbb{Z}_2$ orbifold. Higher-dimensional gauge invariance severely constrains possible sources of Peccei--Quinn symmetry breaking, implying that non-QCD contributions to the axion potential are predominantly generated by nonlocal effects mediated by $U(1)$-charged fields propagating along the compact dimension. We systematically compute these contributions and examine how both the warped geometry and the orbifold fixed points (branes) affect the resulting axion quality. Finally, we classify the parametric suppression of the induced axion potential, thereby identifying the conditions under which warped extra-dimensional axions can achieve sufficiently high quality.

We investigate the sensitivity of a future 10 TeV muon collider to dimension-6 operators in the Standard Model Effective Field Theory (SMEFT), focusing on Higgs and top quark production processes. The analysis includes two-fermion and four-fermion operators that induce electroweak vector and axial-vector interactions, as well as dipole, scalar, and tensor interactions involving muons. Many of these operators are only weakly constrained or difficult to probe at the LHC due to limited sensitivity and large SM backgrounds. We study the processes $μ^+μ^- \to Zh$, $μ^+μ^- \to μ^+μ^-h$, $μ^+μ^- \to t\bar t$, and $μ^+μ^- \to t\bar t h$, exploiting the energy-enhanced SMEFT effects at multi-TeV scales accessible to a muon collider. Using detailed simulations that incorporate differential information and angular distributions, we derive projected bounds on the relevant Wilson coefficients. We find that a 10 TeV muon collider can strengthen existing limits on muon-Higgs-gauge and muon-top interactions by up to an order of magnitude, surpassing even FCC-ee projections. Finally, we interpret these bounds in the context of representative UV scenarios, including models with vector-like lepton and scalar leptoquarks, highlighting the potential of a muon collider to probe new physics at scales well beyond the LHC reach.

The tidal response of compact objects provides a powerful probe of their internal structure and of the surrounding gravitational field. We provide a comprehensive and unified overview of tidal effects in black holes, neutron stars, and exotic compact objects, with emphasis on both static and dynamical responses to external fields, encoded in Love numbers and dissipation numbers. We discuss the vanishing of static bosonic Love numbers for black holes in vacuum General Relativity, their modifications in alternative theories, in non-standard models of compact objects, and in the presence of matter, as well as their role in testing deviations from Einstein's theory and environmental effects. A fundamental distinction between bosonic and fermionic perturbations is highlighted, as the latter yield nonzero static Love numbers even for black holes in General Relativity. For neutron stars, we overview the dependence of tidal Love numbers on the equation of state, the emergence of quasi-universal relations, and the impact of rotation, nonlinearities, and dynamical effects. Exotic compact objects typically feature nonvanishing static tidal Love numbers -- a striking observational signature that differentiates them from black holes. We further review how tidal effects influence the gravitational-wave signals from binary inspirals, and explore their implications for gravitational-wave astronomy. In particular, we stress their significance for current and future detectors as tools to test General Relativity, constrain the nuclear equation of state, and probe the fundamental nature of compact objects and their environments.

We consider mixed four-point correlators of 1/2-BPS operators $φ_{k}$ in SCFTs with eight real Poincaré supercharges, namely the 3d $\mathcal{N}=4$, 4d $\mathcal{N}=2$, 5d $\mathcal{N}=1$, and 6d $\mathcal{N}=(1,0)$ theories. Using the basis of solutions to the superconformal Ward identity introduced in arXiv:hep-th/0405180, we demonstrate that the dynamical data in mixed correlators of extremality $\mathcal{E}=2$ is encoded in certain simpler ``reduced correlator" functions that admit a block expansion, in close analogy to the recent result for maximally supersymmetric CFTs in arXiv:2601.15407. These reduced blocks similarly involve ordinary blocks with shifted kinematics and reproduce what is known in 4d, generalize a known example in 6d, and offer novel results in 3d and 5d.

We study how a fluctuating domain wall in three dimensions modifies bulk observables in a gapped phase. We introduce an effective interaction between the wall and the lightest bulk massive mode, and identify the regime in which this description is controlled: nearly on-shell bulk exchange with small momentum along the wall. In this regime, several observables are controlled by a renormalized dimensionless coupling $λ$, including the large-$L_z$ correction to the wall free energy and the large-$|x_\perp|$ tail of two-point functions in the presence of the wall. Other observables, such as one-point functions and two-point functions in the nearby-regime, retain non-universal dependence on operator data and on the bulk spectral density. We test the universal kinematic consequences of wall fluctuations, and find good agreement with the predicted rough-wall broadening and nearby Gaussian behavior in Monte Carlo simulations of the 3D Ising model with anti-periodic boundary conditions.

We study the loss of hyperbolicity of perturbation equations for black hole solutions of scalar Gauss-Bonnet gravity. We consider a class of coupling functions allowing for static black hole solutions with arbitrary small masses. For masses below a minimum value, such solutions become unphysical, because the perturbation equations become elliptic; this arguably corresponds to the loss of validity of the effective field theory. We analyse the dependence of this minimum mass on the parameters of the theory, finding that with an appropriate choice of the coupling function, such mass can be chosen arbitrarily small. However, this does not correspond to larger deviations from general relativity, since observable quantities like the black hole scalar charge are bounded by above.

We study the exclusive electroproduction process $e+p\to e'+p'+X$, with $X$ a single-particle final state, in the forward-proton kinematics relevant for the future Electron-Ion Collider (EIC). We develop a unified $2\to 3$ framework that provides the full event kinematics and incorporates pseudoscalar and vector mesons, as well as axion-like particles and vector mediators such as dark photons. It is based on phenomenological amplitudes constrained by existing photo- and electroproduction data and constructed to admit systematic refinement as new measurements become available. To benchmark the framework, we compare its predictions to flux-factorized descriptions based on the equivalent-photon approximation, demonstrating close agreement for total rates and selected single-differential distributions in the near-real regime, while highlighting the role of finite-$Q^{2}$ correlations for multi-differential observables at larger photon virtualities. As a case study, we perform a detailed kinematic analysis of the missing-proton-energy signature, illustrating how the full $2\to 3$ treatment informs forward-proton acceptance and signal selection in realistic EIC configurations.

We study how Sachdev-Ye-Kitaev (SYK) interactions can arise from localized single-particle states on a system that is effectively one dimensional. If a local interaction is projected onto coarse localized orbitals, the resulting couplings do not immediately follow the standard SYK distribution. Instead, they have a finite probability of being exactly zero, a broad non-Gaussian distribution for the nonzero values, and strong correlations coming from the geometry of the localized states. We then show that this changes when each localization volume is resolved into $M>1$ smaller microscopic pieces with random phases. As $M$ increases, the distribution of the nonzero couplings moves toward the complex-Gaussian SYK form. At the same time, the large-$M$ limit is a sparse but asymptotically canonical SYK network: the nonzero couplings create SYK clusters, while the pattern of missing or very weak couplings is still determined by the real-space overlap of the localized orbitals. Finally, we map the interaction tensor to a graph in pair space. This makes it possible to follow the formation, merger, and growth of SYK clusters, which we characterize using connected components and clique/simplex counts. The result is a minimal real-space phenomenological theory of SYK-cluster formation, providing clear experimental criteria.

We uncover a combinatorial structure governing the differential equations satisfied by wavefunction coefficients of scalar fields with generic masses in de Sitter space. Using an integral representation of the massive mode functions, we express the Feynman integrals underlying cosmological correlators as twisted integrals of rational functions. In this formulation, the integrals belong to a finite set of master integrals obeying a first-order system of differential equations, which can be derived efficiently in the time-integral representation. We show that these equations admit a simple graphical description in terms of graph tubings, which encode the couplings among basis functions and the evolution of singularities. This structure provides an efficient algorithm to derive the differential equations, and a boundary-centric perspective on massive cosmological correlators in which their analytic structure emerges from underlying combinatorial data. As an illustration, we solve the system in the limits of small and large masses.

We provide a complete first-principles based discussion of quantum tunnelling out of initial states carrying a conserved Noether charge. Our main result is a simple, unambiguous Euclidean-time prescription for the calculation of tunnelling rates out of such states. By relying on a combination of the direct approach and the steadyon framework for the evaluation of real-time path integrals, our derivation offers full transparency of its underlying assumptions, and is independent of any ad-hoc generalisations. This strategy also offers a simple explanation for the emergence of complex saddle points for such systems, justifying techniques postulated by earlier works. Our analysis furthermore offers the first results for initial states with both a conserved Noether charge and a non-trivial energy. We first illustrate the main conceptual points of our analysis for the simple example of a point particle in two spatial dimensions carrying a conserved angular momentum. Then, we generalise our results to the case of multiple dimensions and an arbitrary conserved Noether charge, providing an easy-to-implement prescription for the calculation of the tunnelling rate. We furthermore apply our results to the example of a complex scalar field subject to a global U(1)-symmetry with associated charge. These results provide a reliable foundation for the calculation of tunnelling rates in applications in finite-density and charge-asymmetric systems.

We present a new mechanism for addressing the cosmological constant problem based on global constraints arising from a lapse function in a higher-dimensional gravitational theory. Inspired by Horava-Lifshitz gravity, we consider a 5d spacetime with anisotropic scaling along a compact extra dimension, while preserving Lorentz invariance in four dimensions. In the deep infrared limit, variation with respect to the lapse generates a global constraint on the 4d geometry, closely analogous to that of vacuum energy sequestering. Although the resulting effective gravitational equations differ from standard sequestering, radiative contributions to the Standard Model vacuum energy are nevertheless cancelled at all orders.

Motivated UV frameworks generically predict the existence of multiple axion fields. Their interplay gives rise to novel collective phenomena - including level crossings and the formation of string bundles - which modify the predicted mass and couplings of the QCD axion as a solution to both the strong CP problem and the observed dark matter abundance. Among these effects, the domain wall number is determined by the full anomaly structure of the theory: in the single axion case, the absence of long-lived domain walls imposes $E/N \geq 8/3$ as a theoretical bound on the QCD axion photon coupling, assuming the global structure of the Standard Model gauge group is minimal. We show that this bound can be relaxed in the multi-axion framework. Combined with the fact that the QCD axion can become a subdominant dark matter component, this might render multi-axion scenarios experimentally challenging. Nevertheless, a careful analysis of the parameter space reveals that in most regions where the QCD axion evades detection, an axion-like particle remains visible to next-generation experiments. When all signals fall below future projections, we identify the most promising regions of parameter space to probe in an illustrative two-axion setup.

Conformal supergravity provides an effective off-shell formalism to study higher derivative actions. We show that the $D=4$, $\mathcal{N}=2$ theory admits equivariantly closed forms. These may be used to compute closed-form expressions for supersymmetric observables in a general class of supergravity theories with higher derivative couplings, without any need to solve equations of motion. We discuss applications to holography, presenting results for on-shell actions that are conjecturally valid to all orders in the perturbative $1/N$ expansion.

The Love numbers of a gravitating body are response coefficients encoding its tidal deformability. In compact binary systems, they appear in the gravitational waveform during the inspiral phase and will be measurable by upcoming gravitational-wave observatories. This review provides a comprehensive and pedagogical account of the theoretical foundations of Love numbers and surveys the most recent advances in the study of tidal effects in compact objects, with particular emphasis on black holes and neutron stars. We begin with a gentle introduction to tidal effects in Newtonian gravity, leading into a discussion of how to robustly define tidal responses in general relativity using the effective field theory and post-Newtonian frameworks. After an overview of the perturbation theory of black holes and neutron stars, we review the computation of Love numbers and dissipative response coefficients in a wide range of settings, including the static, dynamical, and nonlinear tidal responses of Kerr black holes and neutron stars in four-dimensional general relativity. We further discuss the extension of these results to charged black holes and a number of "new physics" scenarios, including higher-dimensional black holes and other black objects, (anti-) de Sitter black holes, supergravity black holes, and theories beyond general relativity. Finally we provide an overview of the tantalizing zoo of hidden symmetries of general relativity that have been uncovered in the attempt to explain the famous vanishing of static black hole Love numbers.

We propose a novel cosmological scenario in which baryonic neutron stars could plausibly form in the early universe. If baryogenesis initially produces an excessively-large baryon asymmetry, $Y_B \gg 10^{-10},$ the baryonic mass inside the horizon can exceed the minimum neutron star mass before big bang nucleosynthesis (BBN). While this large asymmetry is present, non-relativistic baryons can dominate the universe and enhanced density perturbations on small scales can gravitationally collapse Hubble patches shortly after horizon re-entry. For some initial perturbations, just below the threshold for black hole formation, this collapse will be arrested only by nuclear pressure, possibly resulting in neutron star formation. Afterwards, there must be a large entropy injection to restore the observed baryon asymmetry, $Y_B \sim 10^{-10}$, and preserve the successful predictions of standard BBN. Unlike neutron stars that form from stellar collapse, primordial neutron stars can, in principle, be as light as $\sim 0.1 M_\odot$, limited only by the nuclear equation of state.

We lay the foundations for a quantum formalism describing the real-time evolution of particles in the Glasma phase of a heavy-ion collision, focusing on the implications of gauge invariance in the definition of the momentum of a particle in a classical background field. We first establish the correspondence between the classical Wong's equations and the Heisenberg equations of motion for a particle in a classical non-Abelian background field. Using this correspondence, we obtain equations of motion for both the kinetic momentum -- the gauge invariant, physically measurable quantity -- and the canonical momentum, which is conjugate to the coordinates in the Hamiltonian. In particular, the kinetic momentum broadening receives non-trivial contributions from the transverse field components, even in the eikonal limit. Finally, we demonstrate that imposing a transverse Coulomb gauge condition at the initial time significantly reduces the accumulation of numerical errors, thereby providing an optimized framework for the forthcoming quantum implementation.

Recently, it has been shown that wave function coefficients (WFCs) admit a natural description in terms of the orthogonal Grassmannian, furnishing homogeneous solutions to the three-dimensional conformal Ward identities in spinor-helicity variables. This, however, presents a challenge for WFCs of conserved currents, which satisfy inhomogeneous Ward identities; correspondingly, the Grassmannian construction reproduces only their \textit{discontinuities}. In this paper, we show that $\mathcal{N}=2$ supersymmetry, by relating spinning and non-spinning WFCs, leads to a Grassmannian formula augmented by a kinematic prefactor that captures the full WFC. Moreover, we show that the positive and negative branches of the Grassmannian formula admit a natural interpretation in terms of supersymmetric invariants, and give rise to distinct helicity amplitudes in the flat-space limit.

Carrollian amplitudes are scattering amplitudes of massless particles written in position space at null infinity. We study various aspects of Carrollian amplitudes for gauge theory and gravity at loop level using primarily the modified Mellin prescription of [1]. Finite one-loop four-point Carrollian amplitudes in gauge theory are shown to maintain an analytic structure similar to tree level results. We compute the one-loop four-point Carrollian MHV amplitudes in planar $N=4$ super Yang-Mills theory, which are expressed as differential operators acting on tree level Carrollian amplitudes. This result is generalized to all loop orders using the Bern-Dixon-Smirnov (BDS) formula. Similar structures are observed at one-loop for Carrollian MHV amplitudes in $N=8$ supergravity. We next consider $2\to 2$ scattering of massless scalars via gravitational interactions in the eikonal regime and show that the corresponding Carrollian amplitudes exhibit logarithmic behavior in the `Carroll time' $u$. We compute the discontinuities of these Carrollian amplitudes up to $O(G^3)$ and show that they are descendants of Carrollian Born amplitudes. We observe similar logarithmic behavior in Carrollian amplitudes associated with the one-loop scalar box diagram. The dependence of this amplitude on dual scaling dimensions also differs from standard tree level results. Finally, we further study the infrared (IR) divergences of Carrollian amplitudes in massless scalar QED, gravity, and Yang-Mills theory. We show that Carrollian amplitudes in these theories naturally factorize, allowing us to provide an IR-safe definition for these objects.

Muon-electron universality is tested in $B^0 \to K^{*0} \ \ell^+ \ell^-$ decays, in the dilepton-invariant-mass region above the $ψ(2S)$ resonance. The analysis uses beauty mesons produced in proton-proton collisions recorded by the LHCb detector at center-of-mass energies of 7, 8, and 13 $\text{TeV}$, corresponding to an integrated luminosity of 9 $\text{fb}^{-1}$. The ratio of branching fractions between the muon and electron channels, $R_{K^{*0}}$, is measured to be $1.08\,^{+0.14}_{-0.12}\text{(stat)} \ \pm 0.07\text{(syst)}$ for a dilepton-invariant-mass squared above 14.0 $\text{GeV}^{2}/\text{c}^{4}$, consistent with the Standard Model prediction. This result represents the most precise measurement of $R_{K^{*0}}$ in this region and the first such measurement performed at a hadron collider.

The present research applies Graph Neural-Networks (GNNs) for energy measurement and particle identification tasks for a proposed second detector at the future Electron Ion Collider (EIC). In particular, an iron-scintillator sampling calorimeter would provide neutral hadron ($K_L$ and neutron) energy measurements and identification, as well as separation of muons from hadrons. Using detector simulations, particle hits in the detector are represented as graphs, and a GNN is trained for either classification or prediction. Furthermore, we developed a parameterization of the scintillator optical photon simulation that yields a 20-fold speed up compared to the default simulation. We find that the GNN method outperforms classical methods at the same tasks, and we report projections for the energy and timing resolution, and identification accuracy of the calorimeter. We also present an integration of the GNN method into a Multi-Objective Optimization framework, enabled by an automated pipeline of data generation, GNN training, and detector performance evaluation. We utilize the optimization to quantify the tradeoffs between different performance metrics at high and low energies when changing the detector design parameters, such as the iron/scintillator thickness.

We investigate an extended 3-4-1 model consisting of a new singly charged Higgs boson, implementing the minimal inverse seesaw mechanism to accommodate the large values of the $(g-2)_{e,μ}$ anomalies as well as the lepton-flavor-violating decay rates of charged leptons, the Standard Model-like Higgs boson, and the $Z$ boson, all consistent with current experimental data. Unlike the previously studied 3-4-1 realization, the model considered here predicts strong correlations among these observables that can be tested in future experiments. In particular, the current upper bound on Br$(τ\to μγ)$ imposes a stringent constraint compatible with the $1σ$ experimental range of $(g-2)_μ$, corresponding to a maximal deviation of $10^{-9}$ from the SM prediction. The forthcoming experimental sensitivity to Br$(τ\to μγ)$ will reduce this deviation to $5\times 10^{-10}$.

In this paper we present a systematic construction of an(isotropic) black brane solutions in arbitrary spacetime dimensions $D$ in particular, with Lifshitz-like asymptotics. Two distinct holographic models are considered. The first model involves a scalar field with a potential coupled to two Maxwell fields, allowing for both electric and magnetic charges. The second model includes a scalar field, a Maxwell field, and a three-form field strength of a Kalb-Ramond field. For each model, exact solutions for the metric, scalar field, gauge fields, and coupling functions are derived, incorporating anisotropic scaling exponents and general warp factors, including Gaussian forms. The results generalize previously known five-dimensional anisotropic black brane solutions to arbitrary dimensions. We show that the third law of thermodynamics, which requires entropy to vanish as temperature approaches zero, is satisfied for a certain range of parameters in both models. However, for specific warp factors or coupling constants, the entropy-temperature relation exhibits non-monotonic or multi-valued behavior, suggesting the possibility of phase transitions and a violation of the third law.

We develop a procedure to solve Einstein-dilaton-Maxwell models in quadratures using the potential reconstruction approach. We then apply this procedure to three distinct models, examining both the null energy condition (NEC) and the validity of the third law of thermodynamics in each case. The explicit knowledge of the blackening function -- as opposed to relying solely on numerical data -- allows us to discuss the validity of the third law in detail. The three models considered are: (I) a 5D model with two Maxwell fields, featuring anisotropy specified by a Gaussian function and a Lifshitz function; (II) the same 5D model as in (I), but with anisotropy parametrized by two Lifshitz parameters; and (III) a 6D model with one 2-form and one 3-form field, with the metric parametrized by two Lifshitz parameters. We show that for models I and II the parameter regions, where both the NEC and the third law are satisfied, exhibit no correlation between the two conditions. In contrast, for model III the validity of the NEC implies the validity of the third law.

The first search for the lepton-flavour violating decays $B^+ \to π^+ μ^{\pm} e^{\mp}$ in proton-proton collisions is presented, using data collected by the LHCb experiment between 2011 and 2018, corresponding to an integrated luminosity of 9 fb$^{-1}$. No significant signal is observed and an upper limit on the branching fraction is set at $\mathcal{B}(B^+ \to π^+ μ^{\pm} e^{\mp}) < 1.8 \times 10^{-9}$ at the $90\%$ confidence level, two orders of magnitude more restrictive than the current world average. This is the first constraint on lepton-flavour violating $b \to d$ quark transitions at the LHC and also sets the most stringent upper limits to date on $b \to d μ^{\pm} e^{\mp}$ transitions. Limits on left-handed and scalar scenarios beyond the Standard Model are also reported.

We introduce a new method for the determination of the ratio of production fractions $R^{\pm0}=\mathcal B(Υ(4S)\to B^+B^-)/\mathcal B(Υ(4S)\to B^0\bar B^0)$ based on $\bar B\to D^{(*)}\ell\bar ν$ decays. Given the importance of these modes, we perform a comprehensive analysis of the available data, extracting the information on their branching fractions and \Rpmz in parallel and providing their correlations in order to avoid double-use of this information in phenomenological analyses. We obtain the most precise value for $R^{\pm0}$ from a single channel so far, about 2$σ$ from unity. The combination with previously available determinations from other channels yields $R^{\pm0}=1.062(19)$, constituting evidence for isospin violation in $Υ(4S)$ decays. This demonstrates the necessity to take this effect into account in experimental and phenomenological analyses. The results for the $\bar B\to D^{(*)}\ell\bar ν$ branching fractions are up to $1.6σ$ larger compared to averages available in the literature, owing to the removal of overlooked inconsistencies in the treatment of older analyses and correcting for d'Agostini bias where possible, thereby reducing the $V_{cb}$ puzzle.

We present the first lattice QCD calculation of the next-to-leading order (NLO) hadronic vacuum polarization (HVP) contribution to the muon anomalous magnetic moment with sub-percent precision. We employ the time-momentum representation combined with the spatially summed vector correlator computed on CLS ensembles with $N_{\mathrm{f}}=2+1$ flavors of $\mathrm{O}(a)$-improved Wilson fermions, spanning six lattice spacings ($0.039$-$0.097\,$fm) and a range of pion masses including the physical value. After accounting for finite-size corrections and isospin-breaking effects, we obtain in the continuum limit $a_μ^{\mathrm{hvp,\,nlo}} = (-101.57 \pm 0.26_{\rm stat} \pm 0.54_{\rm syst}) \times 10^{-11}$, corresponding to a total relative error of 0.6$\%$. Our result lies 1.4$σ$ below the estimate of the 2025 White Paper update and is two times more precise. It also shows a tension of $4.6σ$ with data-driven evaluations based on hadronic cross section measurements prior to the CMD-3 result.

We perform a comprehensive study of a certain class of discrete symmetries of families of Feynman integrals, defined as affine changes of variables that map different sectors of the family into each other. We show that these transformations are always encoded into permutations of the Feynman parameters that relate the Lee-Pomeransky polynomials of the two sectors, irrespective of the integral representation used to define the Feynman integrals. We then construct an affine map in loop-momentum space that encodes such a permutation. We also show that these symmetries can be naturally embedded into the framework of twisted cohomology theories, and the period and intersection parings are invariant under the symmetry transformations. If we focus on symmetries within a fixed sector, we obtain a group acting on the twisted cohomology group, and we study the decomposition of this action into irreducible representations. One of our main mathematical results is that the character of this representation is proportional to the Euler characteristic of the corresponding fixed-point set. We then study the implications for Feynman integrals, in particular for the intersection matrix in a canonical basis. We also present a formula for the number of master integrals in a given sector in the presence of a non-trivial symmetry group in terms of the Euler characteristics of fixed-point sets. As an application, we obtain the numbers of master integrals for banana integrals with up to four loops for arbitrary configurations of non-zero masses. In order to achieve our results, we had to combine tools from various different areas of mathematics, including graph theory, group theory and algebraic topology.

Black hole thermodynamics has recently witnessed three distinct classification schemes: based on local geometric properties of the temperature function, global topological invariants, and Riemann surface foliations in the complex plane. We show that these schemes are equivalent in the real domain via two dictionaries: one linking thermal stability to the monotonicity of the temperature curve, and the other connecting the number of black hole states to the foliation number of a Riemann surface. The number of extremal points of the temperature curve determines the classification in all three frameworks, tracing this unification to the critical point structure of the black hole solution space. As an illustration, several black holes demonstrate how counting extrema yields topological invariants and phase transition information. This unified framework simplifies black hole thermodynamic analysis and provides a foundation for exploring more complex black holes.

We present a detailed investigation of the afterpulse effect in silicon photomultipliers (SiPMs), using a dedicated structure with single-cell readout. This enables direct measurement of intrinsic device properties and observation of individual pulses even after irradiation. Three independent analysis methods to quantify afterpulse induced events were developed and validated by Monte Carlo simulations. The first method is based on charge integration, while the other two methods use multiple linear regression to reconstruct transient waveforms and accurately identify individual pulse positions. These positions are then used either as direct event counts or to construct time interval distributions, enabling comprehensive characterization of the afterpulse probability and providing insights into the dynamics of trapping in silicon. Using this framework, we measured three SiPM samples: one fresh reference device and two irradiated devices exposed to reactor neutron fluences of 2e12 and 1e13 cm^-2. We report systematic measurements of the afterpulse probability and time constant as functions of bias voltage and irradiation fluence. For overvoltages in the range of 3 to 5 V above breakdown, the afterpulse time constant is found to be below 10 ns and the afterpulse probability below 6%. Both quantities show no significant dependence on irradiation fluence.

The T2K long-baseline neutrino oscillation experiment in Japan continues to lead the search for leptonic charge-parity violation while providing precision measurements of mixing and mass splitting parameters. Central to this programme is the mitigation of systematic uncertainties through the near detector complex, which provides high-statistics neutrino-nucleus interaction cross-section measurements across various targets. This contribution presents the latest T2K oscillation results, incorporating the first data with a gadolinium-loaded far detector, and highlights several recent cross-section measurements, including several world-first measurements of rare interaction channels. Together, these results demonstrate the vital synergy between interaction modelling and oscillation analysis in the search for charge-parity violation in the T2K-II era.

We study the impact of multiple vector-like fermions in SU(5) grand unified theory (GUT). Threshold effects from extra fermions allow the observed gauge couplings to be consistently matched to a single unified gauge coupling, and typically raise the unification scale to $M_\mathrm{GUT}\simeq 10^{15.5}\,\mathrm{GeV}$. Because the Standard Model fermions arise as admixtures of several GUT multiplets, the nucleon decay operator coefficients are further suppressed, leading to longer proton lifetimes than in conventional GUTs. We also find that the admixture of multiple GUT multiplets relaxes the rigid Yukawa relations of conventional GUTs and alleviates the bottom-tau unification problem. Overall, our analysis demonstrates that multi-fermion SU(5) GUTs provide a testable framework that simultaneously reconciles gauge coupling unification, realistic flavor structures, and proton stability. Our results highlight the importance of probing multiple proton-decay channels in next-generation experiments such as Hyper-Kamiokande to critically test this scenario.

We prove positivity of a weighted holographic energy for four-dimensional spacetimes with negative cosmological constant whose conformal boundary at infinity is conformally static and admits either spherical sections, or toroidal sections with compatible spin structure.

We construct a consistent anti-de Sitter completion of the static and spherically symmetric black-hole solution sourced by the Palatini-inspired nonlinear electrodynamics \(Y^n\) model. Starting from the Einstein--Hilbert action with a negative cosmological constant and the first-order PINLED sector, we derive the full set of field equations and show that the nonlinear electromagnetic solution preserves its original parametric structure, while the lapse function acquires the standard AdS contribution. We then analyze the horizon structure, Hawking temperature, extended phase-space thermodynamics, and the associated equation of state. In addition, we investigate null and timelike geodesics, with emphasis on the effective potentials, photon sphere, shadow radius for a static observer at finite distance, and innermost stable circular orbit. The resulting framework furnishes the exact AdS extension of the asymptotically flat PINLED black hole and provides a coherent basis for numerical and phenomenological studies of its thermodynamic, optical, and orbital properties.

In this paper, we use a version of the BF formulation of two-dimensional dilaton gravity that allows to define a gauge theory of the two-dimensional Poincaré or Maxwell algebras and several of their higher-spin generalisations, both of finite and infinite dimension. The spectrum of the two-dimensional higher-spin gravity with vanishing cosmological constant based on the extended, infinite-dimensional higher-spin algebra is shown to contain an infinite collection of scalar degrees of freedom with a continuum of ever increasing mass, corresponding to the twisted-(co)adjoint representation. We comment on an approach to include backreaction of the scalar fields on the gravity sector at the level of formal equations of motion, thereby providing a first example of a fully interacting higher-spin gravity theory with vanishing cosmological constant in two dimensions.

The dense environment of our Galactic Center (GC) offers a unique laboratory for probing ultralight dark matter (ULDM). We explore the prospect of detecting a scalar ULDM field through its effects on the orbital dynamics of S-stars around the supermassive black hole in the GC, Sgr A$^*$. We consider both linear and quadratic couplings between the real scalar field $φ$ and Standard Model particles, and analyze two representative ULDM structures: the scalar gravitational atom and the spherical soliton. We find that quadratic coupling induces a non-oscillatory perturbation, leading to a long-term secular orbital evolution. We use the observed periastron precession rate of S2 star to put stringent constraints on the total ULDM mass in the GC and the quadratic coupling constant. For the gravitational atom $|211\rangle$ state, we constrain the mass ratio of ULDM to Sgr A$^*$ to $β\lesssim 10^{-3}$ at $m \sim 10^{-18}$ eV, and for the spherical soliton which extends to $\sim 0.2\,$pc, the mass ratio is limited to $β\lesssim 1$ at $m \sim 3\times10^{-20}$ eV. Notably, the resulting limits on the quadratic coupling constant surpass current bounds in the mass range $10^{-20} \,\text{eV} \lesssim m \lesssim 10^{-18}$ eV.

In this note, we show that the well-known leading low-temperature correction to the Heisenberg-Euler Lagrangian in a constant electromagnetic field arising at two loops can be efficiently extracted from its one-loop zero-temperature analogue. Resorting to the real-time formalism of equilibrium quantum field theory that explicitly separates out the zero-temperature contribution from the finite-temperature corrections the determination becomes essentially trivial. In essence, it only requires taking derivatives of the Heisenberg-Euler Lagrangian at one loop and zero temperature for the field strength. As a bonus, we then effectively dress the low-temperature contribution at two loops by one-particle reducible tadpole structures. This generates a subset of higher-loop contributions to the Heisenberg-Euler Lagrangian in the limit of low temperatures. We extract their leading strong-field behavior at a given loop order, and finally resum these to all loop orders.

We study the heavy quark dynamics in the presence of memory within the framework of a generalized Langevin equation. Time correlated thermal noise with power-law decay is generated by a fractional differential equation, formulated using the Caputo fractional derivative with order parameter $ν$. The effect of memory is calculated through the momentum correlation, the time evolution of the average squared momentum, the average squared displacement, and the average kinetic energy. The effect of memory is further studied for the higher normalised central moments of the heavy quark transverse-momentum distribution. The results indicate that time correlated thermal noise substantially influences heavy quark dynamics in the quark gluon plasma.

The yields of prompt and non-prompt J/$ψ$ and the fraction of non-prompt J/$ψ$ are measured at midrapidity ($|y| < 0.9$) via the dielectron decay channel as a function of the midrapidity charged-particle multiplicity ($|η| < 0.9$) in pp collisions at $\sqrt{s} = 13$ TeV. The J/$ψ$ yields and the multiplicity are normalized by their average value in inelastic collisions. The multiplicity-dependent yield ratio between prompt J/$ψ$ and D$^0$ is reported. The multiplicity is further divided into three azimuthal regions with respect to the J/$ψ$ momentum: toward the J/$ψ$ emission direction, transverse, or opposite to it. A stronger-than-linear increase of the self-normalized yields is observed for both prompt and non-prompt J/$ψ$ production, with similar trends. This behaviour is also observed in the toward region, while a weaker increase is observed in the transverse and away regions.

We develop a systematic classification of asymmetric $\mathbb{Z}_2$ orbifold actions in Pati--Salam heterotic string vacua constructed in the free fermionic formulation. Starting from symmetric $\mathbb{Z}_2 \times \mathbb{Z}_2$ orbifold vacua with an $SO(10)$ GUT, we allow the Pati--Salam breaking vector to act asymmetrically on the internal degrees of freedom. The asymmetric orbifold action freezes geometrical moduli whilst inducing doublet--triplet splitting in the untwisted sector. Notably, this doublet--triplet splitting operates for any asymmetric action, including pure asymmetric shifts that preserve all geometric moduli, and is therefore independent of moduli stabilisation. Classifying the breaking vector according to its twist action, we find six inequivalent classes of geometric moduli spaces characterised by 12, 8, 4 or 0 real untwisted moduli. Through combining these asymmetric twists with all compatible asymmetric shifts, 24 inequivalent cases are identified and characterised by their residual moduli content and internal Narain lattice. For each case we construct representative basis sets admitting three chiral generations, providing the starting point for further classification within each class. We perform explicit GGSO phase enumerations in representative model classes with 12, 8, 4 and 0 moduli, classify the resulting $\mathcal{N} = 1$ and $\mathcal{N} = 0$ vacua according to phenomenological criteria and identify exophobic, phenomenologically viable models. We compute the partition function and corresponding one-loop vacuum energy at the free fermionic point in moduli space for each phenomenologically viable model across the four classes. As the number of geometrical moduli decreases, the number of distinct partition functions for these vacua collapses to a small number, reflecting a pronounced degeneracy under GGSO phase variations.

We investigate quantum field theory in Rindler space with a UV cutoff by considering a free scalar field on a lattice in Rindler coordinates. We find that the Minkowski vacuum is not exactly thermal with respect to the local lattice Rindler Hamiltonian. Nevertheless, for observables sufficiently far from the horizon, the Wightman function and the Unruh--DeWitt detector response reproduce the expected thermal behavior in the continuum limit. Thus, the Unruh effect survives operationally, even though exact thermality is lost at the state level. We also show that the Rindler vacuum energy density reproduces the standard continuum behavior away from the horizon, while the UV singularity at the horizon is replaced by a stretched-horizon contribution. Furthermore, the retarded Green function exhibits a component reflected at the stretched horizon, implying that an ingoing wave packet is reflected at a proper distance of order the cutoff. This provides an effective brick-wall picture in the UV-regulated theory. Our analysis suggests that, once a cutoff is introduced, the global Minkowski description and the wedge description based on a local Rindler Hamiltonian are no longer equivalent at the operator level.

We study the ODE/IM correspondence for the ordinary differential equation associated with the affine Lie algebra $E_6^{(1)}$. The WKB expansion of the solution of the ODE is performed by the diagonalization method, and the period integrals of the WKB coefficients along the Pochhammer contour are calculated. We also compute the integrals of motion on a cylinder in two-dimensional conformal field theory with W-symmetry associated with $E_6^{(1)}$. Their eigenvalues on the highest-weight state are shown to agree with the period integrals up to the sixth order.

We investigate a class of nonminimal derivative couplings between fermions and electromagnetic fields that generate Rashba-like spin--orbit interactions in one-dimensional quantum rings. Starting from a generalized Dirac Lagrangian containing two independent axial structures built from the field strength $F_{μν}$ and its dual $\tilde{F}_{μν}$, we perform a systematic nonrelativistic expansion and show that both couplings induce effective Hamiltonians of the form $\boldsymbol{\mathcal{F}}\cdot(\boldsymbol{p}\times\boldsymbolσ)$. This reveals that magnetic as well as electric background fields may give rise to Rashba-type interactions, in contrast with standard condensed-matter scenarios. Before passing to the nonrelativistic limit, we analyze the relativistic content of the model in detail: the canonical structure of the deformed Dirac operator, the admissible background classes, the effective bilinear current, and the branch splitting of the relativistic dispersion relation, which constitutes the primary relativistic signature of the theory. We derive exact analytical energy levels and normalized eigenspinors for the resulting ring Hamiltonian, compute Aharonov--Anandan geometric phases, and analyze persistent spin currents together with the associated differential spin response $\mathcal{G}_s = \partial\mathcal{J}_\varphi^z/\partialξ$. Exploiting the analytical control offered by the model, we derive the first systematic order-of-magnitude bounds on the two Lorentz-invariant couplings $\mathfrak{g}_1$ and $\mathfrak{g}_2$ from both spectroscopic and mesoscopic scenarios, identifying the experimental channels most sensitive to the new physics encoded in these operators. We discuss physical implications, signatures, and possible experimental analogs, and outline several promising directions involving disorder, noise, and nonequilibrium spin dynamics.

Linearized deformations of the thick-walled (low-amplitude) (1+1)-dimensional Q-ball may be decomposed into relativistic modes, which are roughly plane waves, and also long-wavelength corotating and counterrotating Floquet modes. Each mode oscillates at a pair of mirror frequencies which average to the Q-ball frequency. The corotating modes are those of a breather or oscillon plus a very loosely bound mode. The counterrotating modes are described by an irrational-level Pöschl-Teller potential, with two discrete modes which mix with their unbound mirrors, unbinding them and turning them into Feshbach-type quasinormal modes. Expanding to leading order in the Q-ball amplitude, we present all of these modes in closed form, except for the bound mode which does not exist at leading order.

Extending the Gaussian covariant hydrodynamics approach [1] using torsion as an auxiliary field we formulate a fluctuating hydrodynamics with spin which is covariant with respect to pseudo-gauge transformations as well as generally covariant with respect to foliations. This is done via the second order gravitational Ward identities, derived here in the torsionful case. This ensures that, while angular momentum observables depend covariantly on the pseudo-gauge, the dynamics is pseudo-gauge independent, thus clarifying the role of the pseudo-gauge in hydrodynamics with spin

We propose that probability in quantum theory, like energy in general relativity, acquires a fundamentally quasilocal character in curved spacetime. Interpreting Hermiticity as the symmetry associated with inner-product conservation, we show that gravitational boundaries and horizons convert global probability conservation into a flux balance law. The resulting quasilocal probability naturally induces effective non-Hermiticity for restricted observers while preserving global unitarity. We demonstrate this explicitly in Schwarzschild, Kerr and FLRW spacetimes and after this, we identify observational imprints in black hole ringdowns. Our results suggest that in quantum field theory on curved backgrounds, probability conservation is as geometrically conditioned as energy itself.

We investigate the vacuum-induced current density for a charged scalar field in a $(D+1)$-dimensional cosmic dispiration spacetime threaded by a magnetic flux. This background combines a cosmic string and a screw dislocation, yielding a nontrivial helical geometry. By constructing the normalized mode functions of the Klein--Gordon equation, we evaluate the Wightman function and obtain the vacuum expectation value of the current density. We show that, in addition to the azimuthal component describing a persistent current around the defect, a nonvanishing axial component is induced as a direct consequence of the helical structure of the spacetime. Both components are periodic functions of the magnetic flux, depending only on its fractional part, reflecting the Aharonov--Bohm nature of the effect. Closed expressions are obtained for both massive and massless fields in arbitrary dimensions. We demonstrate that the screw dislocation parameter plays a crucial role in the behavior of the induced currents, leading to the regularization of the axial component at the origin and controlling its magnitude. The asymptotic behavior of both components is analyzed in detail. Our results reduce to known expressions in the absence of the screw dislocation, providing a consistency check. In particular, we examine the physically relevant $(3+1)$-dimensional case, where numerical analysis reveals nontrivial features arising from the interplay between topology and gauge effects.

So far, only a single theory of multiple spin-2 fields is known that features genuine multi-field interactions while remaining free of Boulware-Deser-type ghost instabilities. In this paper we show that this is the most general ghost-free multi spin-2 interaction type possible. We start with the general class of multivielbein interactions containing antisymmetrised products of vielbeins, considered earlier by Hinterbichler and Rosen. We formulate a necessary condition for these theories to be ghost-free. For two vielbeins the theory parameters remain unrestricted, reproducing the ghost-free bimetric theory. But for more than two vielbeins with genuine multi-field interactions, we show that the couplings are restricted precisely to yield the known ghost-free multivielbein theory, thus establishing its uniqueness. We also show that more general interactions, constructed using the ghost-free bimetric and multivielbein potentials as building blocks, satisfy the necessary ghost-free conditions provided the associated interaction graphs have a tree structure.

The Sen action for self-dual fields has been generalised by Hull to include two metrics which allows it to be defined on generic manifolds. In this paper we consider Kaluza-Klein compactifications of this action. The existence of two metrics presents novel challenges as there are two Kaluza-Klein towers of fields. We show that to find a consistent truncation one must include zero-modes associated to each of the two towers. Although this naively leads to a doubling of the massless degrees of freedom we show that on-shell this is not the case. We also discuss a deformation of the Sen action to include an additional form-field but which does not lead to new degrees of freedom on-shell but which arises naturally upon compactification.

We study a minimal model with an imposed $\mathbb{Z}_2$ symmetry incorporating two singlet fermions and a singlet scalar which communicate with the SM particles through a scalar-Higgs portal. We probe regions in the parameter space where stability of the three new particles are guaranteed kinematically, and thus introducing a multi component dark matter (DM) scenario. The regions in the parameter space with predicted total relic density in accordance with the observed DM abundance are found, and the contribution of each species to the total relic density is determined. The elastic DM-nucleon scattering cross section of the two fermions DM are loop suppressed, while that of the scalar DM starts at tree level and thus presumably being dominant. It is found that there exist a viable region in the parameter space that the scalar DM can evade the current direct detection (DD) experimental bounds while it has a minimal contribution to the observed relic density. The DD cross section of the fermion DM being loop suppressed, resides below the lower limit from the {\it neutrino floor}, possessing a large fraction of the total relic density.

We complete our study of the relativistic spatial distributions of the energy-momentum tensor inside polarized nucleons within the quantum phase-space formalism. In the present work, we focus on the components of the energy-momentum tensor involving at least one transverse index. We also explore the multipole structure of the transverse distributions in a moving nucleon. In the infinite-momentum frame, we show that the formalism reproduces the standard light-front distributions, including those with a ``bad'' component, and explains the origin of their structure.

We use the fuzzy-sphere approach to study the Bose-Kondo impurity problem, namely a spin-$S$ impurity coupled to the $(2+1)$-dimensional $O(3)$ Wilson-Fisher CFT (Heisenberg universality class). We demonstrate that for $S=1/2,1,3/2$ the impurity flows to a distinct stable interacting conformal defect for each $S$. Using large-scale exact diagonalization and density-matrix renormalization group methods, we observe integer-spaced defect spectrum consistent with defect conformal symmetry and compute several low-lying defect primary operators as well as the RG monotonic $g$-function. Our findings show that despite sharing the same symmetry and anomaly, Bose-Kondo impurities flow to distinct stable infrared conformal fixed points, which we refer to as \emph{fortuitous universality}. We expect this fortuitous universality to persist for all $S$, extending to $S\rightarrow\infty$, with each spin-$S$ impurity flowing to its own stable infrared conformal fixed point.

In this paper, we present the results for the forward trijet production differential cross section in the gluon initiated channel at leading order in proton-nucleus collisions. The calculations are carried out within the Color Glass Condensate (CGC) effective theory, and in the dilute-dense approximation, using effective vertices for the quark and gluon propagators interacting with the small-$x$ background gluon field. We employ the covariant perturbation theory approach and disentangle the amplitudes into regular and instantaneous contributions. Our results are expressed as convolutions of multiparton color correlators of light-like Wilson lines and perturbative impact factors, organized in compact expressions in terms of the ``bare" topologies of the contributing diagrams. The gluon initiated channel receives contributions from a $q\bar{q}g$ and a $ggg$ final state. Interestingly, when considering the $ggg$ final state, we observe, for the first time, that the four-gluon vertex topology follows a structure similar to the instantaneous contributions. Furthermore, when integrating (one of) the real gluon(s) in the final state, we identify that: i) the rapidity divergence contributes to the real part of JIMWLK of the leading-order color correlator; and ii) the collinear divergence contribute to the evolution of initial-state gluon parton distribution function, and final state fragmentation functions. These results validate the dilute-dense hybrid formalism at one-loop order, and are key ingredients towards the complete next-to-leading order calculation of dijet/dihadron production in proton--nucleus collisions.

In 1918 Weyl introduced Weyl conformal geometry and its associated quadratic action which was the first gauge theory, of a spacetime symmetry, the Weyl gauge theory (of dilatations and Poincaré symmetry). The initial physical interpretation of his theory was however short-lived and led to the downfall of Weyl geometry as a physical theory. We review how this action was re-born into a physical Weyl gauge theory of gravity. This is the only gauge theory of a spacetime symmetry with a physical gauge boson, is Weyl anomaly-free, has {\it exact} geometric interpretation, with all scales of geometric origin, and generates Einstein-Hilbert action and a positive cosmological constant in its spontaneously broken phase. A more fundamental Weyl-Dirac-Born-Infeld gauge theory action exists in Weyl geometry, that does not need a UV regularisation, of which the (geometrically regularised) Weyl gauge theory is the leading order.

We construct a Super-Grassmannian for $n-$point functions in $\mathcal{N}=2$ to $4$ SCFT$_3$. The constraints imposed by super-conformal invariance and $R-$symmetry are completely manifest in this formalism through (operator-valued) delta functions. We test our formalism in $\mathcal{N}=2$ and $\mathcal{N}=4$ AdS$_4$ super Yang-Mills theories. In the $\mathcal{N}=2$ case, for instance, we reproduce the four-gluon correlator using the four-point scalar correlator as input. For $\mathcal{N}=4$, we construct the super-operator in two distinct ways. In one approach, the super-operator has a lowest component of spin zero and includes all states up to spin two. In the other approach, we build the super-operator in a CPT self-conjugate manner, which contains only operators with spin zero, spin half, and spin one mimicking flat space $\mathcal{N}=4$ SYM super-field constructions. The latter construction is particularly interesting, as it matches directly with the $\mathcal{N}=4$ SYM amplitudes in the flat space limit, thereby demonstrating the non-triviality and usefulness of our framework. It is interesting to note that the $R-$symmetry group enhances from $SO(\mathcal{N})$ to $SU(\mathcal{N})$ in the flat space limit.

We compute the excitation function of the global $Λ$ polarization in semicentral heavy-ion collisions within a Core--Corona framework, where the interaction region is described as a dense core and a dilute corona separated by a critical value of the participant density. An important ingredient in the model are the intrinsic polarization functions in each of the two regions. These are computed from a field-theoretical approach where the vortical motion of the medium is included in an effective fermion propagator, which we derive explicitly. The interactions in the core and the corona are transmitted by suitable mediators at finite temperature and baryon chemical potential; gluons for the former and $σ$-mesons for the latter. The temperatures and baryon chemical potentials are related to the collision energies along the chemical freeze-out curve. By allowing the cross section for $Λ$ production in the nuclear environment to take on values below the nucleon-nucleon threshold cross section, the calculation describes the lowest energy polarization data point. For the centralities corresponding to the experimental data, we find that the contribution from the corona is the dominant one and that a lifetime, and correspondingly a volume of this region, which becomes larger for the smaller energies due to stopping, is an essential ingredient in the calculation. Overall, the model provides a good description of the excitation function across the full experimental range and predicts a robust maximum near $\sqrt{s_{NN}}\sim$ 3 GeV that remains stable under reasonable variations of the freeze-out curve and the proton-proton $Λ$ production threshold to account for subthreshold production in a nuclear environment.

A matrix model on a D-dimensional Euclidean space is introduced as a generalization of random matrix models and as a non-perturbative definition of discretized closed string theory. The free energy of the matrix model is formally derived to all orders in string perturbation expansion and shown to be given in terms of invariant graph polynomials, whose coefficients enumerate ribbon graphs and are a refinement of the generalized Catalan numbers. The vacuum diagrams contributing to the free energy are found to be related to Gaussian molecules, known from the study of polymer structures. Coupling the matrix field to a curved background with Riemannian metric yields a non-perturbative definition of discretized string theory on this background. No on-shell condition for the metric is required to arrive at the free energy. Rather, it is shown that the free energy of the matrix model is the Einstein-Hilbert action with cosmological constant term. The gravitational and the cosmological constants are both formally determined to all orders in the string perturbation expansion. In fact, they are explicitly given by the expectation value of a particular graph invariant. Introducing a vector field, minimally coupled to a background gauge field, provides a discretized open-closed string theory, leading to the Yang-Mills action as well as intrinsic and extrinsic curvature terms.

We explore the potential of the Mu2e experiment to probe the lepton-flavor-violating process $μ^- e^- \to e^- e^-$ in a muonic atom within a simplified axion-like particle (ALP) framework featuring flavor-violating $e$-$μ$ couplings and a flavor-diagonal pseudoscalar coupling to electrons, which also allows for possible invisible ALP decays into a dark sector. We compute the ALP-mediated contribution to the transition rate and show that, at fixed couplings, the branching ratio increases for lighter mediators and scales as $(Z-1)^3$, favoring heavier nuclei. We compare the model against constraints from $μ\to eγ$, $μ\to 3e$, $μ\to eγγ$, $μ\to e+\mathrm{inv}$, and muonium-antimuonium conversion, as well as from the anomalous magnetic moments of the electron and muon. Additional astrophysical and beam-dump limits on the electron coupling are also discussed. A key result is that $Δa_e$ provides one of the most stringent probes of the parameter space and, in the global scan, excludes the largest fraction of sampled points. After applying the laboratory constraints used in the scan, the viable branching ratio for $μ^- e^- \to e^- e^-$ in aluminum drops to at most $\mathcal{O}(10^{-20})$, while the resonant region $2m_e<m_a<m_μ-m_e$ is much more heavily suppressed. The highest achievable values are closely tied to $\mathcal{B}(μ\to 3e)$ near its current limit, indicating that the upcoming Mu3e experiment will explore the most promising region relevant for this muonic-atom signal. Our analysis shows that, although a light ALP can parametrically enhance $μ^- e^- \to e^- e^-$ at fixed couplings, existing bounds -- especially $Δa_e$, $μ\to 3e$, $μ\to eγ$, and muonium oscillations -- severely limit the observable rate.

It has been shown that multiway junctions gluing $n$ copies of locally AdS$_3$ spacetimes ($n\geq 2$) can be described by $n-1$ strings obeying non-linear Nambu-Goto equations coupled by Monge-Ampère like terms. Here we study how such junctions along with their stringy degrees of freedom can be interpreted in terms of an interface between $n$ identical holographic conformal theories each defined on a semi-infinite line (wire). We study the gravitational scattering problem at the multiway junction, and show that at the linearized order the dual interfaces correspond to quantum maps which factorize into a product of a scattering matrix determined only by the tension of the dual junction and relative automorphisms of the Virasoro algebra governed by the $n-1$ stringy modes. Both of these are universal in the sense that they are independent of linear modifications of the background state. These generalize earlier results for the 2-way junctions implying that the dual interface is a tunable energy transmitter. We comment on understanding the quantum map corresponding to the full non-linear gravitational problem, and study Ward identities and unitarity bounds.

A new class of symmetries of two Higgs doublet models was recently discovered, the result of an unorthodox transformation on scalar and gauge fields and spacetime coordinates. It was explicitly shown that it is possible to choose Yukawa matrix textures which respect those symmetries up to two-loops. In this work we will establish the fermion field transformations for the two Higgs doublet models to be considered in the context of the new symmetries established. New regions of parameter space, invariant under renormalization to all orders of perturbation theory, are discovered, including scalar-fermion interactions.

Q-balls are large bound-state systems of scalar particles, described classically through localized solutions of the equations of motion. Promoting the required stabilizing $U(1)$ symmetry to a gauge symmetry leads to gauged Q-balls, which cannot grow beyond some maximal size and charge on account of the repulsive gauge interactions. These gauged Q-balls have been studied extensively for scalar potentials that satisfy Coleman's thin-wall criterion; here, we explore gauged Q-balls in flat potentials, which often occur in supersymmetric models. Even though global Q-balls in flat potentials are qualitatively different from Coleman's Q-balls, we find that the gauged versions are remarkably similar. We provide analytic approximations for these solitons and compare to numerical solutions. In addition, we study Proca Q-balls, i.e. make the gauge bosons massive, which interpolates between the global and gauged cases.

We study a non-relativistic realisation of two-dimensional de Sitter gravity both from its boundary and bulk description with the goal of learning about de Sitter space and paving the way for extending the holographic duality into a non-relativistic direction. On the boundary side, we analyse the Schwarzian-type boundary action associated with non-relativistic de Sitter gravity and evaluate its one-loop partition function in order to compute its quantum fluctuations. Rather than relying on the coadjoint-orbit construction, we derive the path integral measure directly from the action using the Ostrogradsky formalism. We find a temperature-dependent prefactor scaling as $T^2$, of which the power agrees with the counting of the four global symmetry generators present. On the bulk side, we construct the corresponding torsionless Newton-Cartan geometry and show that it satisfies the equations of motion of a non-relativistic JT-like action and uplift the geometry to a three-dimensional Lorentzian geometry.

We construct a Super-Grassmannian integral representation for $n-$point functions in $\mathcal{N}=1$ SCFT$_3$. In this formalism, conformal invariance, supersymmetry, and special superconformal invariance are implemented manifestly through (operator-valued) delta function constraints. An important feature of this framework is the fact that we obtain simple algebraic relations among component correlators, which enable us to determine any component correlator in terms of just one of the component correlators. In particular, this formalism enables us to construct (A)dS$_4$ boundary correlators with contact diagrams from those that receive contributions purely from particle exchanges. We illustrate this by determining the (A)dS$_4$ Yang-Mills gluon four-point function from its gluino counterpart. Further, we establish the flat-space limit in super-space, finding a perfect agreement with existing flat-space results.

We analyze the fixed-angle high-energy ($α' \to \infty$) structure of $n$-point tree-level string amplitudes from complementary perspectives: locally via saddle-point expansions, algebraically via difference equations and their asymptotic structure, analytically via Aomoto-Gauss-Manin connection and Mellin-Barnes representation, and geometrically via twisted intersection theory and Lefschetz thimbles. Using, in turn, saddle-point analysis and finite-difference equations in the kinematic variables, we show that the perturbative coefficients in the resulting asymptotic series in $1/α'$ are organized by Bernoulli-number data, rather than by the multiple zeta values characteristic of the low-energy $α' \to 0$ regime. Resurgence theory allows upgrading these divergent series to transseries whose Stokes data capture the analytic continuation between unphysical and physical kinematic regions in the form of non-perturbative monodromy contributions. We derive the transseries for four-point open string amplitudes explicitly. We also construct a differential and Mellin formulation which place their low- and high-energy expansions in a common analytic framework and unifies them as asymptotic sectors of the same underlying object. We extend the difference-equation analysis to $n \geq 5$, where it yields perturbative high-energy asymptotic expansions and leads naturally to a higher-rank connection problem. Finally, translating our asymptotic analysis into the language of twisted de Rham theory, we derive an alternative double-copy representation of the high-energy limit of closed-string amplitudes in terms of Lefschetz thimbles for any $n$.

We study the holographic spread/Krylov complexity of operators with non-trivial internal structure and of genuinely extended operators. We first consider a massive particle in AdS$_5\times S^5$ carrying conserved $R$-charge, and show how motion in the internal space modifies the complexity growth, yielding a natural holographic realisation of symmetry-resolved Krylov complexity. We then move to probes that are effectively pointlike from the field-theory viewpoint but possess an intrinsic structure in the bulk: baryon-vertex configurations and giant gravitons. Our results indicate that, for this broad class of structured but pointlike probes, the leading large-time behaviour retains the characteristic form expected for local operators in conformal theories, while the internal structure and induced charges produce informative subleading effects. We also study a genuinely extended probe, a fundamental string falling in AdS while stretched along a spatial direction, as a model for the spread complexity of a non-local operator. In this case, although the leading behaviour still exhibits the expected growth pattern, the subleading terms and intermediate regimes differ qualitatively from those of pointlike probes. This provides concrete evidence that extended operators carry a finer notion of spread complexity, sensitive to their spatial structure. Our results broaden the class of probes for which holographic Krylov complexity can be analysed explicitly, clarify which features are universal and which depend on the nature of the operator, and open a promising route toward a sharper field-theory understanding of complexity for charged, composite and extended excitations.

Motivated by the recently proposed geometric descriptions of 0A and 0B in M-theory and F-theory, we propose a web of duality among non-supersymmetric strings. In particular we argue that the distinct $\mathbb{Z}_2$ quotients of M-theory on $S^1\vee S^1$ lead to both 0A orientifolds as well as non-supersymmetric 10d heterotic vacua of the E-type, including the tachyon-free $SO(16)\times SO(16)$ strings. Moreover we identify certain $\mathbb{Z}_2$ quotients of F-theory on $(S^1\vee S^1)\times S^1$ with 0B orientifolds (including a tachyon-free model) as well as others with dual to non-supersymmetric heterotic strings of the D-type. Moreover using this picture we resolve some puzzles and provide further evidence for the Bergman-Gaberdiel duality between a particular 0B orientifold in 10 dimensions and the Narain compactification of 26-dimensional bosonic strings on a 16-dimensional torus, as well as the DMS conjecture of a 0A orientifold duality in 10d with a bosonic string orientifold of a Narain compactification to 10d.

Quantum simulations of high-energy physics in $2+1$D can probe dynamical phenomena nonexistent in one spatial dimension and access regimes that are challenging for existing classical simulation methods. For string dynamics -- relevant to hadronization -- a plaquette term is required to realize genuine $2+1$D behavior, as it endows the gauge field with dynamics and enables the propagation of photon-like excitations. Here, we realize a U$(1)$ quantum link model of quantum electrodynamics in two spatial dimensions with a tunable plaquette term on a \texttt{Quantinuum System Model H2} quantum computer. We implement, to our knowledge, the largest quantum simulation of string-breaking dynamics reported to date, on a $5 \times 4$ matter-site square lattice using $51$ qubits. The simulation uses a shallow circuit design with a two-qubit gate depth of $28$ per Trotter step and up to $1540$ entangling gates. Starting from far-from-equilibrium string configurations, we measure the probability for the string to propagate within the lattice plane and find signatures of genuine $2+1$D dynamics only when the plaquette term is present. In a resonant regime, we observe the annihilation of string segments accompanied by the production of electron--positron pairs that screen them. We further find that, only with a nonzero plaquette term, matter creation extends across the lattice plane rather than remaining confined to the initial string path. These results experimentally realize string breaking and demonstrate the emergence of dynamical gauge fields in two spatial dimensions, establishing a route to photon-like propagation in programmable quantum simulators of gauge theories.

A major goal of the quantum simulation of high-energy physics (HEP) is to probe real-time nonperturbative far-from-equilibrium quantum processes underlying phenomena such as hadronization in quantum chromodynamics (QCD). The quantum simulation of the dynamics of confining strings and glueballs, both essential aspects of quark confinement, in a controllable first-principles way is an important step towards this goal. Here, we realize a $\mathbb{Z}_2$ lattice gauge theory in $2+1$D with a tunable plaquette term on a \texttt{Quantinuum System Model H2} trapped-ion quantum computer. We implement a shallow depth-6 Trotter circuit on a $6 \times 5$ matter-site square lattice utilizing all $56$ available qubits to execute over $1000$ entangling gates. We prepare far-from-equilibrium initial string configurations that we quench across a range of parameters to observe rich dynamical phenomena, such as the formation of gauge-invariant closed-loop excitations reminiscent of glueballs in QCD and multi-order string breaking accompanied by spontaneous matter creation. We further demonstrate experimentally that the system displays genuine $2+1$D dynamics, as evidenced by string snapshots over time that cannot be trivially mapped to $1+1$D physics. Our results demonstrate digital quantum simulations of nonequilibrium dynamics in a higher-dimensional lattice gauge theory and provide an experimentally accessible setting for phenomena related to confinement physics.

In 2025, Bernardes, Erler and Firat proposed a novel, elegant expression for the symplectic form on phase space applicable to non-local theories. We show that this BEF symplectic structure can be derived directly from an $L_\infty$-Lagrangian by following the covariant phase space approach. Moreover, we establish a precise relation between the BEF symplectic structure and the Barnich--Brandt symplectic form for general finite-derivative theories. In particular, we prove that for theories with second-order equations of motion, the BEF symplectic structure coincides with the Barnich--Brandt construction, thereby explaining the emergence of the canonical corner term in general relativity within the BEF approach. We further argue that the BEF symplectic structure naturally encodes information about generic corner terms and some information about boundary conditions. In addition, we develop a general expression for the Hamiltonian in theories in $L_\infty$-form and present several explicit examples illustrating the construction.

We propose a novel laboratory test of gravity combining seismic-wave measurements with cosmic-ray muon detections. Quantum-gravity corrections to the anharmonic Debye model are derived, yielding a modified bulk modulus that encodes deviations from standard gravity. The usual dependence on density, a dominant source of uncertainty, is removed via muon tomography and seismic velocities measurement. We show that this setup can constrain gravity parameters at a level comparable to current laboratory experiments. Prospects for further improvements are briefly discussed.

The LHC is achieving energies never reached before, opening up possibilities for the discovery of exotic particles in the TeV mass range. Such states include magnetic monopoles, which can explain the electric charge quantisation and restore the symmetry in Maxwell's equations with respect to the magnetic and electric fields. Scenarios proposed to shed light to dark matter and neutrino masses introduce high-electric-charge objects (HECOs). The existence of both classes of particles can be probed in precision measurements in a manner complementary to direct searches. We focus on the contributions of such virtual particles to light-by-light scattering in the context of effective field theories and a Born-Infeld scenario. Specifically, measurements of central exclusive production of photon pairs with proton tagging carried out by the CMS-TOTEM Precision Proton Spectrometer with LHC Run 2 proton-proton collision data are used to constrain magnetic monopole and HECOs. Resummation techniques have been employed to deal with the large HECO-photon coupling. Masses of up to a few tens of TeV have been excluded for monopoles and HECOs of various spins and magnetic and electric charges, respectively.

In this work, we study exotic theta terms in the 2+1d $φ$-theory, which provides a continuum description of the XY-plaquette model. The $φ$-theory can be viewed as a fractonic analogue of the 1+1d compact boson and exhibits momentum and winding subsystem symmetries. In this theory, discontinuous field configurations play a crucial role. Although such configurations spoil the naive topology of the field, they induce nontrivial backreactions that give rise to new topological terms. We study two types of theta terms, which we call the bulk theta term and the foliated theta term. The foliated theta term is constructed by coupling winding currents on neighboring leaves of a foliation. Remarkably, the corresponding theta angle can vary spatially without affecting the classical equations of motion. Both theta terms lead to generalized Witten effects, in which vortex operators carrying winding subsystem charge acquire momentum subsystem charge. In the case of the foliated theta angle, the Witten effect exhibits a more intricate structure: vortex operators acquire a quadrupolar momentum charge. We demonstrate these features using lattice realizations based on the modified Villain formulation.

We consider the problem of identifying the onset of the conformal window for QCD with $N_f$ massless flavors in the fundamental representation, and propose a new effective method to determine it from lattice simulations. This method is based on the investigation of the so-called Roberge-Weiss transition temperature $T_{RW}$, which is encountered at specific values of the imaginary baryon chemical potential, and can also be interpreted as the inverse of the critical spatial size at which charge conjugation is spontaneously broken in a finite box. Since $T_{RW}$ corresponds to a genuine phase transition for any value of the quark masses, it is a well-defined quantity; we argue that the critical $N_f$ at which $T_{RW}$ vanishes in the chiral limit coincides with the onset of the conformal window. We implement our proposal by investigating QCD with $N_f = 8$ flavors, discretized via stout improved staggered fermions and the tree-level improved Symanzik pure gauge action, at Euclidean temporal extents $N_t = 8, 10, 12, 16, 24$. In this case, we find evidence that $T_{RW}$ already vanishes in the chiral limit, indicating that $N_f = 8$ is already in the conformal window.

In anticipation of the new wave of ALICE experiments on particle-resonance correlation functions, we study the interaction of a kaon with the $D_{s0}^*(2317)$ resonance. Assuming the $D_{s0}^*(2317)$ to be a $DK$ molecular state in isospin $I=0$, we employ the fixed center approximation (FCA) to describe the kaon scattering off the $DK$ cluster, and implement elastic unitarity in the $K D_{s0}^*(2317)$ amplitude via an optical potential and the Lippmann-Schwinger equation. We evaluate the scattering length, effective range, and correlation function, which exhibits a shape characteristic of a strongly attractive interaction. Notably, the amplitude develops a narrow resonant peak about 40~MeV below the $K D_{s0}^*(2317)$ threshold, signaling a three-body bound state. We discuss the experimental feasibility of observing this state through the invariant mass distribution of $K D_s^+ π^0$, and argue that such three-body states, predicted by various theoretical approaches, offer promising targets for future experimental searches, providing valuable insights into the nature of exotic hadronic resonances.

We compute the spectrum of light mesons, composed by up, down, and strange quarks, using a symmetry-preserving approximation that permits the inclusion of fully-dressed quark-gluon vertices in the key dynamical equations. This method is characterized by the use of the standard symmetric kinematic configuration as a seed in the corresponding Schwinger-Dyson equation, yielding finally the full kinematic dependence of all eight form factors composing the transversely-projected quark-gluon vertex. The extension of this approach to the case of distinct nonvanishing current quark masses is discussed, and the compatibility with the fundamental Ward-Takahashi identities demonstrated. The corresponding Bethe-Salpeter kernel is composed by three different diagrammatic structures, which may be deduced from the attendant quark gap equation by applying the standard "cutting" rules. The masses of the light mesons are computed by first determining the eigenvalue of the Bethe-Salpeter equation as a function of Euclidean momenta, and then using the Schlessinger extrapolation method to determine the Minkowski momentum for which this eigenvalue becomes unity. The resulting meson masses are in good agreement with experimental values, and substantially improve upon predictions from the rainbow-ladder approximation.

In this note, we address the question of whether locality, unitarity, and newly discovered hidden zeros can completely determine tree-level amplitudes, from the perspective of soft limit. We reconstruct the single-soft theorems of tree YM amplitudes and the double-soft theorems of tree NLSM amplitudes from locality, unitarity, and hidden zeros. A series of studies have shown that the full YM and NLSM amplitudes can be constructed from these soft theorems; therefore, we conclude that locality, unitarity, and hidden zeros completely determine the tree-level YM and NLSM amplitudes.

This article presents recent measurements by the ALICE Collaboration in proton--oxygen (pO), oxygen--oxygen (OO), and neon--neon (Ne--Ne) collisions delivered by the LHC in July 2025. Measurements of the primary charged-particle pseudorapidity density and the elliptic and triangular flow coefficients of charged particles are reported. Experimental evidence of the suppression of neutral pion yields in OO collisions relative to the proton--proton baseline is also discussed. Comparisons of these new data with theoretical models provide key input to understand particle production, collective phenomena, and parton energy loss in small collision systems.

We investigate the phenomenological impact of incorporating vector-like bottom (VLB) quarks into the Type-II Two-Higgs-Doublet Model (2HDM-II). This framework introduces novel beyond-Standard-Model (BSM) decay channels $B \to Hb$, $B \to Ab$, and $B \to H^-t$, which are typically ignored by LHC pair-production searches focused on Standard Model (SM) final states ($B \to Zb$, $B \to hb$, $B \to Wt$). Our analysis reveals that these BSM pathways significantly weaken current VLB mass constraints. In the 2HDM-II alignment limit, the mass limit for a singlet $B$ shifts from approximately 1.5 TeV down to 1.34 TeV. For $(T, B)$ and $(B, Y)$ doublet configurations, the mass limits relax further to approximately 0.98 TeV, driven by the dominance of $B \to Hb$ and $B \to Ab$ decays, which can reach combined branching ratios of nearly 100\%.

Inverse beta decay (IBD), $\overlineν_e p \to e^+ n \left( γ\right)$, is the main detection channel for reactor and supernova antineutrinos. To provide precise IBD cross sections at antineutrino energies $E_{\overlineν_e} \gtrsim 10~\mathrm{MeV}$, we evaluate radiative corrections from virtual pions within the framework of heavy baryon chiral perturbation theory. At leading order, only the pion isospin-splitting contributions are not suppressed by the electron mass. At next-to-leading order, besides recoil effects, only the Wilson coefficient $c_4$ contributes to the kinematic dependence. However, its precise value is not relevant for IBD at relatively low energies since all next-to-leading order radiative corrections are relatively small. We find the kinematic dependence of the pion-induced QED radiative corrections at the level and below the uncertainty from the momentum dependence of the nucleon form factors. Our results enable sub-permille theoretical precision of charged-current elastic (anti)neutrino-nucleon scattering at antineutrino energies $E_{\overlineν_e} \gtrsim 10~\mathrm{MeV}$.

We postulate that the excesses of di-dijet events observed at the LHC are attributed to the production of four fourth-generation quarks $b'$ with a mass $m_{b'}\approx 2$ TeV at few-TeV scales. The di-dijet signals around the four-jet invariant mass $m_{4j}\approx 8$ TeV arise from a resonant $b'b'\bar b'\bar b'$ tetraquark production, where the dijet resonances of masses about 2 TeV correspond to $b'\bar b'$ first excited states (color-octet scalars with the principal quantum number $n=2$) in a Yukawa potential created by Higgs boson exchanges. Those around $m_{4j}\approx 3.6$ TeV originate from a non-resonant $b'b'\bar b'\bar b'$ production, where the dijet resonances of masses 0.95 TeV correspond to $b'\bar b'$ ground states (color-octet vectors with $n=1$). It is shown that a $b'\bar b'$ system with $m_{b'}\approx 2$ TeV in the Yukawa potential does generate the aforementioned bound state spectrum. We then illustrate that the observed excesses can be accommodated in our setup by translating the fourth-generation model to the effective theories containing color-octet scalars and vectors available in the literature. The di-dijet events at $m_{4j}= 6.6$ TeV and 5.8 TeV with dijet masses about 2 TeV can also be interpreted in the same framework. Simply speaking, our scenario can be viewed as a TeV-scale version of the search for a fully charmed tetraquark via the four-muon channels $X(6900)\to (c\bar c)(c\bar c)\to 4μ$ at a GeV scale.

We report MINERvA's first measurement of inclusive charged-current $\barν_μ$ cross sections on carbon, hydrocarbon, iron, and lead, and their ratios to the cross section on hydrocarbon, as functions of the antimuon transverse momentum, $p_{\mathrm{T}}$. Using a wide-band $\barν_μ$ beam with mean energy $\sim 6~\text{GeV}$, these measurements probe all interaction modes, including the transition from resonance production to deep-inelastic scattering. The total uncertainties are typically $5-10\%$ for the absolute cross sections and $2-5\%$ for the ratios. Comparisons with multiple neutrino interaction models reveal significant discrepancies in the $p_{\mathrm{T}}$ dependence, particularly for heavier nuclei. The disagreements are most pronounced at low $p_{\mathrm{T}}$ but extend across the full $p_{\mathrm{T}}$ range, indicating missing or mis-modelled nuclear effects.

In this paper, we study the internal structure of light pseudoscalar mesons using spin improved power-law wave functions. We choose the pion and the kaon for our work. We use the standard quark-quark correlation functions to calculate the distribution amplitudes (DAs), parton distribution functions (PDFs), transverse momentum dependent parton distribution functions (TMDs), and generalized parton distribution functions (GPDs) at zero skewness and form factors. We present all the above distribution functions through the overlap of light-front wave functions (LFWFs). We use leading-order Efremov-Radyushkin-Brodsky-Lepage (ERBL) equations for DAs and next-to-leading-order (NLO) Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) equations for PDFs to evolve them to higher scales. We find that only 41\% of the longitudinal momentum fraction is carried by the quark and antiquark of both pion and kaon at 16~GeV$^2$. The vector form factors for both the pion and the kaon are found to be in good agreement with experimental data. Similarly, the electromagnetic charge radii are found to be 0.668~fm and 0.704~fm for the pion and kaon, respectively.

The relic abundance of vector dark matter originating from an inherited axion-$SU(2)$ condensate is typically determined by implementing an adiabatic matching procedure across the symmetry-breaking transition. We demonstrate that this outcome does not arise in the generic case. The post-inflationary crossover can instead be formulated as a dynamical quantum quench problem, in which the residual coherent component of the field is characterized by a survival factor that induces an $\mathcal{O}(1)$ renormalization of the standard abundance relation. Expressed in conformal time, the spatially homogeneous condensate dynamics reduce to those of a canonical oscillator with quartic and quadratic self-interactions. This representation enables an analytic determination of the matching conditions across the symmetry-breaking transition, the derivation of the corresponding quench work and excess energy relations, and a quantitative validation of the coherent sector description via numerical simulations in both Minkowski and Friedmann--Robertson--Walker backgrounds. We also formulate the homogeneous fluctuation theory via the diagonal-$SO(3)$ $1 \oplus 3 \oplus 5$ decomposition and isolate a soft traceless-symmetric quintet with a $k=0$ vacuum obstruction, a regulated ultraviolet adiabatic bound, and a positive quartic stabilization term. Collectively, these results refine the theoretical description of inherited non-Abelian dark matter production and establish the necessary infrared framework for subsequent investigations of finite-$k$ gauge--Higgs transfer dynamics.

Accelerator-based neutrino physics is entering an energy-frontier regime in which interactions reach the TeV scale and produce exceptionally dense, overlapping detector signatures. In this regime, event interpretation becomes impractical for conventional reconstruction approaches, particularly when labelled data are scarce and the analysis spans diverse downstream objectives. We present a sparse ViT framework for learning reusable representations from heterogeneous detector data. Self-supervised pre-training combines masked autoencoder reconstruction with relational voxel-level objectives for hierarchy, ghost and particle identification, and the resulting shared encoder is then jointly fine-tuned across classification and regression tasks. Evaluated on simulated events from the proposed FASERCal concept at the LHC, we find that pre-training consistently improves neutrino flavour and charm-quark identification, momentum regression, and vertex reconstruction over training from scratch, with the addition of relational objectives yielding further gains in the most topologically complex channels. Interpretability analyses further show that pre-training yields a more structured latent space, while detector-subsystem ablations recover physically plausible channel-dependent roles for the heterogeneous inputs. A data-efficiency study shows that, with roughly $10^3$ labelled events, the pre-trained encoder already matches the flavour-classification performance of a randomly initialised model trained on an order of magnitude more data. The learned representations also transfer effectively to publicly available benchmarks spanning different detector technologies and energy scales, matching or exceeding published baselines. These results support self-supervised pre-training on multimodal detector data as a scalable route towards reusable representations for neutrino and particle-detector analysis.

We construct a causal and covariantly stable kinetic model whose spectrum at real wavenumbers $k$ reproduces any rest-frame stable dissipative dispersion relation $ω(k)$ via suitable initialization of the microscopic degrees of freedom. Macroscopic observables can therefore obey arbitrary linear evolution equations (including forms that would be acausal if taken as fundamental), while the underlying dynamics remains causal, and all apparent propagation is encoded in the initial data. This provides an explicit counterexample to the idea that microscopic causality alone constrains the analytic form of dispersion relations at real $k$. In particular, bounds on transport coefficients based solely on the analytic structure of $ω(k)$, such as the hydrohedron bounds, require additional assumptions about the region in the complex $k$-plane where $ω(k)$ corresponds to physical modes.

We investigate finite volume effects on the transverse momentum spectra of charged pions produced in the most central heavy-ion collisions at RHIC and LHC energies. A cylindrically symmetric finite volume Boltzmann-Gibbs blast-wave model is employed that fully incorporates the finite longitudinal extent of the fire cylinder at kinetic freeze-out. The model applies Planck transformations to convert the local rest frame temperature and chemical potential of each fluid element into laboratory frame values, ensuring full Lorentz covariance. This approach is compared with the conventional infinite volume blast-wave model, in which the thermodynamic parameters remain defined in the local rest frame while the particle momenta are expressed in the laboratory frame. Both models are fitted to the experimental transverse momentum distributions of charged pions measured by the HADES, STAR, PHENIX, and ALICE collaborations over the center-of-mass energy range $\sqrt{s_{NN}} = 2.4$ GeV to $5.44$ TeV. The finite volume model with Planck transformed laboratory frame parameters yields temperature values fully consistent with relativistic thermodynamics (except for a small anomaly at $\sqrt{s_{NN}} = 193$ and $200$ GeV) and produces realistic fire cylinder volumes several times larger than the initial nuclear overlap volume. In contrast, the conventional infinite volume model yields unphysical results: infinite volume, infinite maximum half-length, and maximum longitudinal flow velocity equal to the speed of light at all energies. These findings demonstrate that a proper treatment of finite system size, together with the correct Lorentz (Planck form) transformation of the thermodynamic variables, is essential for the reliable extraction of freeze-out parameters in heavy-ion collisions.

Similar to some other two-Higgs doublet models (2HDM), the flipped 2HDM admits of a light pseudoscalar physical state whose mass can be well below 50 GeV. The fact that the pseudoscalar decays dominantly into a $b{\bar b}$ pair makes its identification at the Large Hadron Collider (LHC) difficult. Moreover, the regions of the parameter space corresponding to a light pseudoscalar tend to jeopardize perturbativity at a rather low scale. One possibility that ameliorates this problem is to postulate that the light physical state has the admixture of an SU(2) singlet field. In such a situation, however, the production mode of the pseudoscalar along with a $Z$ (which provides a useful tag) gets suppressed. We have here chosen to fall back on the QCD-driven final state, namely, one or two jets, together with an energetic squeezed $b{\bar b}$-pair. We utilize boosted di-b-jet tagging techniques and a strategy based on boosted decision trees (BDT) to analyze the signals, considering all backgrounds and likely fakes (mostly from charmed quarks). We find that, including 10\% systematics, one can expect signal significance of 5-10$σ$ with an integrated luminosity of 3 $ab^{-1}$.

We calculate direct detection constraints on inelastic dark matter (DM) for a scalar portal scenario with leptophilic couplings. The p-wave velocity suppression of the annihilation cross section of scalar-mediated inelastic Dirac DM implies the opening of viable regions of DM parameter space in the MeV-GeV mass range. Xenon-based experiments can provide a constraints on scalar-mediated inelastic fermion dark matter for sub-MeV mass splitting, via endothermic and exothermic spin-independent DM-electron scattering. To estimate the relevant constraints, we use public data from the XENON1T, PandaX-4T, and LZ liquid-xenon experiments that measure ionization electron signals.

Most quantum divergences derive their structure from classical f-divergences or Renyi-type constructions, a dependence that obscures several quantum geometric effects. We introduce a quantum relative-alpha-entropy that extends Umegaki's relative entropy while falling outside the f-divergence class. The proposed divergence exhibits a nonlinear convexity property, which yields a generalized convexity result for the Petz-Renyi divergence for alpha greater than one, complementing the known convexity for alpha less than one. It is additive under tensor products, invariant under unitary transformations, and depends only on the relative geometry of quantum states rather than their absolute magnitudes. Using Nussbaum-Szkola-type distributions, we also establish an exact correspondence of this divergence with classical relative-alpha-entropy. This reveals relative-alpha-entropy as a fundamentally geometric notion of quantum distinguishability not captured by existing divergence frameworks.

Lattice QCD (LQCD) calculations play a key role in the establishment of flavor anomalies. One of the most recent advancements in LQCD to this end has been the publication of several calculations of the $B\to D^\ast\ellν$ form factors, but despite all the anticipation, the LQCD results have been unable to give a final answer to the questions it was destined to answer. In this work I briefly review what is the current status of heavy-to-heavy and heavy-to-light semileptonic decays calculations in LQCD, and what we can expect for the near and not-so-near future.

The quantum nature of gravity remains experimentally unverified, despite recent proposals to probe it using tabletop experiments such as gravity-mediated entanglement schemes. In parallel, consistent formulations of classical--quantum dynamics have been developed as alternative descriptions of gravity, in which quantum matter interacts with a classical mediator assumed to be fundamentally classical. In this work, we show that classical--quantum dynamics arise generically as an effective description of fully quantum systems under decoherence, providing a bridge between fully quantum and classical--quantum dynamics. We derive the reduced dynamics, which are generically non-Markovian, using an explicit hidden model in which the mediator is coupled to unobserved environmental degrees of freedom. We identify a concrete criterion for when a classical--quantum interpretation is valid: the semi-Wigner operator associated with the mediator sector must remain positive semidefinite, which can be expressed as a positivity condition on nonlocal kernels governing the evolution. In the short-memory limit, the reduced evolution reproduces Markovian classical--quantum dynamics of Oppenheim and collaborators. Our results imply that a classical mediator can arise effectively from decohered quantum dynamics, so that experimental agreement with classical-quantum models does not uniquely determine whether the mediator is fundamentally classical.

We investigate the impact of quantum gravitational memory-burden effects on high-energy neutrino signals from evaporating primordial black holes and the resulting constraints from IceCube observations. Treating the backreaction as an energy-dependent deformation of the Hawking emission spectrum, we show that the high-energy tail is suppressed while the infrared behaviour remains unchanged. We derive analytically that this modification reduces the total luminosity and extends the evaporation lifetime by a mass-independent factor determined solely by the suppression parameter. Using an effective treatment of cosmological redshift, we compute the diffuse neutrino flux from a primordial black hole population and compare it with the observed astrophysical neutrino spectrum to constrain the primordial black hole dark matter fraction. We find that the suppression onset lies within the IceCube sensitivity window, leading to a direct reduction of the observable signal and a systematic weakening of the inferred bounds. Our results provide a controlled phenomenological framework for assessing the impact of quantum gravitational corrections on neutrino probes of primordial black hole evaporation.

We develop a unified mathematical method for the pole structure of frequency-domain Green's functions and the associated quasinormal spectra in radial boundary value problems reducible to the Gauss hypergeometric equation. By systematically employing connection formulas for Kummer solutions, we construct an explicit quantization function that encodes arbitrary linear asymptotic boundary conditions. We demonstrate that the frequency-dependent spectral factor entering the residue formula is controlled algebraically by the closed-form Digamma derivative of this quantization function, bypassing integral evaluation. Furthermore, we establish the simultaneous vanishing of the quantization function and its first derivative as a direct algebraic criterion for double-pole QNMs. The formalism is successfully benchmarked against the exact BTZ black hole spectrum and provides an analytic diagnostic for the exceptional lines and nearly double-pole excitations in the Nariai/Pöschl-Teller limit.

We present a systematic analysis of the Higgs signal strengths at 125 GeV and 95 GeV in a non-supersymmetric $U(1)_X$ model with vector-like fermions ($U(1)_X$VLFM). This model extends the SM by introducing an additional $U(1)_X$ gauge symmetry, three right-handed neutrinos, two singlet Higgs fields ($φ$ and $S$), and one generation of vector-like quarks and leptons. The scalar fields mix with each other in the neutral CP-even sector, leading to two Higgs-like states around 95 GeV and 125 GeV. A $χ^2$ analysis is performed by combining the Higgs signal strength measurements at 125 GeV from ATLAS and CMS, including the $γγ$, $WW^*$, $ZZ^*$, $b\bar{b}$, and $τ\barτ$ channels, with the 95 GeV excesses observed in the diphoton and $b\bar{b}$ final states reported by CMS and LEP. Our results indicate that the $U(1)_X$VLFM can successfully reproduce the observed signal strengths of the 125 GeV Higgs while simultaneously explaining the 95 GeV excess. The parameters $g_X$, $g_{YX}$, $v_S$, $v_P$, and the new Yukawa couplings play a crucial role in achieving this consistency.

We report x-ray diffraction and emission spectroscopy of FeO under laser-driven shock compression between 31-199 GPa. FeO retains the B1 (rocksalt) structure along the Hugoniot to the melt boundary at 191 GPa. While the phase and volume are broadly consistent with results from static compression, we observe an anomalous 7-10% volume collapse around 60 GPa absent in static experiments. We identify this as an isostructural high-spin to low-spin metallic transition in FeO. The low-spin state is directly evidenced by x-ray emission spectroscopy at 180 GPa.

We present a first real-time study of hadronic scattering in a $(1+1)$-dimensional SU(2) lattice gauge theory with fundamental fermions using tensor-network techniques. Working in the gaugeless Hamiltonian formulation, we investigate scattering processes across sectors of fixed global baryon number $B = 0, 1, 2$, corresponding respectively to meson--meson, meson--baryon, and baryon--baryon collisions. At strong coupling, the $B = 0$ and $B = 2$ channels exhibit predominantly elastic dynamics closely resembling the U(1) Schwinger model. The mixed $B = 1$ sector displays qualitatively new behavior: meson and baryon wavepackets become entangled during the collision, with the slower state becoming spatially delocalized while the faster one propagates ballistically. We characterize these processes through local observables, entanglement entropy, and the information lattice.

We study massive modes on a magnetized blow-up manifold of $T^2/\mathbb{Z}_N$. The blow-up manifold can be constructed by appropriately replacing orbifold singular points with a part of $S^2$. To ensure a smooth connection between the massive modes on magnetized $T^2/\mathbb{Z}_N$ orbifold and those on magnetized $S^2$, it is required that not only the total magnetic flux as well as the total curvature but also the effective magnetic flux on the connected line remain invariant under the blow-up procedure. Furthermore, we find that the number of the localized modes at each orbifold singular point increases by one for each unit increment of the mass level.

A study has been carried out to evaluate the performance stability of Gas Electron Multiplier (GEM) chamber prototypes in the laboratory using $^{55}$Fe radiation source with Argon and CO$_2$ gas mixture. This research focuses on the characterisation of the GEM detector's gain, efficiency (count rate with radioactive source), and energy resolution under varying operational conditions. A patch on the detector has been subjected to continuous and absolutely uninterrupted radiation for about 98 days. The gain and energy resolution of the detector are measured along with the ambient parameters temperature (t), pressure (p) and relative humidity (RH). In addition to that, the long-term behaviour of the count rate with a strong radioactive source are also studied. This work is very relevant for Micro Pattern Gaseous Detectors (MPGD) such as GEM before installing on large experiment. The experimental setup, methodology, and results are presented in this article.

We have evaluated fluctuations in neutrino fluxes caused by solar gravity (g) modes based on the analysis of linear adiabatic oscillation of a spherically symmetric star. We find that the first-order fluctuation is zero due to geometrical cancellation. We still find that the second-order fluctuation is non-zero, which consists of time-varying and non-time-varying components. The amplitude of the time-varying component is small (${\sim} 10^{-9}$ in relative difference, in the case of $\mathrm{^{8}B}$ neutrino) and well below the detection limits of the current neutrino detectors, when we assume the g-mode amplitude parameter $A_{n \ell}$ to be $10^{-5}$, which corresponds to the assumed maximum relative temperature perturbation inside the Sun. Thus, it is at the moment fair to say that detecting individual solar g-modes via the solar neutrino flux measurement is almost impossible. However, the net increase in the mean neutrino flux that originates from the non-time-varying component could be non-negligible. In particular, since $A_{n \ell}$ may be related to convection amplitude, which could change in accordance with the solar magnetic activity, the total net increase in the neutrino flux, which is proportional to $A_{n \ell}^2$, should also change with the solar activity cycle. Such a long-period variation~(${\sim} 11$~years) in the neutrino flux could thus be interpreted as evidence for a bunch of solar g-modes. Comparison of the theoretical prediction with the solar neutrino measurements by, e.g., Super-Kamiokande, may have a potential to put constraints on the theory of the excitation mechanism of solar g-modes.

We study the classical stability of a non-topological $Z^\prime$ string in the minimal 331 model, which arises from the maximal symmetry breaking pattern of an ${{\mathfrak s}{\mathfrak u}}(6)$ toy model. Two Higgs triplets are introduced according to the emergent global symmetries in the fermionic sector of the ${{\mathfrak s}{\mathfrak u}}(6)$ toy model, which will achieve the sequential symmetry breaking of ${{\mathfrak s}{\mathfrak u}}(3)_c\oplus {{\mathfrak s}{\mathfrak u}}(3)_W \oplus {\mathfrak u}(1)_X\to {{\mathfrak s}{\mathfrak u}}(3)_c\oplus {{\mathfrak s}{\mathfrak u}}(2)_W \oplus {\mathfrak u}(1)_Y$. By analyzing small perturbations around the string background and solving the coupled Helmholtz equations numerically, we find that the string is stable only near the semilocal limit of $\vartheta_S \approx \fracπ{2}$, even when Higgs self-couplings are tuned to minimize instabilities. This suggests that such non-topological strings are unlikely to exist in unified theories based on ${{\mathfrak s}{\mathfrak u}}(N>5)$ Lie algebras.

We combine the highest resolution N-body simulation of a $\sim 10^{12}\, M_\odot$ $Λ$CDM halo (Aquarius-A) with the {\sc GALFORM} galaxy formation semianalytic model to study the full satellite population expected in a MW-like galaxy. The model assumes that galaxies only form in subhalos whose peak circular velocity exceeds the H-cooling threshold, all of which are well resolved in the simulation. The number of luminous subhalos ever accreted into the main halo is thus well defined, and implies that the total number of MW satellites, down to arbitrarily low luminosity, should not exceed a few hundred. The model tracks satellites even after their halos cease to be resolved ("orphan" galaxies), and includes a novel treatment of dark matter and stellar tidal stripping which takes into account that all $Λ$CDM subhalos survive until the present because of their cuspy inner density profiles. After accounting for tides, our results match well the massive end of the observed MW satellite mass function and predict that a large number of ultra-faint dwarfs are missing from the current MW satellite census. The missing UFDs are predicted to avoid the innermost regions of the host, and to have properties that overlap with those of the many ultra-faint compact MW satellites (UFCSs) discovered recently, with properties intermediate between globular clusters and dwarf galaxies. Our results suggest that many UFCS systems are dark matter-dominated dwarfs with velocity dispersions between $1-3$km/s, which have survived disruption because they reside in the dense cusp of $Λ$CDM subhalos. UFCSs should have mean densities of order $10^{10}$-$10^{11}\,M_\odot/$kpc$^3$, higher than those of more extended ultra-faint systems. If confirmed, our results would provide support for the cuspy nature of $Λ$CDM dark matter halos and for the hydrogen-cooling threshold for galaxy formation.

The vertical redistribution of materials in the lunar regolith - ranging from continuously produced space-weathering products to sporadic pulses of supernova- or kilonova-derived isotopes - remains a fundamental problem in planetary science. We present a unified stochastic model of regolith gardening induced by the impact flux. Treating gardening as a competition between impact-driven advection and diffusion predicts the maturity profiles of Apollo cores over more than two orders of magnitude in time ($1.4 \times 10^7$ to $4.5 \times 10^8$ years). This model describes well the depth profiles of live Fe60 in Apollo regolith samples, suggesting that supernova dust capture is independent of native iron abundance, and is consistent with a uniform influx at the latitudes of the Apollo landing sites. We extend our model to predict lunar signals for live r-process species that might originate from supernovae or kilonovae: Pu244 tied to terrestrial detections, and I129, Hf182, and Cm247 based on r-process calculations. The Pu244/Fe60 depth profile can probe the origin of Pu244, motivating searches in Artemis regolith samples down to depths O(100) cm.

Low-mass galaxies with intense starbursts exhibit spectra dominated by extreme nebular emission and faint stellar continua. These extreme emission-line galaxies (EELGs) are key laboratories to study star formation, feedback, and ionizing photon escape in low-metallicity environments. We exploit the DESI survey to assemble the k-Means of Extreme Nebulae from DEsi outLiers (k-MENDEL), a statistically robust sample of ~16,000 EELGs at 0.01 < z < 0.96 selected via automatic k-means classification. Using SED fitting and Te-based metallicities, we characterize EELGs including "blueberry" and "green pea" galaxies, spanning stellar masses of 10^6-10^10 Msun and SFRs of 0.1-100 Msun/yr. k-MENDEL extends previous SDSS samples toward higher redshifts and lower metallicities (12+log(O/H) ~ 7.0-8.5). EELGs lie systematically above the star-forming main sequence, with sSFRs up to ~100 Gyr^-1. They follow a shallower mass-metallicity relation offset by 0.3-0.5 dex from local relations, closely resembling young galaxies observed with JWST at z > 3-10. The large intrinsic metallicity scatter, even after projecting along the fundamental metallicity relation, indicates strong departures from simple "bathtub" models, suggesting massive inflows of metal-poor gas followed by strong feedback. While ~6% of the sample shows AGN-like signatures, the most extreme star-forming systems reach high ionization (O32 ~ 5-60) comparable to confirmed Lyman-continuum emitters. Our results support the interpretation of EELGs as short-lived, non-equilibrium phases in the evolution of low-mass galaxies and highlight their importance as nearby analogs of galaxies likely driving cosmic reionization (Abridged).

Recent XRISM observations of active galactic nuclei such as PDS 456 have revealed ``forests'' of absorption lines best modeled by five distinct absorption zones with varying large blueshifts. We propose a model in which these relativistic blueshifts originate from the motion of the accretion disc itself, rather than from a clumpy super-Eddington outflow at hundreds of gravitational radii $r_g\equiv GM/c^2$. We demonstrate that thin rings of absorbing material lying just above the accretion disc at varying radii can produce the observed energy shifts and separations of the absorption zones. In this model, the PDS 456 transmission spectrum is well reproduced by rings with widths $Δr\lesssim1r_g$ at locations between the black hole's innermost stable circular orbit (ISCO) and $\approx15r_g$. This model suggests that the absorption forests seen in XRISM observations can probe the surface structure of the innermost ($\lesssim15r_g$) regions of quasar accretion discs.

The 1D flux power spectrum ($P_{\mathrm{1D}}$) of the Ly$α$ forest provides an exceptionally high-resolution probe of structure formation down to small scales ($k\approx1-10~\text{$h~$Mpc$^{-1}$}$). These scales carry the imprints of massive neutrinos, warm dark matter, and the running of the primordial power spectrum spectral index. The effective field theory (EFT) is a promising perturbative approach to systematically and efficiently describe the Ly$α$ forest, but it faces challenges in its application to $P_{\mathrm{1D}}$, as many EFT parameters become degenerate when projected along the line of sight. In addition, this projection generates new stochastic terms from the integration over small-scale modes. In this work, we address these issues by compressing the EFT model space using the Fisher matrix formalism and linearizing the resulting compression directions, enabling analytic template marginalization and significantly reducing the computational cost of likelihood evaluation. We use hydrodynamical simulations to obtain a baseline estimate of EFT parameters, and use the DESI DR1 $P_{\mathrm{1D}}$ measurements to derive compression directions. We then marginalize over deviations from the baseline using these compression directions and forecast the constraining power of our formalism. We find that even in conservative scenarios where each data redshift bin requires its own set of EFT parameters, the cosmological constraints saturate with the linear bias, two leading-order 1D stochastic terms, and three principal combinations of the remaining EFT templates. In this case, our forecasted precision of the amplitude ($Δ^2_p$) and the logarithmic slope ($n_p$) of the linear matter power spectrum at the pivot scale ($k_p=0.7~\text{Mpc}^{-1}$) is $10\%$ and $2.0\%$, respectively, which is similar to emulator-based analyses that include observational data systematics.

''Little Red Dots'' (LRDs) are broad-line sources at high redshift, initially identified by their compact morphologies, red colours and prominent Balmer breaks. The origin of their optical-to-near-infrared continua is debated, with proposed explanations ranging from direct recombination emission to thermalised blackbodies from stellar-like atmospheres. Here we report evidence for Paschen jumps in a subset of LRDs, consistent with free-bound recombination to hydrogen $n=3$. The Paschen and Brackett continuum shapes across the sample are consistent with minimally reddened emission from low-temperature gas with $T_e\lesssim10\,000$ K, while the presence of Paschen jump signatures limits scenarios in which the emission is thermalised. Further, the extreme H$α$ equivalent widths and the tight observed correlation between H$α$ and the continuum follow naturally if both originate in recombination emission. This provides an observational upper limit on the contribution of any direct AGN accretion component and any stellar-atmosphere-like component, as well as on the fraction of line emission that can be thermalised as it traverses the cocoon. Ultimately, nebular radiative-transfer models provide a self-consistent explanation of the continuum, line strengths and line profiles without requiring multiple separately fitted components.

The extragalactic background light (EBL), the cumulative radiation from all extragalactic sources, traces galaxy formation and cosmic evolution. High-energy $γ$ rays attenuated via pair production with EBL photons are a powerful probe of the EBL. In this work, we use very-high-energy (VHE; $E_γ> 100\,\mathrm{GeV}$) $γ$ rays to measure the local EBL intensity and test its consistency with galaxy counts and direct measurements. Our analysis employs a sample of 268 spectra from 45 sources observed with Imaging Atmospheric Cherenkov telescopes. A model-dependent study shows seven EBL templates require only $\le 10\%$ rescaling to fit the observed $γ$-ray attenuation. The galaxy-count-anchored model gives the closest match. We then derive template-marginalized TeV optical depths from a representative model subset. We combine them with \textit{Fermi}-LAT GeV measurements to reconstruct the EBL at $z = 0$ using empirical and physically motivated models. The two reconstructions agree and follow the integrated galaxy light to within $2$--$3\,\mathrm{nW\,m^{-2}\,sr^{-1}}$ (typically $<25\%$) over $0.5$--$30\,μ$m. Both are consistent with low-zodiacal-light observations, including outer solar system and dark cloud measurements. In contrast, the near-IR excess reported by IRTS and CIBER exceeds our reconstructed intensity by $3$--$5σ$, implying an additional $\gtrsim 5$--$10\,\mathrm{nW\,m^{-2}\,sr^{-1}}$ incompatible with the $γ$-ray optical depths. Combined with GeV constraints on EBL evolution to $z \simeq 4$, these TeV optical depths provide a VHE-anchored determination of the local EBL intensity. The agreement with galaxy counts and deep-space measurements indicates that known galaxy populations account for most of the optical and near-IR background, leaving limited room for an additional diffuse component.

The relationship between the star formation rate surface density and the molecular gas surface density in galaxies is key to understanding galaxy evolution. To investigate the molecular Kennicutt-Schmidt (K-S) relation and its dependence on gas density, we analyze a uniform sample of 36 nearby galaxies from the AMISS survey, focusing on the CO(1-0), CO(2-1), and CO(3-2) transitions, which trace progressively denser and warmer molecular gas. Using statistical methods that combine binning with Markov Chain Monte Carlo (MCMC) fitting, we derive the slope, scatter, and intercept of the $Σ_{\mathrm{SFR}}$-$Σ_{\mathrm{CO}}$ relation for each transition. We find power-law slopes of 1.26, 1.14, and 1.07 for CO(1-0), CO(2-1), and CO(3-2), respectively, consistent with a trend toward increasingly linear star formation relations at higher-J transitions. This behavior supports the idea that denser gas is more directly linked to ongoing star formation and is consistent with previous findings of near-linear correlations between HCN or high-J CO luminosities and global SFR. The observed trend suggests an underlying relation between gas and SFR volume densities with a power-law index of $\sim$1.5, indicating enhanced star formation efficiency in denser environments. These findings underscore the critical role of dense gas in regulating star formation and highlight the importance of tracer selection and excitation conditions when interpreting the K-S relation across different environments.

We present ALMA detections (or stringent upper limits) of the [CII] 158 $μm$ emission line and underlying dust continuum from five massive quenched galaxies (QGs) at 2<z<4.7. We find extreme variations in the molecular gas fractions ($\rm{f_g=M_{mol}/M_{\star}}$), spanning 0.1%-25%, if a standard $\rm{α_{[CII]}}$ applies. We attempt a first empirical calibration of $\rm{α_{[CII]}}$ with respect to dust continuum in a $z=2$ lensed QG and with respect to CO(3-2) in a $z=3.1$ QG, finding no evidence of strong deviations from the standard value. Dust continuum measurements, coupled with JWST/MIRI fluxes, suggest higher dust temperatures compared to expectations from $z<2$ QGs, reaching $T_{d}\sim40-50 \,K$ in two galaxies. Coupled with remarkably high total infrared luminosities (LIR) not explained by observed JWST colors not by energy balance based on literature dust extinction measurements, and with [CII] deficits down to $\rm{[CII]/LIR\sim 2\times10^{-4}}$ typical of (Ultra)Luminous Infrared Galaxies, our findings point to additional dust-heating mechanisms other than dust-absorbed stellar radiation. Surprisingly, JWST/NIRCam and ALMA imaging reveal widespread disturbed stellar morphologies and offsets/tails in dust and gas, indicative of ongoing interactions. While larger samples are needed to assess how common these features are in high-z QGs, these findings support a merger-driven origin for the phenomenology observed in these systems, with key similarities with respect to local post-starburst galaxies where low-velocity shocks and turbulence also inject energy into the residual ISM.

We aim to provide a systematic and quantitative description of the hydrogenation network connecting HCN and HNC to methylamine on interstellar water ices, while identifying dominant pathways and bottlenecks. To this end, we performed a comprehensive quantum-chemical investigation of H-addition, H-abstraction, reactions with H2, and water-assisted H-transfer isomerization, covering intermediates linking HCN and HNC to CH3NH2. Calculations were carried out on amorphous solid water clusters of 14 molecules. Using benchmarked density functional theory, we derived activation barriers, elucidated mechanisms, and determined the binding energy distribution of H2CN and CNH2, also assessing deuterium substitution effects. H-addition reactions generally involve activation barriers, except for radical species. Considering both barrier heights and tunneling crossover temperatures, the most favorable sequence originates from HNC rather than HCN. The network evolves toward methanimine (H2CNH), the central species, or the singlet carbene HC:NH2, from which further hydrogenation leads to methylamine. Along these paths, several reactions are barrierless, while some H-abstraction processes compete with addition. Reactions involving H2 are uncommon, as most are endoergic. Deuterium substitution weakly affects classical barriers but significantly influences tunneling efficiencies. Our results support efficient formation of methanimine and methylamine from HNC on cold interstellar ices, with methanimine acting as a chemical sink, whereas HCN is less reactive and more likely to persist. These findings provide quantitative constraints for astrochemical models.

Supernova (SN) 1987A provides a unique laboratory for investigating many aspects of SN physics and evolution. An observation at Day 12927 (35.4 yr) since the explosion with the Mid-Infrared Instrument (MIRI) Medium Resolution Spectrometer (MRS) on the James Webb Space Telescope (JWST) provided the first spatially resolved spectroscopic study of SN 1987A in the mid-IR, yielding insights into the evolution of dust, the ejecta, the equatorial ring (ER), and shocks in the system. Here we present a second epoch with MIRI/MRS at Day 13311 (36.4 yr) allowing the mid-IR spatially resolved spectroscopic temporal evolution of SN 1987A to be probed for the first time. Analysis of the ER-dominated dust continuum showed little evolution between Days 12927 and 13311. However, a spatial analysis reveals the inner ER to be fading while the outermost regions are brightening. Broad ejecta emission lines detected at Day 12927 are evolving rapidly, driven by the recent onset of the ejecta/equatorial ring interaction in the northeast and southwest of the ER. Most lines from the ER show no change during the 384 days between the epochs, though some such as [Ne II] and [Ar II] have faded. We identify mid-IR H2 emission associated with the ejecta for the first time. Using the near- and mid-IR [Fe II] lines as density and temperature diagnostics of the ejecta in the interaction region we find it likely that the dense inner Fe-rich ejecta has now reached the reverse shock. Continued monitoring of SN 1987A is essential to observe the evolving ejecta/ER interaction and dust components.

Utilising the optical imaging Fourier transform spectrograph SITELLE, the Star-formation, Ionized Gas and Nebular Abundances Legacy Survey (SIGNALS) is designed to study the connection between star-forming regions and their environments. Targeting $31$ local star-forming galaxies, its data products also lend themselves to planetary nebula (PN) surveys. We present here a new pipeline to find PNe using automated emission-line diagnostics and morphology tests, that is able to distinguish PNe from contaminants with an accuracy similar to that of past visual methods. We also perform thorough completeness tests using mock PNe inserted into the data cubes with full spectra. We apply these tools to a pilot sample of two dwarf irregular galaxies from the SIGNALS survey, NGC 4214 and NGC 4449, with other galaxies to follow. For these two galaxies, we identify $25$ PNe (including $6$ new discoveries) and $23$ PNe (including $13$ new discoveries), respectively, and calculate PN luminosity function distances of $3.09^{+0.25}_{-0.46}$ and $3.91^{+0.33}_{-0.52}$ Mpc, respectively, the latter consistent with previous estimates. We also calculate the bolometric PN specific frequency of our galaxies ($α_\mathrm{bol}$), as well as a newly defined $V$-band PN specific frequency ($α_\mathrm{V}$) based solely on the galaxies' total luminosities in that band.

We develop a self-consistent analytical two-fluid framework for plasma evolution in the short-time regime, elucidating the fundamental mechanism underlying the coupled generation of flow and magnetic fields. We show that consistency between ion momentum and mass conservation imposes a structural constraint on the system: the total pressure must satisfy the Laplace equation, $\nabla^2 P = 0$. This constraint enables a class of exact analytical solutions in which pressure gradients simultaneously drive plasma flow and generate magnetic fields through a Biermann-type mechanism. Using representative parameters, we obtain magnetic-field strengths and flow velocities consistent with both laser-produced plasmas and large-scale astrophysical systems. This framework provides a unified description of pressure-driven magnetogenesis and plasma flow in the short-time regime.

Stacked prism lenses (SPLs) are a type of refractive X-ray optics currently under development with the potential to greatly improve on current X-ray telescope optics in terms of focal length, angular resolution, efficiency and scalability. For this work, SPLs are manufactured using two-photon polymerization (2PP), with production being significantly faster and with higher geometric fidelity than previous methods. Preliminary laboratory tests show improved efficiency compared to previous manufacturing methods and promising optical capabilities. Two-photon polymerization is shown to be a reliable method for producing SPLs, and when challenges around printing time and assembly are addressed, the path towards an SPL X-ray telescope lies open.

JWST has revealed a large population of massive black holes (BHs) in the early Universe with unusual properties which mark them as distinct from low-redshift active galactic nuclei. Such findings have prompted the development of new models of BH formation and growth, and of their co-evolution with host galaxies. Linking the gas-phase metallicity of BH environments to seed masses is key to understanding which evolutionary pathways could explain the population of JWST-discovered BHs. We present new high-resolution JWST NIRSpec/IFU observations covering the rest-frame optical emission lines of a Little Red Dot (LRD) at $z=3.55$, known as The Cliff, from the `Red Unknowns: Bright Infrared Extragalactic Survey' (RUBIES). We find evidence for low metallicity ($Z=0.017\pm0.004 \ Z_\odot$) based on the low narrow-line [OIII]$\lambda5007$/H$β$ ratio, supported by the non-detection of low-ionisation emission lines such as [OII]$λ\lambda3727,3729$ and [NII]$λ\lambda6548,6583$. We find that the observed properties of The Cliff, including its overmassive BH, can be reproduced by some simulations of black hole growth and evolution down to $z\sim3.5$. However, these simulation runs require high seed masses ($10^4 - 10^5\ M_\odot$) and appear as rarely in the simulation volume as in the RUBIES survey volume over redshifts $3<z<4$, highlighting the unusual nature of The Cliff. Future simulations and numerical models will help to uncover how such a metal poor system managed to develop a massive black hole and persist to such low redshift.

We present an extended optical monitoring of the quadruply-imaged gravitationally lensed quasar QSO 2237+0305, the Einstein Cross, including observations from different observatories in both hemispheres and using a new photometric technique. This technique uses a region far enough from the lens system to determine accurately the sky background level, and minimises contamination from the lensing galaxy by combining analytical and numerical modeling of its structure. The resulting light curves of the four quasar images describe variations across practically the entire optical spectrum and span about 9000 days in the $VRI$ bands. The multi-band microlensing variability is captured with an unprecedented level of detail, and a preliminary microlensing analysis reveals an almost linear scaling of source radius with wavelength, providing direct evidence for the wavelength-dependent structure of the region contributing to optical passband fluxes. Specifically, assuming a mean microlens mass $\langle M \rangle$ = 0.3 $\rm{M_{\odot}}$ and concentric Gaussian sources that move according to the velocity distribution peaks (speed and direction) reported in a previous microlensing analysis, we find that the half-light radius of the $g$-band source is 9.6 $\pm$ 2.7 lt-day and the size of the sources grows with wavelength with a power-law index of $α$ = 0.94 $\pm$ 0.05. We conclude that these long-term light curves set stringent empirical constraints on models of quasar emission and microlensing physics.

We apply discrete Morse theory, global topology, and persistent homology to characterize the impact of massive neutrinos on the multiscale cosmic web, focusing on filaments. The topology of the cosmic web is sensitive to neutrino imprints, and persistence diagrams provide more information than commonly used summary statistics by quantifying the longevity of topological features across densities. This scale-adaptive, parameter-free formalism is powerful, as massive neutrinos affect halos, walls, filaments, and voids in distinct ways. Within this framework, we simultaneously assess their impact on tracers and skeleton structures and capture their multiscale signals across cosmic time. Discrete Morse theory is also well suited for particle-based neutrino implementations, often affected by Poisson shot noise, as it preserves the salient features of the underlying smooth field. Using two independent sets of N-body simulations, we present filament statistics and persistence diagrams in massive-neutrino cosmologies. Our results show that neutrinos leave distinct imprints on filaments and skeleton connectivity, producing mass-dependent signatures most pronounced at high redshift (z~2) and detectable at the few-percent level for masses as small as $M_ν\sim 0.1$ eV. Filaments thus provide an ideal environment for isolating neutrino effects. We also compare two implementations of massive neutrinos to assess systematics. Our study establishes a promising avenue for leveraging cosmic web topology, persistent homology, and environment-based statistics to constrain or directly detect neutrino mass and infer the mass hierarchy - a long-standing challenge in particle physics and a major objective of ongoing and upcoming galaxy redshift surveys (e.g., DES, DESI, Euclid, Rubin-LSST).

The extragalactic background Light (EBL) from ultraviolet to infrared comprises the emission from all stars, galaxies, and actively accreting black holes in the observable Universe. A precise measurement of the EBL is critically important to probe models of star formation and galaxy evolution. The EBL can be measured via the absorption imprint left on the spectra of gamma-ray blazars. In this work, we rely on 15 years of {\it Fermi}-LAT data and 1576 blazars to measure the EBL optical depth in the $0<z<4.3$ range. We detect the EBL attenuation with $\sim23σ$ significance and measure the optical depth in 19 redshift bins, extending the coverage and improving on our previous results. This allows us to reconstruct the EBL evolution and find general consistency with recent EBL models. These results represent the most precise determination of the EBL with GeV $γ$ rays to date.

The nucleosynthetic heterogeneity between different asteroids and planets is well established. These isotopic variations manifest themselves at the part per millions level or larger, in isotopes that were synthesised in various stellar environments. To escape homogenisation, some of these isotopic signatures must have been preserved in dust, which ended up being heterogeneously distributed in the solar protoplanetary disc. The origin of the nucleosynthetic heterogeneity is still poorly constrained, potentially reflecting inherited isotope variations from the Sun's parental molecular cloud and/or processing and redistribution during the subsequent protoplanetary disc phase with thermal processing and size sorting as major processes. This chapter aims to provide a broad review of the dynamical, collisional, and thermal processes in protoplanetary discs -- from initial infall to gas dispersal -- that may have influenced the distribution and survival of the anomalous carrier phases, which finally accreted into asteroids and planets. While several of these mechanisms have been considered in past studies, they are often examined in isolation, which impedes the assessment of how their effects may be altered or amplified by additional disc processes. Size sorting in particular has received little attention, and here we highlight that this process likely occurred in the disc and can induce nucleosynthetic heterogeneity. By placing previous studies within the context of a comprehensive overview, we aim to clarify the broader physical framework in which anomalous carrier transport occurs and identify previously underexplored mechanisms that may have contributed to the final isotopic structure of the Solar System we see today.

FRB 20220912A is a highly active repeating fast radio burst (FRB) source, discovered by the Canadian Hydrogen Intensity Mapping Experiment (CHIME) using its real-time FRB detection system (CHIME/FRB). Here, we present results from a radio monitoring campaign of FRB 20220912A using CHIME, including ~200 hours of data collected by CHIME/Pulsar, spanning 1.5 years following the source's discovery. We present an analysis of a sample of 828 CHIME-detected bursts from FRB 20220912A, in the 400-800 MHz radio frequency band. The source remains highly active for ~10 weeks and has a bimodal wait-time distribution with peaks at $160^{+120}_{-70}$ ms and $306^{+14}_{-13}$ s. Assuming a radio efficiency factor of $10^{-4}$ and a beaming angle of 0.1, we estimate the total emitted energy from the source over the entire observing campaign to be $2 \times 10^{43}$ ergs. We report a 2.3$σ$ detection of a linear increase in the DM of $1.4 \pm 0.6$ pc cm$^{-3}$ yr$^{-1}$, with no significant trend in rotation measure (with a 3$σ$ upper limit of 13.4 rad m$^{-2}$ yr$^{-1}$). We contrast our findings with other active repeaters, which exhibit different DM and RM evolution to indicate that FRB 20220912A may reside in a unique local environment.

We extracted a list of 662 nearby (within $\sim16$ Mpc) HI-detection sources from the Five-hundred-meter Aperture Spherical radio Telescope (FAST) All Sky HI Survey (FASHI) and made a visual identification of them with optical counterpart. This inspection led to the discovery of 71 new dwarf galaxies. All of them are dwarf irregular galaxies with ongoing star formation. They are characterized by the following median parameters: visual magnitude of $g=17.8$ mag and color $(g-r)=0.29$ mag, HI-flux $S_\mathrm{HI}=718$ mJy km/s, HI-mass $M_\mathrm{HI}=3.7\times10^7$ $M_\odot$, as well as the HI line-width of $W_{50}=37$ km/s.

We continued the search for single pulses (SPs) in the northern part of the all-sky High Time Resolution Universe survey, whose aim is to detect pulsars and other radio transients. This search is now about 21% complete and has yielded the first discovery of a fast radio burst (FRB) with the 100 m Effelsberg Radio Telescope. FRB20110220A was detected with an S/N-optimised dispersion measure of 501.0 pc/cm$^{3}$ and a width of 11.9 $\pm$ 3.5 ms, for a fluence of 0.6 $\pm$ 0.1 Jy ms. We obtained the first L-band detection of the rotating radio transient (RRAT) J2028+28, from which we obtained upper limits on the source's period and burst rate, as well as an improved position. We also discovered a new RRAT, J0404+53, which had previously been reported as an isolated SP candidate. Eight new SP trains and 272 faint isolated SP candidates were detected too. We used these candidates to demonstrate that their all-sky detection rates depend on Galactic latitude and longitude. This direction dependence suggests the existence of a faint Galactic SP population.

With the Event Horizon Telescope and future Very Long Baseline Interferometry arrays poised to image supermassive black holes, there is an urgent need to understand dynamic aspects of small-scale structure near the supermassive black hole. In this study, we introduce the relative magnification factor to characterize point sources distributed on the surface of the accretion disk near a black hole. We investigate the influence of source motion on this factor, comparing static sources with those corotating with the disk. In contrast to the static case, which can be well-understood in the standard framework of gravitational lensing, corotating sources exhibit significant distortions in the distribution of the magnification factor on both the image and source planes, indicating that the caustic structure is substantially modulated by source motion. This magnification factor pattern encodes signatures of the kinematics of accretion flow when the time-delayed images are incorporated. This potentially offers a novel probe for investigating the interplay between spacetime geometry and properties of accretion flow.

The origin of low surface brightness (LSB) galaxies remains a key open question in galaxy formation, reflecting the balance internal mechanisms and environmental influence. Using MaNGA integral-field spectroscopy, we investigate whether LSB and high surface brightness (HSB) galaxies of comparable stellar mass ($9 < \log M_\ast < 10$) occupy distinct environments or differ primarily through internal evolution. Our late-type sample comprises 113 central and 29 satellite LSB galaxies, and 374 central and 142 satellite HSB galaxies. We characterize environments on scales from 100 kpc to 10 Mpc, analyzing radial profiles of stellar mass surface density ($Σ_\ast$), star formation activity, and gas-phase metallicity. Central LSB and HSB galaxies inhabit similarly low-density large-scale ($>$200 kpc) environments, but LSB galaxies are more isolated on small scales ($\sim$100 kpc). Even after matching in stellar mass and environment, LSB galaxies show systematically lower $Σ_\ast$, $Σ_{SFR}$, and metallicities, often hosting diffuse, weakly star-forming bulges embedded in extended disks. These results indicate that LSB structure and star formation are not primarily governed by large-scale environment or halo mass. While secondary halo properties such as spin, concentration, or gas accretion history are often invoked, their environmental dependence appears weak. Instead, LSB-HSB differences for centrals likely reflect divergent assembly or interaction histories and internal processes -- such as angular momentum-driven disk evolution or inefficient gas conversion -- largely decoupled from large-scale environment. Nonetheless, environment still influences the observed star formation and chemical differences between central and satellite LSB galaxies.

We study the effects of using the optical properties of irregular hexahedral grains in photoionization models of circumstellar nebulae around evolved stars. Dust opacities for the irregular grains were obtained from the scattering properties available in the TAMUdust2020 database and these were implemented in the spectral synthesis code cloudy. A sample of photoionization models that use opacities from both spherical and irregular hexahedral grains across a standard MRN size distribution (0.005 to 0.25 um) was produced. We consider the optical properties of graphite, amorphous carbon and silicate dust grains and find that differences between the model nebula continua calculated using spherical and irregular dust grains increase with the grain size, especially for graphite. In particular, we find that the luminosities at the infrared peak for the hexahedral grain models can be up to 60% higher than those from the equivalent spherical grain models for the largest grains. This result suggests that traditional spherical grain assumptions may lead to an overestimate of the dust mass in photoionized nebulae.

Previous studies using p-H$_2$CO $J=3$--$2$ transitions at 218 GHz suggested widespread high-temperature gas exceeding 60 K and even 100 K in the CMZ, with heating mechanisms possibly related to cosmic rays or turbulent dissipation. However, at temperatures above 100 K, p-H$_2$CO $J=3$--$2$ line emission may lead to significant overestimates of kinetic temperature. This study combines o-H$_2$CO $J=5$--$4$ data from JCMT with p-H$_2$CO $J=3$--$2$ data from APEX to analyze three molecular clouds (The Brick, Sgr A1, and Sgr A2) with high temperatures. We used the non-LTE radiative transfer code RADEX to model spectral lines and constrain physical parameters with multiple line ratios, obtaining more reliable kinetic temperatures. Our results show that the previously reported extreme temperatures ($>100$ K) based on p-H$_2$CO $J=3$--$2$ line ratios are revised downward, with the average kinetic temperatures now constrained to 84--95 K using o-H$_2$CO $J=5$--$4$ line ratios, indicating systematic overestimation in the earlier studies. Further analysis reveals that the relationship between temperature and gas line width aligns more closely with predictions from models incorporating both high cosmic ray ionization rate and turbulent heating, suggesting that these molecular clouds are likely heated by a combination of cosmic-ray and turbulent dissipation mechanisms.

Fast and reliable inference of gravitational-wave source parameters is crucial for analyzing large catalogs that are reaching the size of hundreds of detections, and for identifying short-lived electromagnetic counterparts. Neural posterior estimation has emerged as a powerful inference method, where the model is trained on simulated gravitational-wave data at considerable computational cost, but thereafter enables extremely fast and inexpensive inference at test time. Here, we extend this approach by incorporating domain-specific physical insights and methods in the model architecture. These include compressing the data by heterodyning against a reference waveform chosen via approximate likelihood maximization, removing parameter degeneracies through tailored coordinate systems, and eliminating known multimodalities by folding the parameter space. As a result, the network is approximately equivariant to changes in the source parameters, and achieves a reduced training cost and improved model interpretability. Our implementation, called labrador, can be trained end-to-end on a 1-day timescale on $\sim 10^2$ CPU cores and a V100 GPU, achieving a median importance-sampling efficiency of 1% on quadrupolar, aligned-spin signals in a broad mass range (chirp mass $\mathcal M \in 1\text{-}50\,\mathrm{M}_\odot$, mass ratio $q > 0.1$). labrador is the first neural inference code to achieve extensive coverage of long-duration signals with secondary masses $m_2 < 10\,\mathrm{M}_\odot$, rendered possible by its equivariance property. Among our novel contributions is a numerically stable procedure that enables neural posterior estimation when the simulation and inference priors differ.

We present a theoretical model for the power spectrum and bispectrum of galaxy clustering that exploits the complementarity between small-scale power spectrum information and large-scale bispectrum measurements. We extend the FOLPS code by combining its one-loop EFT galaxy power spectrum with a tree-level galaxy bispectrum projected onto the tripolar spherical harmonics (Sugiyama) basis. To access additional small-scale information, we also consider a line-of-sight damping factor in both statistics, mirroring approaches commonly used in studies of redshift-space distortions. We test the model using DESI DR2 galaxy mocks. Even without damping, the joint analysis of the EFT power spectrum and bispectrum significantly improves constraints and reduces parameter degeneracies relative to power spectrum analyses alone. For LRG-like samples, including the damping further extends the range beyond $k\sim 0.3 \,h \text{Mpc}^{-1}$ in the power spectrum and $k \sim 0.24 \,h \text{Mpc}^{-1}$ in the bispectrum without introducing statistically significant parameter biases. This leads to up to $\sim 30\%$ tighter constraints on $A_s$ and $ω_{cdm}$. For low signal-to-noise tracers such as QSOs, however, the damping parameters are weakly constrained and can absorb noise fluctuations, leading to shifts in inferred parameters. Similar limitations may arise in models where cosmological information is encoded in power-spectrum shape features degenerate with the damping, such as scenarios with massive neutrinos. In contrast, for $w_0w_a$CDM we obtain $15\%$ and $21\%$ tighter constraints on $w_0$ and $w_a$, respectively, yielding a deviation from constant dark energy at slightly more than the $1σ$ level using full-shape information alone. The code is publicly available at https://github.com/cosmodesi/FolpsD

cTreeBalls (cBalls for short) is a Python/C package useful to measure (2,3)-point clustering statistics. cBalls can efficiently calculate 3-point correlations of more than 200 million HEALPix pixels ( a full sky simulation with Nside = 4096) in less than 10 minutes on a single high-performance computing node, enabling a feasible analysis for the upcoming LSST data. It builds upon octree (Barnes & Hut, 1986) and kd-tree algorithms (Bentley, 1975), and supplies a user-friendly interface with flexible input/output (I/O) of catalogue data and measurement results, with the built program configurable through external parameter files and tracked through enhanced logging and warning/exception handling. For completeness and complementarity, methods for measuring two-point clustering statistics for periodic boxes are also included in the package. cTreeBalls was developed for its use in the Dark Energy Science Collaboration (DESC) of the Rubin Observatory Legacy Survey of Space and Time (LSST).

Most scientific research begins in the context of the then-contemporary state of knowledge of the field and moves toward a deeper understanding of the subject. This colloquium presents the Arecibo telescope$^{'}$s contribution to radio astronomy from the point of view of its contemporary impact and relevance and how that evolved over the 57 years of its long life. Sometimes, serendipitous discoveries at Arecibo and elsewhere brought revolutionary changes to the field. Further, significant upgrades to the reflector, optics, receiving, and data-taking systems helped clear a pathway for a leap ahead. Charting these movements through time and without any claim to completeness, this Colloquium presents a progression through Arecibo telescope$^{'}$s role in the history of radio astronomy.

The detection of gravitational waves (GWs) from a binary neutron star (BNS) merger by Advanced LIGO and Advanced Virgo (GW170817), together with its electromagnetic counterpart, the short gamma-ray burst GRB~170817A, heralded the birth of multi-messenger astronomy. The detection of TeV emission from GRBs motivates follow-up observations with the Cherenkov Telescope Array Observatory (CTAO), ideal for detecting such signals due to its unprecedented sensitivity, rapid response, and wide-field survey capabilities. The aim of this work is to evaluate GeV--TeV GW follow-up strategies for CTAO using a multi-step simulation pipeline and to estimate the expected rate of joint GW-GRB detections during observing run O5. Using a simulated sample of BNS systems with corresponding GW detections, gamma-ray emission is simulated through phenomenological prescriptions based on the observed population of short GRBs, including off-axis jet scenarios. CTAO observations are simulated to account for instrument response, sky tiling strategies, integration times, and varying observing conditions. Strategies with variable and constant integration times are investigated. We find that, via an optimized follow-up strategy, about 5% of simulated GW-associated short GRBs produce GeV--TeV radiation detectable by CTAO. Detectability is strongly influenced by the jet opening angle and viewing angle, suggesting that even rough estimates of the viewing angle in GW alerts could enhance targeting. This framework motivates future follow-ups of GW-detectable events, including neutron star-black hole mergers, and further supports the development of advanced strategies incorporating galaxy distributions and synergies with future detectors such as the Einstein Telescope.

We report the discovery of GLIMPSED-329380, a z=6.64 galaxy behind Abell S1063, which shows signs of an extreme ionised outflow driven by an active galactic nucleus (AGN). The deep JWST/NIRSpec medium grating observations show spatially resolved structures of a host galaxy containing the very fast outflow and an AGN, which we analyse separately. The outflow, mainly traced by broad [O III]λ5008 and Hα emissions in the host, reaches a full-width half-maximum velocity of ~5500km/s, velocities only observed in AGN-dominated systems. From the Balmer decrement, we observe that while the narrow emission lines show no dust attenuation, the outflowing gas is dusty. We use emission lines diagnostics to infer gas abundances within the host galaxy. The oxygen abundance is 12+log(O/H) ~ 7.95 (~18% solar) and the host is slightly nitrogen-enriched with log(N/O) ~ -0.75. Despite its extreme velocity, the mass loading factor (<0.1) and the kinematic energy of the outflow (~10^43 erg/s) suggest limited impact on star formation. The AGN component shows many similarities with little red dots (LRDs): a characteristic "V-shape", exponential profile in hydrogen lines, numerous detection of forbidden [Fe II] lines, a Balmer break, and a broad absorption feature at ~4550 Å. This detection of a fast outflow in an LRD, rare in surveys dominated by low-resolution (e.g. PRISM) spectra, provides direct evidence of AGN activity in these systems.

The Galactic core-collapse supernova (SN) rate is estimated at $\approx 1-3$ per century; however, no optically visible SN has been discovered in the past 400 years. Although records of the last optically detected SN (Cassiopeia A) are debated, it is revealed today via its bright, variable mid-infrared (MIR) dust echoes -- offering the possibility of identifying dust-obscured, missed events via their dust echoes. We present the first all-sky, untargeted search for thermal dust echoes of luminous Galactic transients using difference imaging on 12 years of time-resolved NEOWISE co-adds (spanning $2009-2022$) followed by statistical detection of variable extended sources. We use echo features around Cas A, together with archival catalogs to train a convolutional neural network to classify transient candidates as dust echoes, point sources, artifacts, and high proper motion stars. Our model achieves $\approx 94$% accuracy in distinguishing echoes from other variable sources. Applying the classifier to $\approx 11$ million transient candidates, we search for spatial over-densities of echoes across the Galactic plane. We find that Cas A is the only region exhibiting echoes at the WISE sensitivity threshold of $W2$ surface brightness of $\approx 20$ Vega mag arcsec$^{-2}$ -- reflecting its unique combination of young age and luminous shock breakout. We present the largest catalog of time-resolved echo positions of Cas A (20477 within 10$^\circ$) that are being used for studies of the surrounding interstellar medium with the James Webb Space Telescope. Our results lay the groundwork for the imminent Roman space telescope surveys -- which will achieve $\approx 100\times$ higher sensitivity and $\approx 30\times$ better spatial resolution at wavelengths of $\lesssim 2.5\,μ$m.

Whether the "Hubble sequence" of galaxy morphologies exists up to z~4 is still disputed, and one of the challenges is characterizing galaxy structure consistently across a wide range of redshifts. To enable a fair comparison across cosmic time, we constructed "absolute" images of galaxies spanning 0.15<z<4.5 and 8<log $M_{\star}$<11 from HST CANDELS and JWST CEERS surveys, by matching the effective resolution and surface brightness limit of galaxies, accounting for cosmological dimming and evolution in size and mass-to-light ratio. We measured the structural parameters of 2825 galaxies and used the UMAP technique to study the evolution of the morphological phase space. We find a continuous sequence spanning late-type to early-type galaxies, with no redshift gradient - indicating that a Hubble-like sequence is established by z~4. We show that our approach recovers a cleaner separation between early- and late-type galaxies than visual classifications. By tracing progenitors using empirical mass assembly histories, we find that progenitors of low-mass galaxies are predominantly star-forming disks at all epochs. Progenitors of massive galaxies follow two distinct paths: a stable star-forming disk population with little structural evolution, and an early-type population that builds up rapidly from irregular progenitors and quenches within a few Gyr, consistent with a compaction-driven quenching scenario.

The Milky Way is surrounded by a hot diffuse circumgalactic medium (CGM) with temperatures of millions of degrees. Recent X-ray observations with the eROSITA satellite discovered a significant temperature asymmetry of this hot CGM, with the southern hemisphere being on average hotter than the northern one by a relative difference of $Δ T/T \approx 12\%$, where $T$ is averaged over the entire CGM. In this Letter, we investigate whether the passage of the Magellanic Clouds can be responsible for this asymmetry by means of a hydrodynamical/N-body simulation. In the simulation, the Magellanic Clouds induce a relative motion of the Milky Way's disc of up to 40 km/s. This motion leads to compression of the CGM gas in the southern hemisphere, resulting in an overall temperature increase in that region. We estimate a south-north temperature difference of $Δ T/T \approx 13-20\%$, consistent with the observations. We find that this temperature asymmetry is a recent phenomenon that began ~100 Myr ago.

We consider a scenario in which the Milky Way (MW) and M31 have had a previous pericentric passage, and investigate its compatibility with self-interacting dark matter (SIDM). Using initial conditions sampled from Local Group (LG) analogues in the IllustrisTNG simulation, we perform controlled re-simulations of the MW-M31 orbit, evolving the system under both standard cold dark matter (CDM) and various SIDM cross-sections. We find that the deep baryonic potential of the MW preconditions the halo's thermal structure, establishing an initial negative temperature gradient. This drives SIDM halos to bypass the standard core-formation phase and enter immediate core-collapse, resulting in monotonically increasing central densities. In full orbital simulations, the compact stellar component (disk/bulge) of the MW analog remains robust against tidal disruption for pericenter distances as close as $r_{\rm peri}\lesssim20$ kpc during an encounter at cosmic time $\sim8$ Gyr. The diffuse stellar halo is comparatively more susceptible, facing disruption for $r_{\rm peri}\lesssim100$ kpc. Our results demonstrate a dichotomy in structural evolution: the compact disk/bulge is sensitive to intrinsic SIDM thermodynamics but dynamically robust against the pericenter encounter, whereas the diffuse stellar halo is largely independent of the specific SIDM model but more vulnerable to orbital tidal disruptions.

The inverse energy cascade in turbulent Taylor-Couette flow is studied in line with the results of the large eddy simulation. The simulation results show that the inverse energy cascade first occurs within the core region of the flow channel of the Taylor-Couette flow at higher Reynolds number. It is uncovered that this phenomenon is induced by the pulsed zero shear stress resulting from the singularities of the Navier-Stokes equation. In the core area between the two cylinders, the shear stress is nearly zero at higher Reynolds number. The turbulence generated there has high turbulent energy due to discontinuity of the tangential velocity. Since the energy transfer between the fluid layers is inhibited due to the low shear stress, the turbulent energy cannot be transferred along the radial direction, and small-scale vortices with high turbulent energy are produced. These small-scale vortices are located with the large-scale vortices and cannot be dissipated owing to low shear stress. A peak in the energy spectrum at middle frequency (or wave number) is formed due to the concentration of the small-scale vortices. As the number of the singular points of the Navier-Stokes equation increases with the increasing Reynolds number, the region with zero shear stress expands along the radial direction, intensifying nonlinear instability and energy accumulation. This, in turn, leads to more prominent peaks in the energy spectrum, resulting in a more pronounced inverse energy cascade.