arXiv Daily Digest - 2026-04-14

Profiled Sparse Networks (PSN) replace uniform connectivity with deterministic, heterogeneous fan-in profiles defined by continuous, nonlinear functions, creating neurons with both dense and sparse receptive fields. We benchmark PSN across four classification datasets spanning vision and tabular domains, input dimensions from 54 to 784, and network depths of 2--3 hidden layers. At 90% sparsity, all static profiles, including the uniform random baseline, achieve accuracy within 0.2-0.6% of dense baselines on every dataset, demonstrating that heterogeneous connectivity provides no accuracy advantage when hub placement is arbitrary rather than task-aligned. This result holds across sparsity levels (80-99.9%), profile shapes (eight parametric families, lognormal, and power-law), and fan-in coefficients of variation from 0 to 2.5. Internal gradient analysis reveals that structured profiles create a 2-5x gradient concentration at hub neurons compared to the ~1x uniform distribution in random baselines, with the hierarchy strength predicted by fan-in coefficient of variation ($r = 0.93$). When PSN fan-in distributions are used to initialise RigL dynamic sparse training, lognormal profiles matched to the equilibrium fan-in distribution consistently outperform standard ERK initialisation, with advantages growing on harder tasks, achieving +0.16% on Fashion-MNIST ($p = 0.036$, $d = 1.07$), +0.43% on EMNIST, and +0.49% on Forest Cover. RigL converges to a characteristic fan-in distribution regardless of initialisation. Starting at this equilibrium allows the optimiser to refine weights rather than rearrange topology. Which neurons become hubs matters more than the degree of connectivity variance, i.e., random hub placement provides no advantage, while optimisation-driven placement does.

This paper argues that Large Language Models (LLMs) should incorporate explicit mechanisms for human empathy. As LLMs become increasingly deployed in high-stakes human-centered settings, their success depends not only on correctness or fluency but on faithful preservation of human perspectives. Yet, current LLMs systematically fail at this requirement: even when well-aligned and policy-compliant, they often attenuate affect, misrepresent contextual salience, and rigidify relational stance in ways that distort meaning. We formalize empathy as an observable behavioral property: the capacity to model and respond to human perspectives while preserving intention, affect, and context. Under this framing, we identify four recurring mechanisms of empathic failure in contemporary LLMs--sentiment attenuation, empathic granularity mismatch, conflict avoidance, and linguistic distancing--arising as structural consequences of prevailing training and alignment practices. We further organize these failures along three dimensions: cognitive, cultural, and relational empathy, to explain their manifestation across tasks. Empirical analyses show that strong benchmark performance can mask systematic empathic distortions, motivating empathy-aware objectives, benchmarks, and training signals as first-class components of LLM development.

While Diffusion Large Language Models (dLLMs) have emerged as a promising non-autoregressive paradigm comparable to autoregressive (AR) models, their faithfulness, specifically regarding hallucination, remains largely underexplored. To bridge this gap, we present the first controlled comparative study to evaluate hallucination patterns in dLLMs. Our results demonstrate that current dLLMs exhibit a higher propensity for hallucination than AR counterparts controlled for architecture, scale, and pre-training weights. Furthermore, an analysis of inference-time compute reveals divergent dynamics: while quasi-autoregressive generation suffers from early saturation, non-sequential decoding unlocks potential for continuous refinement. Finally, we identify distinct failure modes unique to the diffusion process, including premature termination, incomplete denoising, and context intrusion. Our findings underscore that although dLLMs have narrowed the performance gap on general tasks, their distinct hallucination mechanisms pose a critical challenge to model reliability. Our code is available at https://github.com/ZeroLoss-Lab/Lost-in-Diffusion

Graph neural networks have demonstrated excellent applicability to a wide range of domains, including social networks, biological systems, recommendation systems, and wireless communications. Yet a principled theoretical understanding of their generalization behavior remains limited, particularly for graph classification tasks where complex interactions between model parameters and graph structure play a crucial role. Among existing theoretical tools, PAC-Bayesian norm-based generalization bounds provide a flexible and data-dependent framework; however, current results for GNNs often restrict the exploitation of graph structures. In this work, we propose a topology-aware PAC-Bayesian norm-based generalization framework for graph convolutional networks (GCNs) that extends a previously developed framework to graph-structured models. Our approach reformulates the derivation of generalization bounds as a stochastic optimization problem and introduces sensitivity matrices that measure the response of classification outputs with respect to structured weight perturbations. By imposing different structures on sensitivity matrices from both spatial and spectral perspectives, we derive a family of generalization error bounds with graph structures explicitly embedded. Such bounds could recover existing results as special cases, while yielding bounds that are tighter than state-of-the-art PAC-Bayesian bounds for GNNs. Notably, the proposed framework explicitly integrates graph structural properties into the generalization analysis, enabling a unified inspection of GNN generalization behavior from both spatial aggregation and spectral filtering viewpoints.

Personalized learning systems are almost universally designed around a single objective: help people acquire knowledge and skills more efficiently. We argue this framing misses the more consequential problem. The most damaging failures in human life-financial ruin, health collapse, professional obsolescence-are rarely caused by insufficient knowledge acquisition. They arise from the systematic absence of entire conceptual territories from a person's cognitive map: domains they never thought to explore because, from within their existing worldview, those domains did not appear to exist or to matter. We call such absences Ontological Blind Spots and introduce Failure Ontology (F), a formal framework for detecting, classifying, and remediating them across a human lifetime. The framework introduces three original contributions: (1) a four-type taxonomy of blind spots distinguishing domain blindness, structural blindness, weight blindness, and temporal blindness; (2) five convergent failure patterns characterizing how blind spots interact with external disruption to produce catastrophic outcomes; and (3) the Failure Learning Efficiency Theorem, proving that failure-based learning achieves higher sample efficiency than success-based learning under bounded historical data. We illustrate the framework through historical case analysis of the 1997 Asian Financial Crisis and the 2008 subprime mortgage crisis, and through alongitudinal individual case study spanning five life stages.

We introduce Agent^2 RL-Bench, a benchmark for evaluating agentic RL post-training -- whether LLM agents can autonomously design, implement, and run complete RL pipelines that improve foundation models. This capability is important because RL post-training increasingly drives model alignment and specialization, yet existing benchmarks remain largely static: supervised fine-tuning alone yields strong results, leaving interactive RL engineering untested. Agent^2 RL-Bench addresses this with six tasks across three levels -- from static rule-based training to closed-loop online RL with trajectory collection -- each adding a structural requirement that prior levels do not impose. The benchmark provides isolated workspaces with a grading API, runtime instrumentation that records every submission and code revision, and automated post-hoc analysis that generates structured run reports, enabling the first automated diagnostic of agent-driven post-training behavior. Across multiple agent stacks spanning five agent systems and six driver LLMs, we find that agents achieve striking interactive gains -- on ALFWorld, an RL-only agent improves from 5.97 to 93.28 via SFT warm-up and GRPO with online rollouts -- yet make only marginal progress on others (DeepSearchQA: +2.75 within evaluation noise), and that driver choice has a large effect on interactive tasks -- within the same scaffold, switching drivers changes interactive improvement from near-zero to +78pp. More broadly, the benchmark reveals that supervised pipelines dominate agent-driven post-training under fixed budgets, with online RL succeeding as the final best route only on ALFWorld. Code is available at https://github.com/microsoft/RD-Agent/tree/main/rdagent/scenarios/rl/autorl_bench.

Time series foundation models (TSFMs) have recently achieved remarkable success in universal forecasting by leveraging large-scale pretraining on diverse time series data. Complementing this progress, incorporating frequency-domain information yields promising performance in enhancing the modeling of complex temporal patterns, such as periodicity and localized high-frequency dynamics, which are prevalent in real-world time series. To advance this direction, we propose a new perspective that integrates explicit frequency-domain representations into scalable foundation models, and introduce WaveMoE, a wavelet-enhanced mixture-of-experts foundation model for time series forecasting. WaveMoE adopts a dual-path architecture that jointly processes time series tokens and wavelet tokens aligned along a unified temporal axis, and coordinates them through a shared expert routing mechanism that enables consistent expert specialization while efficiently scaling model capacity. Preliminary experimental results on 16 diverse benchmark datasets indicate that WaveMoE has the potential to further improve forecasting performance by incorporating wavelet-domain corpora.

Video-to-Audio (V2A) generation is essential for immersive multimedia experiences, yet its evaluation remains underexplored. Existing benchmarks typically assess diverse audio types under a unified protocol, overlooking the fine-grained requirements of distinct audio categories. To address this gap, we propose VidAudio-Bench, a multi-task benchmark for V2A evaluation with four key features: (1) Broad Coverage: It encompasses four representative audio categories - sound effects, music, speech, and singing - under both V2A and Video-Text-to-Audio (VT2A) settings. (2) Extensive Evaluation: It comprises 1,634 video-text pairs and benchmarks 11 state-of-the-art generation models. (3) Comprehensive Metrics: It introduces 13 task-specific, reference-free metrics to systematically assess audio quality, video-audio consistency, and text-audio consistency. (4) Human Alignment: It validates all metrics through subjective studies, demonstrating strong consistency with human preferences. Experimental results reveal that current V2A models perform poorly in speech and singing compared to sound effects. Our VT2A results further highlight a fundamental tension between instruction following and visually grounded generation: stronger visual conditioning improves video-audio alignment, but often at the cost of generating the intended audio category. These findings establish VidAudio-Bench as a comprehensive and scalable framework for diagnosing V2A systems and provide new insights into multimodal audio generation.

Key-Value (KV) cache plays a crucial role in accelerating inference in large language models (LLMs) by storing intermediate attention states and avoiding redundant computation during autoregressive generation. However, its memory footprint scales linearly with sequence length, often leading to severe memory bottlenecks on resource-constrained hardware. Prior work has explored offloading KV cache to the CPU while retaining only a subset on the GPU, but these approaches often rely on imprecise token selection and suffer performance degradation in long-generation tasks such as chain-of-thought reasoning. In this paper, we propose a novel KV cache management strategy, IceCache, which integrates semantic token clustering with PagedAttention. By organizing semantically related tokens into contiguous memory regions managed by a hierarchical, dynamically updatable data structure, our method enables more efficient token selection and better utilization of memory bandwidth during CPU-GPU transfers. Experimental results on LongBench show that, with a 256-token budget, IceCache maintains 99% of the original accuracy achieved by the full KV cache model. Moreover, compared to other offloading-based methods, IceCache attains competitive or even superior latency and accuracy while using only 25% of the KV cache token budget, demonstrating its effectiveness in long-sequence scenarios. The code is available on our project website at https://yuzhenmao.github.io/IceCache/.

Incorporating large language models (LLMs) in medical question answering demands more than high average accuracy: a model that returns substantively different answers each time it is queried is not a reliable medical tool. Online health communities such as Reddit have become a primary source of medical information for millions of users, yet they remain highly susceptible to misinformation; deploying LLMs as assistants in these settings amplifies the need for output consistency alongside correctness. We present a practical, open-source evaluation framework for assessing small, locally-deployable open-weight LLMs on medical question answering, treating reproducibility as a first-class metric alongside lexical and semantic accuracy. Our pipeline computes eight quality metrics, including BERTScore, ROUGE-L, and an LLM-as-judge rubric, together with two within-model reproducibility metrics derived from repeated inference (N=10 runs per question). Evaluating three models (Llama 3.1 8B, Gemma 3 12B, MedGemma 1.5 4B) on 50 MedQuAD questions (N=1,500 total responses) reveals that despite low-temperature generation (T=0.2), self-agreement across runs reaches at most 0.20, while 87-97% of all outputs per model are unique -- a safety gap that single-pass benchmarks entirely miss. The clinically fine-tuned MedGemma 1.5 4B underperforms the larger general-purpose models on both quality and reproducibility; however, because MedGemma is also the smallest model, this comparison confounds domain fine-tuning with model scale. We describe the methodology in sufficient detail for practitioners to replicate or extend the evaluation for their own model-selection workflows. All code and data pipelines are available at https://github.com/aviad-buskila/llm_medical_reproducibility.

The Model Context Protocol (MCP) is a new and emerging technology that extends the functionality of large language models, improving workflows but also exposing users to a new attack surface. Several studies have highlighted related security flaws, but MCP attack detection remains underexplored. To address this research gap, this study develops and evaluates a range of supervised machine learning approaches, including both traditional and deep-learning models. We evaluated the systems on the detection of malicious MCP tool descriptions in two scenarios: (1) a binary classification task distinguishing malicious from benign tools, and (2) a multiclass classification task identifying the attack type while separating benign from malicious tools. In addition to the machine learning models, we compared a rule-based approach that serves as a baseline. The results indicate that several of the developed models achieved 100\% F1-score on the binary classification task. In the multiclass scenario, the SVC and BERT models performed best, achieving F1 scores of 90.56\% and 88.33\%, respectively. Confusion matrices were also used to visualize the full distribution of predictions often missed by traditional metrics, providing additional insight for selecting the best-fitting solution in real-world scenarios. This study presents an addition to the MCP defence area, showing that machine learning models can perform exceptionally well in separating malicious and benign data points. To apply the solution in a live environment, a middleware was developed to classify which MCP tools are safe to use before execution, and block the ones that are not safe. Furthermore, the study shows that these models can outperform traditional rule-based solutions currently in use in the field.

Conventional Vision-and-Language Navigation (VLN) benchmarks assume instructions are feasible and the referenced target exists, leaving agents ill-equipped to handle false-premise goals. We introduce VLN-NF, a benchmark with false-premise instructions where the target is absent from the specified room and agents must navigate, gather evidence through in-room exploration, and explicitly output NOT-FOUND. VLN-NF is constructed via a scalable pipeline that rewrites VLN instructions using an LLM and verifies target absence with a VLM, producing plausible yet factually incorrect goals. We further propose REV-SPL to jointly evaluate room reaching, exploration coverage, and decision correctness. To address this challenge, we present ROAM, a two-stage hybrid that combines supervised room-level navigation with LLM/VLM-driven in-room exploration guided by a free-space clearance prior. ROAM achieves the best REV-SPL among compared methods, while baselines often under-explore and terminate prematurely under unreliable instructions. VLN-NF project page can be found at https://vln-nf.github.io/.

Peptide therapeutics are widely regarded as the "third generation" of drugs, yet progress in peptide Machine Learning (ML) are hindered by the absence of standardized benchmarks. Here we present PepBenchmark, which unifies datasets, preprocessing, and evaluation protocols for peptide drug discovery. PepBenchmark comprises three components: (1) PepBenchData, a well-curated collection comprising 29 canonical-peptide and 6 non-canonical-peptide datasets across 7 groups, systematically covering key aspects of peptide drug development, representing, to the best of our knowledge, the most comprehensive AI-ready dataset resource to date; (2) PepBenchPipeline, a standardized preprocessing pipeline that ensures consistent dataset cleaning, construction, splitting, and feature transformation, mitigating quality issues common in ad hoc pipelines; and (3) PepBenchLeaderboard, a unified evaluation protocol and leaderboard with strong baselines across 4 major methodological families: Fingerprint-based, GNN-based, PLM-based, and SMILES-based models. Together, PepBenchmark provides the first standardized and comparable foundation for peptide drug discovery, facilitating methodological advances and translation into real-world applications. The data and code are publicly available at https://github.com/ZGCI-AI4S-Pep/PepBenchmark/.

How software developers interact with Artificial Intelligence (AI)-powered tools, including Large Language Models (LLMs), plays a vital role in how these AI-powered tools impact them. While overreliance on AI may lead to long-term negative consequences (e.g., atrophy of critical thinking skills); underreliance might deprive software developers of potential gains in productivity and quality. Based on twenty-two interviews with software developers on using LLMs for software development, we propose a preliminary reliance-control framework where the level of control can be used as a way to identify AI overreliance and underreliance. We also use it to recommend future research to further explore the different control levels supported by the current and emergent LLM-driven tools. Our paper contributes to the emerging discourse on AI overreliance and provides an understanding of the appropriate degree of reliance as essential to developers making the most of these powerful technologies. Our findings can help practitioners, educators, and policymakers promote responsible and effective use of AI tools.

We develop a high-precision classifier to measure artificial intelligence (AI) patents by fine-tuning PatentSBERTa on manually labeled data from the USPTO's AI Patent Dataset. Our classifier substantially improves the existing USPTO approach, achieving 97.0% precision, 91.3% recall, and a 94.0% F1 score, and it generalizes well to Chinese patents based on citation and lexical validation. Applying it to granted U.S. patents (1976-2023) and Chinese patents (2010-2023), we document rapid growth in AI patenting in both countries and broad convergence in AI patenting intensity and subfield composition, even as China surpasses the United States in recent annual patent counts. The organization of AI innovation nevertheless differs sharply: U.S. AI patenting is concentrated among large private incumbents and established hubs, whereas Chinese AI patenting is more geographically diffuse and institutionally diverse, with larger roles for universities and state-owned enterprises. For listed firms, AI patents command a robust market-value premium in both countries. Cross-border citations show continued technological interdependence rather than decoupling, with Chinese AI inventors relying more heavily on U.S. frontier knowledge than vice versa.

Referring multi-object tracking (RMOT) is a task of associating all the objects in a video that semantically match with given textual queries or referring expressions. Existing RMOT approaches decompose object grounding and tracking into separated modules and exhibit limited performance due to the scarcity of training videos, ambiguous annotations, and restricted domains. In this work, we introduce STORM, an end-to-end MLLM that jointly performs grounding and tracking within a unified framework, eliminating external detectors and enabling coherent reasoning over appearance, motion, and language. To improve data efficiency, we propose a task-composition learning (TCL) strategy that decomposes RMOT into image grounding and object tracking, allowing STORM to leverage data-rich sub-tasks and learn structured spatial--temporal reasoning. We further construct STORM-Bench, a new RMOT dataset with accurate trajectories and diverse, unambiguous referring expressions generated through a bottom-up annotation pipeline. Extensive experiments show that STORM achieves state-of-the-art performance on image grounding, single-object tracking, and RMOT benchmarks, demonstrating strong generalization and robust spatial--temporal grounding in complex real-world scenarios. STORM-Bench is released at https://github.com/amazon-science/storm-referring-multi-object-grounding.

As Large Language Models (LLMs) have become capable of generating long and descriptive code summaries, accurate and reliable evaluation of factual consistency has become a critical challenge. However, previous evaluation methods are primarily designed for short summaries of isolated code snippets. Consequently, they struggle to provide fine-grained evaluation of multi-sentence functionalities and fail to accurately assess dependency context commonly found in real-world code summaries. To address this, we propose ReFEree, a reference-free and fine-grained method for evaluating factual consistency in real-world code summaries. We define factual inconsistency criteria specific to code summaries and evaluate them at the segment level using these criteria along with dependency information. These segment-level results are then aggregated into a fine-grained score. We construct a code summarization benchmark with human-annotated factual consistency labels. The evaluation results demonstrate that ReFEree achieves the highest correlation with human judgment among 13 baselines, improving 15-18% over the previous state-of-the-art. Our code and data are available at https://github.com/bsy99615/ReFEree.git.

Modern vision-language models achieve strong performance in static perception, but remain limited in the complex spatiotemporal reasoning required for embodied, egocentric tasks. A major source of failure is their reliance on temporal priors learned from passive video data, which often leads to spatiotemporal hallucinations and poor generalization in dynamic environments. To address this, we present EgoTSR, a curriculum-based framework for learning task-oriented spatiotemporal reasoning. EgoTSR is built on the premise that embodied reasoning should evolve from explicit spatial understanding to internalized task-state assessment and finally to long-horizon planning. To support this paradigm, we construct EgoTSR-Data, a large-scale dataset comprising 46 million samples organized into three stages: Chain-of-Thought (CoT) supervision, weakly supervised tagging, and long-horizon sequences. Extensive experiments demonstrate that EgoTSR effectively eliminates chronological biases, achieving 92.4% accuracy on long-horizon logical reasoning tasks while maintaining high fine-grained perceptual precision, significantly outperforming existing open-source and closed-source state-of-the-art models.

Selecting the right knowledge is critical when using large language models (LLMs) to solve domain-specific data analysis tasks. However, most retrieval-augmented approaches rely primarily on lexical or embedding similarity, which is often a weak proxy for the task-critical knowledge needed for multi-step reasoning. In many such tasks, the relevant knowledge is not merely textually related to the query, but is instead grounded in executable code and the dependency structure through which computations are carried out. To address this mismatch, we propose SGKR (Structure-Grounded Knowledge Retrieval), a retrieval framework that organizes domain knowledge with a graph induced by function-call dependencies. Given a question, SGKR extracts semantic input and output tags, identifies dependency paths connecting them, and constructs a task-relevant subgraph. The associated knowledge and corresponding function implementations are then assembled as a structured context for LLM-based code generation. Experiments on multi-step data analysis benchmarks show that SGKR consistently improves solution correctness over no-retrieval and similarity-based retrieval baselines for both vanilla LLMs and coding agents.

Surgical phase segmentation is central to computer-assisted surgery, yet robust models remain difficult to develop when labeled surgical videos are scarce. We study data-efficient phase segmentation for manual small-incision cataract surgery (SICS) through a controlled comparison of visual representations. To isolate representation quality, we pair each visual encoder with the same temporal model (MS-TCN++) under identical training and evaluation settings on SICS-155 (19 phases). We compare supervised encoders (ResNet-50, I3D) against large self-supervised foundation models (DINOv3, V-JEPA2), and use a cached-feature pipeline that decouples expensive visual encoding from lightweight temporal learning. Foundation-model features improve segmentation performance in this setup, with DINOv3 ViT-7B achieving the best overall results (83.4% accuracy, 87.0 edit score). We further examine cataract-domain transfer using unlabeled videos and lightweight adaptation, and analyze when it helps or hurts. Overall, the study indicates strong transferability of modern vision foundation models to surgical workflow understanding and provides practical guidance for low-label medical video settings. The project website is available at: https://sl2005.github.io/DataEfficient-sics-phase-seg/

AI agent development relies heavily on natural language prompting to define agents' tasks, knowledge, and goals. These prompts are interpreted by Large Language Models (LLMs), which govern agent behavior. Consequently, agentic performance is susceptible to variability arising from imprecise or ambiguous prompt formulations. Identifying and correcting such issues requires examining not only the agent's code, but also the internal system prompts generated throughout its execution lifecycle, as reflected in execution logs. In this work, we introduce an analytics pipeline implemented as part of the Agent Mentor open-source library that monitors and incrementally adapts the system prompts defining another agent's behavior. The pipeline improves performance by systematically injecting corrective instructions into the agent's knowledge. We describe its underlying mechanism, with particular emphasis on identifying semantic features associated with undesired behaviors and using them to derive corrective statements. We evaluate the proposed pipeline across three exemplar agent configurations and benchmark tasks using repeated execution runs to assess effectiveness. These experiments provide an initial exploration of automating such a mentoring pipeline within future agentic governance frameworks. Overall, the approach demonstrates consistent and measurable accuracy improvements across diverse configurations, particularly in settings dominated by specification ambiguity. For reproducibility, we released our code as open source under the Agent Mentor library.

Large language models (LLMs) are increasingly used for causal and counterfactual reasoning, yet their reliability in real-world policy evaluation remains underexplored. We construct a benchmark of 40 empirical policy evaluation cases drawn from economics and social science, each grounded in peer-reviewed evidence and classified by intuitiveness -- whether the empirical finding aligns with (obvious), is unclear relative to (ambiguous), or contradicts (counter-intuitive) common prior expectations. We evaluate four frontier LLMs across five prompting strategies with 2,400 experimental trials and analyze the results using mixed-effects logistic regression. Our findings reveal three key results: (1) a chain-of-thought (CoT) paradox, where chain-of-thought prompting dramatically improves performance on obvious cases but this benefit is nearly eliminated on counter-intuitive ones (interaction OR = 0.053, $p < 0.001$); (2) intuitiveness as the dominant factor, explaining more variance than model choice or prompting strategy (ICC = 0.537); and (3) a knowledge-reasoning dissociation, where citation-based familiarity is unrelated to accuracy ($p = 0.53$), suggesting models possess relevant knowledge but fail to reason with it when findings contradict intuition. We frame these results through the lens of dual-process theory (System 1 vs. System 2) and argue that current LLMs' "slow thinking" may be little more than "slow talking" -- they produce the form of deliberative reasoning without the substance.

Large language models frequently fail to produce correct code on their first attempt, yet most benchmarks evaluate them in a single-shot setting. We investigate iterative self-repair (feeding execution errors back to the model for correction) across seven models spanning three families and both open-weight and proprietary providers: Llama 3.1 8B, Llama 3.3 70B, Llama 4 Scout (MoE, 16 experts), Llama 4 Maverick (MoE, 128 experts), Qwen3 32B, Gemini 2.5 Flash, and Gemini 2.5 Pro. On HumanEval (164 problems) and MBPP Sanitized (257 problems) with up to five attempts, self-repair universally improves pass rates: +4.9 to +17.1 pp on HumanEval and +16.0 to +30.0 pp on MBPP. Gemini 2.5 Flash achieves the highest final pass rates (96.3% HumanEval, 93.8% MBPP). Most gains concentrate in the first two rounds.Error-type analysis shows assertion errors (logical mistakes) are the hardest to repair at ~45%, while syntax and name errors are repaired at substantially higher rates, connecting to broader findings on the limits of LLM self-correction. Prior work found that weaker models fail at self-repair or require fine-tuning; we show that modern instruction-tuned models succeed with prompting alone, even at 8B scale. We also provide the first comparison of dense and MoE architectures for self-repair, and extend the repair-vs-resampling tradeoff analysis to modern models. A prompt ablation reveals chain-of-thought repair yields up to +5.5 pp additional self-repair gain (measured as improvement in repair delta) over minimal prompting for capable models.

Psychological client simulators have emerged as a scalable solution for training and evaluating counselor trainees and psychological LLMs. Yet existing simulators exhibit unrealistic over-compliance, leaving counselors underprepared for the challenging behaviors common in real-world practice. To bridge this gap, we present ResistClient, which systematically models challenging client behaviors grounded in Client Resistance Theory by integrating external behaviors with underlying motivational mechanisms. To this end, we propose Resistance-Informed Motivation Reasoning (RIMR), a two-stage training framework. First, RIMR mitigates compliance bias via supervised fine-tuning on RPC, a large-scale resistance-oriented psychological conversation dataset covering diverse client profiles. Second, beyond surface-level response imitation, RIMR models psychologically coherent motivation reasoning before response generation, jointly optimizing motivation authenticity and response consistency via process-supervised reinforcement learning. Extensive automatic and expert evaluations show that ResistClient substantially outperforms existing simulators in challenge fidelity, behavioral plausibility, and reasoning coherence. Moreover, ResistClient facilities evaluation of psychological LLMs under challenging conditions, offering new optimization directions for mental health dialogue systems.

Vision-Language Models (VLMs) have made significant strides in static image understanding but continue to face critical hurdles in spatiotemporal reasoning. A major bottleneck is "multi-image reasoning hallucination", where a massive performance drop between forward and reverse temporal queries reveals a dependence on superficial shortcuts instead of genuine causal understanding. To mitigate this, we first develop a new Chain-of-Thought (CoT) dataset that decomposes intricate reasoning into detailed spatiotemporal steps and definitive judgments. Building on this, we present a progressive training framework: it initiates with supervised pre-training on our CoT dataset to instill logical structures, followed by fine-tuning with scalable weakly-labeled data for broader generalization. Our experiments demonstrate that this approach not only improves backbone accuracy but also slashes the forward-backward performance gap from over 70\% to only 6.53\%. This confirms the method's ability to develop authentic dynamic reasoning and reduce the inherent temporal biases of current VLMs.

Agent based systems are more common than we may think. A Promise Theory perspective on cooperation, in systems of human-machine agents, offers a unified perspective on organization and functional design with semi-automated efforts, in terms of the abstract properties of autonomous agents, This applies to human efforts, hardware systems, software, and artificial intelligence, with and without management. One may ask how does a reasoning system of components keep to an intended purpose? As the agent paradigm is now being revived, in connection with artificial intelligence agents, I revisit established principles of agent cooperation, as applied to humans, machines, and their mutual interactions. Promise Theory represents the fundamentals of signalling, comprehension, trust, risk, and feedback between agents, and offers some lessons about success and failure.

Current large language models (LLMs), even those explicitly trained for reasoning, often struggle with ambiguous content moderation cases due to misleading "decision shortcuts" embedded in context. Inspired by cognitive psychology insights into expert moderation, we introduce \caro (Chain-of-Analogy Reasoning Optimization), a novel two-stage training framework to induce robust analogical reasoning in LLMs. First, \caro bootstraps analogical reasoning chains via retrieval-augmented generation (RAG) on moderation data and performs supervised fine-tuning (SFT). Second, we propose a customized direct preference optimization (DPO) approach to reinforce analogical reasoning behaviors explicitly. Unlike static retrieval methods, \caro dynamically generates tailored analogical references during inference, effectively mitigating harmful decision shortcuts. Extensive experiments demonstrate that \caro substantially outperforms state-of-the-art reasoning models (DeepSeek R1, QwQ), specialized moderation models (LLaMA Guard), and advanced fine-tuning and retrieval-augmented methods, achieving an average F1 score improvement of 24.9\% on challenging ambiguous moderation benchmarks.

Modern audio systems universally employ mel-scale representations derived from 1940s Western psychoacoustic studies, potentially encoding cultural biases that create systematic performance disparities. We present a comprehensive evaluation of cross-cultural bias in audio front-ends, comparing mel-scale features with learnable alternatives (LEAF, SincNet) and psychoacoustic variants (ERB, Bark, CQT) across speech recognition (11 languages), music analysis (6 collections), and European acoustic scene classification (10 European cities). Our controlled experiments isolate front-end contributions while holding architecture and training protocols minimal and constant. Results demonstrate that mel-scale features yield 31.2% WER for tonal languages compared to 18.7% for non-tonal languages (12.5% gap), and show 15.7% F1 degradation between Western and non-Western music. Alternative representations significantly reduce these disparities: LEAF reduces the speech gap by 34% through adaptive frequency allocation, CQT achieves 52% reduction in music performance gaps, and ERB-scale filtering cuts disparities by 31% with only 1% computational overhead. We also release FairAudioBench, enabling cross-cultural evaluation, and demonstrate that adaptive frequency decomposition offers practical paths toward equitable audio processing. These findings reveal how foundational signal processing choices propagate bias, providing crucial guidance for developing inclusive audio systems.

Content moderation in online platforms faces persistent challenges due to the evolving complexity of user-generated content and the limitations of traditional rule-based and machine learning approaches. While recent advances in large language models (LLMs) have enabled more sophisticated moderation via direct prompting or fine-tuning, these approaches often exhibit limited generalization, interpretability, and adaptability to unseen or ambiguous cases. In this work, we propose a novel moderation framework that leverages analogical examples to enhance rule induction and decision reliability. Our approach integrates end-to-end optimization of analogical retrieval, rule generation, and moderation classification, enabling the dynamic adaptation of moderation rules to diverse content scenarios. Through comprehensive experiments, we demonstrate that our method significantly outperforms both rule-injected fine-tuning baselines and multi-stage static RAG pipelines in terms of moderation accuracy and rule quality. Further evaluations, including human assessments and external model generalization tests, confirm that our framework produces rules with better clarity, interpretability, and applicability. These findings show that analogical example-driven methods can advance robust, explainable, and generalizable content moderation in real-world applications.

Federated learning has become a popular paradigm for privacy protection and edge-based machine learning. However, defending against differential attacks and devising incentive strategies remain significant bottlenecks in this field. Despite recent works on privacy-aware incentive mechanism design for federated learning, few of them consider both data volume and noise level. In this paper, we propose a novel federated learning system called FEDBUD, which combines privacy and economic concerns together by considering the joint influence of data volume and noise level on incentive strategy determination. In this system, the cloud server controls monetary payments to edge nodes, while edge nodes control data volume and noise level that potentially impact the model performance of the cloud server. To determine the mutually optimal strategies for both sides, we model FEDBUD as a two-stage Stackelberg Game and derive the Nash Equilibrium using the mean-field estimator and virtual queue. Experimental results on real-world datasets demonstrate the outstanding performance of FEDBUD.

Although optimization is one of the most promising applications of quantum computers, the development of effective optimization strategies requires real-world test cases. When planning our recent wedding reception, we realized that the problem of optimally seating our guests, given constraints related to guests' relatedness, shared interests, and physical needs, could be mapped to a cost function network (CFN) form solvable with classical or quantum optimization algorithms. We compared the seating optimization performance of classical Monte Carlo CFN solvers in the Masala software suite to that of quantum annealing-based CFN optimization algorithms using one-hot, domain-wall, and approximate binary mappings, which we had developed for protein design problems. Surprisingly, the D-Wave Advantage 2 system, which performs well on similarly-structured CFN problems for protein design, struggled to return optimal seating arrangements that were easily found by classical Monte Carlo methods. We provide our seating optimization benchmark set, and code to convert seating optimization problems to CFN problems, as a plugin library for Masala, permitting this class of real-world problems to be used to benchmark performance of current and future classical CFN solvers, quantum optimization algorithms, and quantum computing hardware.

Outliers have emerged as a fundamental bottleneck in preserving accuracy for low-precision large models, particularly within Mixture-of-Experts (MoE) architectures that are increasingly central to large-scale language modeling. Under post-training quantization (PTQ), these outliers induce substantial quantization errors, leading to severe accuracy degradation. While recent rotation-based smoothing techniques alleviate the problem by redistributing outlier magnitudes, residual errors remain and continue to impede reliable low-precision deployment. In this work, we tackle this challenge by introducing \textit{CodeQuant}, a unified quantization-and-clustering scheme that contains smoothing activation outliers via learnable rotation and absorbing weight outliers into fine-tuned cluster centroids for MoE. This design reduces the influence of extreme values by fitting them within cluster centroids, thereby lowering quantization error while maintaining expressive capacity. Coupled with a dedicated kernel design for GPU and CPU, CodeQuant achieves up to $4.15\times$ speedup while delivering significantly higher accuracy than state-of-the-art quantization approaches across diverse MoE models. Our results highlight CodeQuant as a promising direction for efficient and accurate deployment of MoE-based large language models under low-precision constraints. Our code is available at https://github.com/SAI-Lab-NYU/CodeQuant.

As Large Language Models (LLMs) are increasingly deployed in real-world applications, reliable uncertainty quantification (UQ) becomes critical for safe and effective use. Most existing UQ approaches for language models aim to produce a single confidence score -- for example, estimating the probability that a model's answer is correct. However, uncertainty in natural language tasks arises from multiple distinct sources, including model knowledge gaps, output variability, and input ambiguity, which have different implications for system behavior and user interaction. In this work, we study how the source of uncertainty impacts the behavior and effectiveness of existing UQ methods. To enable controlled analysis, we introduce a new dataset that explicitly categorizes uncertainty sources, allowing systematic evaluation of UQ performance under each condition. Our experiments reveal that while many UQ methods perform well when uncertainty stems solely from model knowledge limitations, their performance degrades or becomes misleading when other sources are introduced. These findings highlight the need for uncertainty-aware methods that explicitly account for the source of uncertainty in large language models.

Block Floating-Point (BFP) is emerging as an attractive data format for edge Neural Processing Units (NPUs), combining wide dynamic range with high hardware efficiency. However, its behavior under hardware faults and suitability for safety-critical deployments remain underexplored. Here, we present the first in-depth empirical reliability study of BFP-based NPUs. Using RTL-level fault injection on NPUs, our bit- and path-level analysis reveals pronounced heterogeneous vulnerabilities and shows conventional end-to-end check becomes ineffective under nonlinear block scaling. Guided by these insights, we design a fault-tolerant BFP-based NPU microarchitecture that aligns the BFP computational semantics with reliability constraints. The design uses a row/column-wise blocking strategy to decouple the fixed-point mantissa computations from the scalar exponent path, and introduces ultra-lightweight protection mechanisms for each. Experimental results demonstrate our design achieves near-dual modular redundancy reliability with only $3.55\%$ geometric mean performance overhead and less than $2\%$ hardware cost.

Automating real-world software engineering tasks remains challenging for large language model (LLM)-based agents due to the need for long-horizon reasoning over large, evolving codebases and making consistent decisions across interdependent actions. Existing approaches typically rely on static prompting strategies or handcrafted heuristics to select actions such as code editing, file navigation, and test execution, but they lack fine-grained feedback on intermediate decisions. This leads to inefficient exploration, error propagation, and brittle solution trajectories. To address this limitation, we propose SWE-Shepherd, a framework that introduces Process Reward Models (PRMs) to provide dense, step-level supervision for repository-level code agents. Using trajectories from SWE-Bench, we construct an action-level reward dataset and train a lightweight reward model on a base LLM to estimate the usefulness of intermediate actions. During inference, the PRM evaluates candidate actions and guides the agent toward higher-reward decisions without requiring full reinforcement learning. Experiments on SWE-Bench Verified demonstrate improved interaction efficiency and action quality, while also highlighting challenges in aligning intermediate rewards with final task success.

Low-visibility scenarios, such as low-light conditions, pose significant challenges to human pose estimation due to the scarcity of annotated low-light datasets and the loss of visual information under poor illumination. Recent domain adaptation techniques attempt to utilize well-lit labels by augmenting well-lit images to mimic low-light conditions. But handcrafted augmentations oversimplify noise patterns, while learning-based methods often fail to preserve high-frequency low-light characteristics, producing unrealistic images that lead pose models to generalize poorly to real low-light scenes. Moreover, recent pose estimators rely on image cues through image-to-keypoint cross-attention, but these cues become unreliable under low-light conditions. To address these issues, we propose Unsupervised Domain Adaptation for Pose Estimation (UDAPose), a novel framework that synthesizes low-light images and dynamically fuses visual cues with pose priors for improved pose estimation. Specifically, our synthesis method incorporates a Direct-Current-based High-Pass Filter (DHF) and a Low-light Characteristics Injection Module (LCIM) to inject high-frequency details from input low-light images, overcoming rigidity or the detail loss in existing approaches. Furthermore, we introduce a Dynamic Control of Attention (DCA) module that adaptively balances image cues with learned pose priors in the Transformer architecture. Experiments show that UDAPose outperforms state-of-the-art methods, with notable AP gains of 10.1 (56.4%) on the ExLPose-test hard set (LL-H) and 7.4 (31.4%) in cross-dataset validation on EHPT-XC. Code: https://github.com/Vision-and-Multimodal-Intelligence-Lab/UDAPose

DNNs and LLMs increasingly rely on hardware accelerators, including in safety-critical domains, while technology scaling and growing model complexity make hardware faults more frequent. Existing system-level mechanisms typically treat the NPU as a monolithic unit, using coarse-grained replication that incurs prohibitive performance and hardware overheads, leaving a gap between reliability requirements and deployable solutions. To bridge this gap, we present Strix, a full-stack NPU reliability framework on an open-source SoC, spanning micro-architecture, ISA, and programming methods. Strix re-partitions the NPU along the system inference pipeline, identifies dominant failure modes, and attaches targeted safeguards, achieving sub-micro-second fault localisation, error detection, and correction with only 1.04$\times$ slowdown and minimal hardware overhead.

Retrieving the correct set of files from a large codebase is a crucial step in Automated Program Repair (APR). High recall is necessary to ensure that the relevant files are included, but simply increasing the number of retrieved files introduces noise and degrades efficiency. To address this tradeoff, we propose PatchRecall, a hybrid retrieval approach that balances recall with conciseness. Our method combines two complementary strategies: (1) codebase retrieval, where the current issue description is matched against the codebase to surface potentially relevant files, and (2) history-based retrieval, where similar past issues are leveraged to identify edited files as candidate targets. Candidate files from both strategies are merged and reranked to produce the final retrieval set. Experiments on SWE-Bench demonstrate that PatchRecall achieves higher recall without significantly increasing retrieved file count, enabling more effective APR.

Post-training data plays a pivotal role in shaping the capabilities of Large Language Models (LLMs), yet datasets are often treated as isolated artifacts, overlooking the systemic connections that underlie their evolution. To disentangle these complex relationships, we introduce the concept of \textbf{data lineage} to the LLM ecosystem and propose an automated multi-agent framework to reconstruct the evolutionary graph of dataset development. Through large-scale lineage analysis, we characterize domain-specific structural patterns, such as vertical refinement in math-oriented datasets and horizontal aggregation in general-domain corpora. Moreover, we uncover pervasive systemic issues, including \textit{structural redundancy} induced by implicit dataset intersections and the \textit{propagation of benchmark contamination} along lineage paths. To demonstrate the practical value of lineage analysis for data construction, we leverage the reconstructed lineage graph to create a \textit{lineage-aware diversity-oriented dataset}. By anchoring instruction sampling at upstream root sources, this approach mitigates downstream homogenization and hidden redundancy, yielding a more diverse post-training corpus. We further highlight lineage-centric analysis as an efficient and robust topological alternative to sample-level dataset comparison for large-scale data ecosystems. By grounding data construction in explicit lineage structures, our work advances post-training data curation toward a more systematic and controllable paradigm.

Modeling household-level trip generation is fundamental to accurate demand forecasting, traffic flow estimation, and urban system planning. Existing studies were mostly based on classical machine learning models with limited predictive capability, while recent LLM-based approaches have yet to incorporate behavioral theory or intra-household interaction dynamics, both of which are critical for modeling realistic collective travel decisions. To address these limitations, we propose a novel LLM-based framework, named Persona-Enriched Multi-Agent Negotiation for Travel (PEMANT), which first integrates behavioral theory for individualized persona modeling and then conducts household-level trip planning negotiations via a structured multi-agent conversation. Specifically, PEMANT transforms static sociodemographic attributes into coherent narrative profiles that explicitly encode household-level attitudes, subjective norms, and perceived behavioral controls, following our proposed Household-Aware Chain-of-Planned-Behavior (HA-CoPB) framework. Building on these theory-grounded personas, PEMANT captures real-world household decision negotiation via a structured two-phase multi-agent conversation framework with a novel persona-alignment control mechanism. Evaluated on both national and regional household travel survey datasets, PEMANT consistently outperforms state-of-the-art benchmarks across datasets.

Legal consultation question answering (Legal CQA) presents unique challenges compared to traditional legal QA tasks, including the scarcity of high-quality training data, complex task composition, and strong contextual dependencies. To address these, we construct JurisCQAD, a large-scale dataset of over 43,000 real-world Chinese legal queries annotated with expert-validated positive and negative responses, and design a structured task decomposition that converts each query into a legal element graph integrating entities, events, intents, and legal issues. We further propose JurisMA, a modular multi-agent framework supporting dynamic routing, statutory grounding, and stylistic optimization. Combined with the element graph, the framework enables strong context-aware reasoning, effectively capturing dependencies across legal facts, norms, and procedural logic. Trained on JurisCQAD and evaluated on a refined LawBench, our system significantly outperforms both general-purpose and legal-domain LLMs across multiple lexical and semantic metrics, demonstrating the benefits of interpretable decomposition and modular collaboration in Legal CQA.

Standard resampling ratios (e.g., $α\approx 0.632$) are widely used as default baselines in ensemble learning for three decades. However, how these ratios interact with a base learner's intrinsic functional complexity in finite samples lacks a exact mathematical characterization. We leverage the Hoeffding-ANOVA decomposition to derive the first exact, finite-sample variance decomposition for subagging, applicable to any symmetric base learner without requiring asymptotic limits or smoothness assumptions. We establish that subagging operates as a deterministic low-pass spectral filter: it preserves low-order structural signals while attenuating $c$-th order interaction variance by a geometric factor approaching $α^c$. This decoupling reveals why default baselines often under-regularize high-capacity interpolators, which instead require smaller $α$ to exponentially suppress spurious high-order noise. To operationalize these insights, we propose a complexity-guided adaptive subsampling algorithm, empirically demonstrating that dynamically calibrating $α$ to the learner's complexity spectrum consistently improves generalization over static baselines.

Diffusion models are often introduced from multiple perspectives, such as VAEs, score matching, or flow matching, accompanied by dense and technically demanding mathematics that can be difficult for beginners to grasp. One classic question is: how does the reverse process invert the forward process to generate data from pure noise? This article systematically organizes the diffusion model from a fresh Langevin perspective, offering a simpler, clearer, and more intuitive answer. We also address the following questions: how can ODE-based and SDE-based diffusion models be unified under a single framework? Why are diffusion models theoretically superior to ordinary VAEs? Why is flow matching not fundamentally simpler than denoising or score matching, but equivalent under maximum-likelihood? We demonstrate that the Langevin perspective offers clear and straightforward answers to these questions, bridging existing interpretations of diffusion models, showing how different formulations can be converted into one another within a common framework, and offering pedagogical value for both learners and experienced researchers seeking deeper intuition.

The rapid growth of generative AI has introduced new challenges in content moderation and digital forensics. In particular, benign AI-generated images can be paired with harmful or misleading text, creating difficult-to-detect misuse. This contextual misuse undermines the traditional moderation framework and complicates attribution, as synthetic images typically lack persistent metadata or device signatures. We introduce a steganography enabled attribution framework that embeds cryptographically signed identifiers into images at creation time and uses multimodal harmful content detection as a trigger for attribution verification. Our system evaluates five watermarking methods across spatial, frequency, and wavelet domains. It also integrates a CLIP-based fusion model for multimodal harmful-content detection. Experiments demonstrate that spread-spectrum watermarking, especially in the wavelet domain, provides strong robustness under blur distortions, and our multimodal fusion detector achieves an AUC-ROC of 0.99, enabling reliable cross-modal attribution verification. These components form an end-to-end forensic pipeline that enables reliable tracing of harmful deployments of AI-generated imagery, supporting accountability in modern synthetic media environments. Our code is available at GitHub: https://github.com/bli1/steganography

The exponential growth of user-generated movie reviews on digital platforms has made accurate text sentiment classification a cornerstone task in natural language processing. Traditional models, including standard BERT and recurrent architectures, frequently struggle to capture long-distance semantic dependencies and resolve ambiguous emotional expressions in lengthy review texts. This paper proposes a novel hybrid framework that seamlessly integrates dynamic adaptive multi-head attention with supervised contrastive learning into a BERT-based Transformer encoder. The dynamic adaptive attention module employs a global context pooling vector to dynamically regulate the contribution of each attention head, thereby focusing on critical sentiment-bearing tokens while suppressing noise. Simultaneously, the supervised contrastive learning branch enforces tighter intra-class compactness and larger inter-class separation in the embedding space. Extensive experiments on the IMDB dataset demonstrate that the proposed model achieves competitive performance with an accuracy of 94.67\%, outperforming strong baselines by 1.5--2.5 percentage points. The framework is lightweight, efficient, and readily extensible to other text classification tasks.

Wearable IMU-based Human Activity Recognition (HAR) relies heavily on Deep Neural Networks (DNNs), which are burdened by immense computational and buffering demands. Their power-hungry floating-point operations and rigid requirement to process complete temporal windows severely cripple battery-constrained edge devices. While Spiking Neural Networks (SNNs) offer extreme event-driven energy efficiency, standard architectures struggle with complex biomechanical topologies and temporal gradient degradation. To bridge this gap, we propose the Physics-Aware Spiking Neural Network (PAS-Net), a fully multiplier-free architecture explicitly tailored for Green HAR. Spatially, an adaptive symmetric topology mixer enforces human-joint physical constraints. Temporally, an $O(1)$-memory causal neuromodulator yields context-aware dynamic threshold neurons, adapting actively to non-stationary movement rhythms. Furthermore, we leverage a temporal spike error objective to unlock a flexible early-exit mechanism for continuous IMU streams. Evaluated across seven diverse datasets, PAS-Net achieves state-of-the-art accuracy while replacing dense operations with sparse 0.1 pJ integer accumulations. Crucially, its confidence-driven early-exit capability drastically reduces dynamic energy consumption by up to 98\%. PAS-Net establishes a robust, ultra-low-power neuromorphic standard for always-on wearable sensing.

Recent advances in large language models (LLMs) have enabled promising progress in diagnosis prediction from electronic health records (EHRs). However, existing LLM-based approaches tend to overfit to historically observed diagnoses, often overlooking novel yet clinically important conditions that are critical for early intervention. To address this, we propose EviCare, an in-context reasoning framework that integrates deep model guidance into LLM-based diagnosis prediction. Rather than prompting LLMs directly with raw EHR inputs, EviCare performs (1) deep model inference for candidate selection, (2) evidential prioritization for set-based EHRs, and (3) relational evidence construction for novel diagnosis prediction. These signals are then composed into an adaptive in-context prompt to guide LLM reasoning in an accurate and interpretable manner. Extensive experiments on two real-world EHR benchmarks (MIMIC-III and MIMIC-IV) demonstrate that EviCare achieves significant performance gains, which consistently outperforms both LLM-only and deep model-only baselines by an average of 20.65\% across precision and accuracy metrics. The improvements are particularly notable in challenging novel diagnosis prediction, yielding average improvements of 30.97\%.

Olfaction lies at the intersection of chemical structure, neural encoding, and linguistic perception, yet existing representation methods fail to fully capture this pathway. Current approaches typically model only isolated segments of the olfactory pathway, overlooking the complete chain from molecule to receptors to linguistic descriptions. Such fragmentation yields learned embeddings that lack both biological grounding and semantic interpretability. We propose NOSE (Neural Olfactory-Semantic Embedding), a representation learning framework that aligns three modalities along the olfactory pathway: molecular structure, receptor sequence, and natural language description. Rather than simply fusing these signals, we decouple their contributions via orthogonal constraints, preserving the unique encoded information of each modality. To address the sparsity of olfactory language, we introduce a weak positive sample strategy to calibrate semantic similarity, preventing erroneous repulsion of similar odors in the feature space. Extensive experiments demonstrate that NOSE achieves state-of-the-art (SOTA) performance and excellent zero-shot generalization, confirming the strong alignment between its representation space and human olfactory intuition.

Test optimization contains test case selection and minimization, which is an important challenge in software testing and has been addressed with search-based approaches intensively in the past. Inspired by the recent advancement of using quantum optimization solutions for addressing test optimization problems, we looked into Coherent Ising Machines (CIM), which offer potential for solving combinatorial optimization problems, but have not yet been exploited in test optimization. Hence, in this paper, we present IsingTester, an open-source, Python-based command-line tool that provides an end-to-end pipeline for solving test optimization problems that are formulated as Ising models. With IsingTester, we reformulate test selection and minimization as Ising spin configurations, encode multiple optimization strategies into Ising Hamiltonians, and implement solvers including CIM simulation and brute-force search. Given a user-provided dataset and solver configuration, IsingTester automatically performs problem encoding, optimization, and spin decoding, returning selected test cases back to the user. Along with IsingTester, we also present the accompanying IsingBench for evaluating and comparing optimization techniques across Ising-based paradigms against baseline approaches. A screencast demonstrating the tool is available at: https://github.com/WSE-Lab/IsingBench.

Recent advancements in Large Language Models (LLMs) have successfully employed search-based strategies to enhance code generation. However, existing methods typically rely on static, sparse public test cases for verification, leading to pseudo-correctness -- where solutions overfit the visible public tests but fail to generalize to hidden test cases. We argue that optimizing against a fixed, weak environment inherently limits robustness. To address this, we propose AdverMCTS, a novel adversarial Monte Carlo Tree Search framework that combats pseudo-correctness by coupling code search with active vulnerability discovery. AdverMCTS formulates generation as a minimax-style game between a Solver agent, which synthesizes code candidates, and an Attacker agent, which evolves to generate targeted corner test cases that exploit logical divergences in the current code pool. These discovered tests form a dynamic, progressively hostile filter that penalizes fragile reasoning. Extensive experiments demonstrate that AdverMCTS significantly outperforms state-of-the-art baselines, effectively reducing false positive rates and forcing the model to generalize beyond the initial constraints. The resources of this work are available at https://anonymous.4open.science/r/AdverMCTS_open-A255.

Instruction tuning relies on large instruction-response corpora whose quality and composition strongly affect downstream performance. We propose Answer Divergence-Guided Selection (ADG), which selects instruction data based on the geometric structure of multi-sample outputs. ADG draws several high-temperature generations per instruction, maps responses into an embedding space, and computes an output divergence score that jointly encodes dispersion magnitude and shape anisotropy. High scores correspond to instructions whose answers are both far apart and multi-modal, rather than clustered paraphrases along a single direction. Across two backbones and three public instruction pools, fine-tuning on only 10K ADG-selected examples consistently outperforms strong selectors on six benchmarks spanning reasoning, knowledge, and coding. Analyses further show that both dispersion magnitude and shape anisotropy are necessary, supporting answer divergence as a practical signal for instruction data selection. Code and appendix are included in the supplementary materials.

Medical large language models (LLMs) achieve impressive performance on standardized benchmarks, yet these evaluations fail to capture the complexity of real clinical encounters where patients exhibit memory gaps, limited health literacy, anxiety, and other communication barriers. We introduce VeriSim, a truth-preserving patient simulation framework that injects controllable, clinically evidence-grounded noise into patient responses while maintaining strict adherence to medical ground truth through a hybrid UMLS-LLM verification mechanism. Our framework operationalizes six noise dimensions derived from peer-reviewed medical communication literature, capturing authentic clinical phenomena such as patient recall limitations, health literacy barriers, and stigma-driven non-disclosure. Experiments across seven open-weight LLMs reveal that all models degrade significantly under realistic patient noise, with diagnostic accuracy dropping 15-25% and conversation length increasing 34-55%. Notably, smaller models (7B) show 40% greater degradation than larger models (70B+), while medical fine-tuning on standard corpora provides limited robustness benefits against patient communication noise. Evaluation by board-certified clinicians demonstrates high-quality simulation with strong inter-annotator agreement (kappa > 0.80), while LLM-as-a-Judge serves as a validated auxiliary evaluator achieving comparable reliability for scalable assessment. Our results highlight a critical Sim-to-Real gap in current medical AI. We release VeriSim as an open-source noise-injection framework, establishing a rigorous testbed for evaluating clinical robustness.

This paper considers the problem of zero-shot safety guarantees for cascade dynamical systems. These are systems where a subset of the states (the inner states) affects the dynamics of the remaining states (the outer states) but not vice-versa. We define safety as remaining on a set deemed safe for all times with high probability. We propose to train a safe RL policy on a reduced-order model, which ignores the dynamics of the inner states, but it treats it as an action that influences the outer state. Thus, reducing the complexity of the training. When deployed in the full system the trained policy is combined with a low-level controller whose task is to track the reference provided by the RL policy. Our main theoretical contribution is a bound on the safe probability in the full-order system. In particular, we establish the interplay between the probability of remaining safe after the zero-shot deployment and the quality of the tracking of the inner states. We validate our theoretical findings on a quadrotor navigation task, demonstrating that the preservation of the safety guarantees is tied to the bandwidth and tracking capabilities of the low-level controller.

We develop a queueing-theoretic framework to model the temporal evolution of cyber-attack surfaces, where the number of active vulnerabilities is represented as the backlog of a queue. Vulnerabilities arrive as they are discovered or created, and leave the system when they are patched or successfully exploited. Building on this model, we study how automation affects attack and defense dynamics by introducing an AI amplification factor that scales arrival, exploit, and patching rates. Our analysis shows that even symmetric automation can increase the rate of successful exploits. We validate the model using vulnerability data collected from an open source software supply chain and show that it closely matches real-world attack surface dynamics. Empirical results reveal heavy-tailed patching times, which we prove induce long-range dependence in vulnerability backlog and help explain persistent cyber risk. Utilizing our queueing abstraction for the attack surface, we develop a systematic approach for cyber risk mitigation. We formulate the dynamic defense problem as a constrained Markov decision process with resource-budget and switching-cost constraints, and develop a reinforcement learning (RL) algorithm that achieves provably near-optimal regret. Numerical experiments validate the approach and demonstrate that our adaptive RL-based defense policies significantly reduce successful exploits and mitigate heavy-tail queue events. Using trace-driven experiments on the ARVO dataset, we show that the proposed RL-based defense policy reduces the average number of active vulnerabilities in a software supply chain by over 90% compared to existing defense practices, without increasing the overall maintenance budget. Our results allow defenders to quantify cumulative exposure risk under long-range dependent attack dynamics and to design adaptive defense strategies with provable efficiency.

Large Language Models (LLMs) struggle with knowledge-intensive tasks due to hallucinations and fragmented reasoning over dispersed information. While Retrieval-Augmented Generation (RAG) grounds generation in external sources, existing methods often treat evidence as isolated units, failing to reconstruct the logical chains that connect these dots. Inspired by Complementary Learning Systems (CLS), we propose CodaRAG, a framework that evolves retrieval from passive lookup into active associative discovery. CodaRAG operates via a three-stage pipeline: (1) Knowledge Consolidation to unify fragmented extractions into a stable memory substrate; (2) Associative Navigation to traverse the graph via multi-dimensional pathways-semantic, contextualized, and functional-explicitly recovering dispersed evidence chains; and (3) Interference Elimination to prune hyper-associative noise, ensuring a coherent, high-precision reasoning context. On GraphRAG-Bench, CodaRAG achieves absolute gains of 7-10% in retrieval recall and 3-11% in generation accuracy. These results demonstrate CodaRAG's superior ability to systematically robustify associative evidence retrieval for factual, reasoning, and creative tasks.

Foundation-style ECG encoders pretrained with self-supervised learning are increasingly reused across tasks, institutions, and deployment contexts, often through model-as-a-service interfaces that expose scalar scores or latent representations. While such reuse improves data efficiency and generalization, it raises a participation privacy concern: can an adversary infer whether a specific individual or cohort contributed ECG data to pretraining, even when raw waveforms and diagnostic labels are never disclosed? In connected-health settings, training participation itself may reveal institutional affiliation, study enrollment, or sensitive health context. We present an implementation-grounded audit of membership inference attacks (MIAs) against modern self-supervised ECG foundation encoders, covering contrastive objectives (SimCLR, TS2Vec) and masked reconstruction objectives (CNN- and Transformer-based MAE). We evaluate three realistic attacker interfaces: (i) score-only black-box access to scalar outputs, (ii) adaptive learned attackers that aggregate subject-level statistics across repeated queries, and (iii) embedding-access attackers that probe latent representation geometry. Using a subject-centric protocol with window-to-subject aggregation and calibration at fixed false-positive rates under a cross-dataset auditing setting, we observe heterogeneous and objective-dependent participation leakage: leakage is most pronounced in small or institution-specific cohorts and, for contrastive encoders, can saturate in embedding space, while larger and more diverse datasets substantially attenuate operational tail risk. Overall, our results show that restricting access to raw signals or labels is insufficient to guarantee participation privacy, underscoring the need for deployment-aware auditing of reusable biosignal foundation encoders in connected-health systems.

Replicability requires that algorithmic conclusions remain consistent when rerun on independently drawn data. A central structural question is composition: given $k$ problems each admitting a $ρ$-replicable algorithm with sample complexity $n$, how many samples are needed to solve all jointly while preserving replicability? The naive analysis yields $\widetilde{O}(nk^2)$ samples, and Bun et al. (STOC'23) observed that reductions through differential privacy give an alternative $\widetilde{O}(n^2k)$ bound, leaving open whether the optimal $\widetilde{O}(nk)$ scaling is achievable. We resolve this open problem and, more generally, show that problems with sample complexities $n_1,\ldots,n_k$ can be jointly solved with $\widetilde{O}(\sum_i n_i)$ samples while preserving constant replicability. Our approach converts each replicable algorithm into a perfectly generalizing one, composes them via a privacy-style analysis, and maps back via correlated sampling. This yields the first advanced composition theorem for replicability. En route, we obtain new bounds for the composition of perfectly generalizing algorithms with heterogeneous parameters. As part of our results, we provide a boosting theorem for the success probability of replicable algorithms. For a broad class of problems, the failure probability appears as a separate additive term independent of $ρ$, immediately yielding improved sample complexity bounds for several problems. Finally, we prove an $Ω(nk^2)$ lower bound for adaptive composition, establishing a quadratic separation from the non-adaptive setting. The key technique, which we call the phantom run, yields structural results of independent interest.

Large language models (LLMs) enable waveform-to-text ECG interpretation and interactive clinical questioning, yet most ECG-LLM systems still rely on weak signal-text alignment and retrieval without explicit physiological or causal structure. This limits grounding, temporal reasoning, and counterfactual "what-if" analysis central to clinical decision-making. We propose CARE-ECG, a causally structured ECG-language reasoning framework that unifies representation learning, diagnosis, and explanation in a single pipeline. CARE-ECG encodes multi-lead ECGs into temporally organized latent biomarkers, performs causal graph inference for probabilistic diagnosis, and supports counterfactual assessment via structural causal models. To improve faithfulness, CARE-ECG grounds language outputs through causal retrieval-augmented generation and a modular agentic pipeline that integrates history, diagnosis, and response with verification. Across multiple ECG benchmarks and expert QA settings, CARE-ECG improves diagnostic accuracy and explanation faithfulness while reducing hallucinations (e.g., 0.84 accuracy on Expert-ECG-QA and 0.76 on SCP-mapped PTB-XL under GPT-4). Overall, CARE-ECG provides traceable reasoning by exposing key latent drivers, causal evidence paths, and how alternative physiological states would change outcomes.

Urban intersections expose the limitations of single-vehicle perception under occlusion and partial observability. In this study, we present an auditable roadside LiDAR framework for infrastructure-assisted safety analysis at a signalized urban intersection in New York City, developed and evaluated using real-world data. The proposed framework integrates trajectory construction, iterative human-in-the-loop quality assurance (QA), and interpretable near-miss analytics to produce defensible safety evidence from infrastructure sensing. Using a human-labeled heavy vehicle--bicycle interaction as an anchor case, we show that direction-agnostic time-to-collision (TTC) drops below 1s, while longitudinal TTC remains above conservative braking thresholds, revealing a lateral-intrusion-dominated conflict mechanism. Beyond individual cases, continuous-window evaluation and multi-round QA analysis demonstrate that the framework systematically reduces failure modes such as track fragmentation, spurious TTC triggers, unstable geometry, and cross-lane false conflicts. These results position roadside LiDAR as a practical post-hoc auditing mechanism for cooperative perception systems, with broader statistical validation discussed. This work provides a pathway toward scalable, data-driven safety auditing of urban intersections, enabling transportation agencies to identify and mitigate high-risk interactions beyond crash-based analyses.

Alan Turing is considered as a founder of current computer science together with Kurt Godel, Alonzo Church and John von Neumann. In this paper multiple new research results are presented. It is demonstrated that there would not be Alan Turing's achievements without earlier seminal contributions by Georg Cantor in the set theory and foundations of mathematics. It is proposed to introduce the measure of undecidability of problems unsolvable by Turing machines based on probability distribution of its input data, i.e., to provide the degree of unsolvabilty based on the number of undecidable instances of input data versus decidable ones. It is proposed as well to extend the Turing's work on infinite logics and Oracle machines to a whole class of super-Turing models of computation. Next, the three new complexity classes for TM undecidable problems have been defined: U-complete (Universal complete), D-complete (Diagonalization complete) and H-complete (Hypercomputation complete) classes. The above has never been defined explicitly before by other scientists, and has been inspired by Cook/Levin NP-complete class for intractable problems. Finally, an equivalent to famous P is not equal to NP unanswered question for NP-complete class, has been answered negatively for U-complete class of complexity for undecidable problems.

In recent years, aspect-based sentiment analysis (ABSA) has made rapid progress and shown strong practical value. However, existing research and benchmarks are largely concentrated on high-resource languages, leaving fine-grained sentiment extraction in low-resource languages under-explored. To address this gap, we constructed the first Low-resource languages Aspect-based Sentiment Quadruple dataset, named LASQ, which includes two low-resource languages: Uzbek and Uyghur. Secondly, it includes a fine-grained target-aspect-opinion-sentiment quadruple extraction task. To facilitate future research, we designed a grid-tagging model that integrates syntactic knowledge. This model incorporates part-of-speech (POS) and dependency knowledge into the model through our designed Syntax Knowledge Embedding Module (SKEM), thereby alleviating the lexical sparsity problem caused by agglutinative languages. Experiments on LASQ demonstrate consistent gains over competitive baselines, validating both the dataset's utility and the effectiveness of the proposed modeling approach.

Accurate precipitation estimation is critical for flood forecasting, water resource management, and disaster preparedness. Satellite products provide global hourly coverage but contain systematic biases; ground-based gauges are accurate at point locations but too sparse for direct gridded correction. Existing methods fuse these sources by interpolating gauge observations onto the satellite grid, but treat each time step independently and therefore discard temporal structure in precipitation fields. We propose Neural Stochastic Process (NSP), a model that pairs a Neural Process encoder conditioning on arbitrary sets of gauge observations with a latent Neural SDE on a 2D spatial representation. NSP is trained under a single variational objective with simulation-free cost. We also introduce QPEBench, a benchmark of 43{,}756 hourly samples over the Contiguous United States (2021--2025) with four aligned data sources and six evaluation metrics. On QPEBench, NSP outperforms 13 baselines across all six metrics and surpasses JAXA's operational gauge-calibrated product. An additional experiment on Kyushu, Japan confirms generalization to a different region with independent data sources.

Conditional effects are commonly used measures for understanding how treatment effects vary across different groups, and are often used to target treatments/interventions to groups who benefit most. In this work we review existing methods and propose novel ones, focusing on the odds ratio (OR) and the risk ratio (RR). While estimation of the conditional average treatment effect (ATE) has been widely studied, estimators for the OR and RR lag behind, and cutting edge estimators such as those based on doubly robust transformations or orthogonal risk functions have not been generalized to these parameters. We propose such a generalization here, focusing on the DR-learner and the R-learner. We derive orthogonal risk functions for the OR and RR and show that the associated pseudo-outcomes satisfy second-order conditional-mean remainder properties analogous to the ATE case. We also evaluate estimators for the conditional ATE, OR, and RR in a comprehensive nonparametric Monte Carlo simulation study to compare them with common alternatives under hundreds of different data-generating distributions. Our numerical studies provide empirical guidance for choosing an estimator. For instance, they show that while parametric models are useful in very simple settings, the proposed nonparametric estimators significantly reduce bias and mean squared error in the more complex settings expected in the real world. We illustrate the methods in the analysis of physical activity and sleep trouble in U.S. adults using data from the National Health and Nutrition Examination Survey (NHANES). The results demonstrate that our estimators uncover substantial treatment effect heterogeneity that is obscured by traditional regression approaches and lead to improved treatment decision rules, highlighting the importance of data-adaptive methods for advancing precision health research.

In modern cloud and heterogeneous distributed infrastructures, container images are widely used as the deployment unit for machine learning applications. An image bundles the application with its entire platform-specific execution environment and can be directly launched into a container instance. However, this approach forces developers to build and maintain separate images for each target deployment platform. This limitation is particularly evident for widely used interpreted languages such as Python and R in data analytics and machine learning, where application code is inherently cross-platform, yet the runtime dependencies are highly platform-specific. With emerging computing paradigms such as sky computing and edge computing, which demand seamless workload migration and cross-platform deployment, traditional images not only introduce inefficiencies in storage and network usage, but also impose substantial burdens on developers, who must repeatedly craft and manage platform-specific builds. To address these challenges, we propose a lazy-build approach that defers platform-specific construction to the deployment stage, thus keeping the image itself cross-platform. To enable this, we introduce a new image format, CIR (Container Intermediate Representation), together with its pre-builder and lazy-builder. CIR targets interpreted-language applications and only stores the identifiers of the application's direct dependencies, leaving platform adaptation to the lazy-builder, which at deployment time assembles the actual dependencies into runnable containers. A single CIR can therefore be deployed across heterogeneous platforms while reducing image size by 95% compared to conventional images that bundle all dependencies. In our evaluation, CIR reduces deployment time by 40-60% compared with pre-built images, outperforming state-of-the-art systems such as Docker, Buildah, and Apptainer.

Interpreting chest X-rays is inherently challenging due to the overlap between anatomical structures and the subtle presentation of many clinically significant pathologies, making accurate diagnosis time-consuming even for experienced radiologists. Recent radiology-focused foundation models, such as LLaVA-Rad and Maira-2, have positioned multi-modal large language models (MLLMs) at the forefront of automated radiology report generation (RRG). However, despite these advances, current foundation models generate reports in a single forward pass. This decoding strategy diminishes attention to visual tokens and increases reliance on language priors as generation proceeds, which in turn introduces spurious pathology co-occurrences in the generated reports. To mitigate these limitations, we propose Category-Wise Contrastive Decoding (CWCD), a novel and modular framework designed to enhance structured radiology report generation (SRRG). Our approach introduces category-specific parameterization and generates category-wise reports by contrasting normal X-rays with masked X-rays using category-specific visual prompts. Experimental results demonstrate that CWCD consistently outperforms baseline methods across both clinical efficacy and natural language generation metrics. An ablation study further elucidates the contribution of each architectural component to overall performance.

We introduce IMPACT, a synchronized five-view RGB-D dataset for deployment-oriented industrial procedural understanding, built around real assembly and disassembly of a commercial angle grinder with professional-grade tools. To our knowledge, IMPACT is the first real industrial assembly benchmark that jointly provides synchronized ego-exo RGB-D capture, decoupled bimanual annotation, compliance-aware state tracking, and explicit anomaly--recovery supervision within a single real industrial workflow. It comprises 112 trials from 13 participants totaling 39.5 hours, with multi-route execution governed by a partial-order prerequisite graph, a six-category anomaly taxonomy, and operator cognitive load measured via NASA-TLX. The annotation hierarchy links hand-specific atomic actions to coarse procedural steps, component assembly states, and per-hand compliance phases, with synchronized null spans across views to decouple perceptual limitations from algorithmic failure. Systematic baselines reveal fundamental limitations that remain invisible to single-task benchmarks, particularly under realistic deployment conditions that involve incomplete observations, flexible execution paths, and corrective behavior. The full dataset, annotations, and evaluation code are available at https://github.com/Kratos-Wen/IMPACT.

Edge-based multimodal medical monitoring requires models that balance diagnostic accuracy with severe energy constraints. Continuous acquisition of ECG, PPG, EMG, and IMU streams rapidly drains wearable batteries, often limiting operation to under 10 hours, while existing systems overlook the high temporal redundancy present in physiological signals. We introduce Adaptive Multimodal Intelligence (AMI), an end-to-end framework that jointly learns when to sense and how to infer. AMI integrates three components: (1) a lightweight Agentic Modality Controller that uses differentiable Gumbel-Sigmoid gating to dynamically select active sensors based on model confidence and task relevance; (2) a Learned Sigma-Delta Sensing module that applies patch-wise Delta-Sigma operations with learnable thresholds to skip temporally redundant samples; and (3) a Foundation-backed Multimodal Prediction Model built on unimodal foundation encoders and a cross-modal transformer with temporal context, enabling robust fusion even under gated or missing inputs. These components are trained jointly via a multi-objective loss combining classification accuracy, sparsity regularization, cross-modal alignment, and predictive coding. AMI is hardware-aware, supporting dynamic computation graphs and masked operations, leading to real energy and latency savings. Across MHEALTH, HMC Sleep, and WESAD datasets, it reduces sensor usage by 48.8% while improving state-of-the-art accuracy by 1.9% on average.

We address jailbreaks, backdoors, and unlearning for large language models (LLMs). Unlike prior work, which trains LLMs based on their actions when given malign instructions, our method specifically trains the model to change how it interprets instructions. Our method, Latent Instruction Representation Alignment (LIRA), greatly improves generalization. We further boost generalization through an internally adversarial training algorithm. Our methods block over 99% of PEZ jailbreak attacks; remove a challenging insecure code backdoor; and achieve optimal forgetting on WMDP cyber with negligible loss of benign capabilities.

Inferring nationality from personal names is a critical capability for equity and bias monitoring, personalization, and a valuable tool in biomedical and sociological research. However, existing name-based nationality classifiers are typically trained on relatively small or source-specific labeled datasets, which can introduce coverage gaps and limit performance for underrepresented countries. While large language models (LLMs) demonstrate strong zero-shot performance for name-based nationality prediction, their computational cost and latency make them impractical for real-time, large-scale deployment. In this work, we created a large-scale name-nationality dataset from the Open Academic Graph (OAG) and introduce a framework that leverages LLMs as dataset enrichers rather than inference engines. We augment low-resource countries with LLM-generated names and evaluate on real and synthetic-tail test sets. We find that augmentation produces large gains when evaluation includes synthetic tail names and still offers a modest lift on tail-country metrics otherwise. Overall, NameBERT models achieve significantly higher accuracy than state-of-the-art baselines across both in- and out-of-domain tasks, while remaining efficient for large-scale inference compared to LLMs.

Video-based human-object interaction (HOI) understanding requires both detecting ongoing interactions and anticipating their future evolution. However, existing methods usually treat anticipation as a downstream forecasting task built on externally constructed human-object pairs, limiting joint reasoning between detection and prediction. In addition, sparse keyframe annotations in current benchmarks can temporally misalign nominal future labels from actual future dynamics, reducing the reliability of anticipation evaluation. To address these issues, we introduce DETAnt-HOI, a temporally corrected benchmark derived from VidHOI and Action Genome for more faithful multi-horizon evaluation, and HOI-DA, a pair-centric framework that jointly performs subject-object localization, present HOI detection, and future anticipation by modeling future interactions as residual transitions from current pair states. Experiments show consistent improvements in both detection and anticipation, with larger gains at longer horizons. Our results highlight that anticipation is most effective when learned jointly with detection as a structural constraint on pair-level video representation learning. Benchmark and code will be publicly available.

Large language models are increasingly used to generate code from natural language, but ensuring correctness remains challenging. Formal verification offers a principled way to obtain such guarantees by proving that a program satisfies a formal specification. However, specifications are frequently missing in real-world codebases, and writing high-quality specifications remains expensive and expertise-intensive. We present VeriSpecGen, a traceable refinement framework that synthesizes intent-aligned specifications in Lean through requirement-level attribution and localized repair. VeriSpecGen decomposes natural language into atomic requirements and generates requirement-targeted tests with explicit traceability maps to validate generated specifications. When validation fails, traceability maps attribute failures to specific requirements, enabling targeted clause-level repairs. VeriSpecGen achieve 86.6% on VERINA SpecGen task using Claude Opus 4.5, improving over baselines by up to 31.8 points across different model families and scales. Beyond inference-time gains, we generate 343K training examples from VeriSpecGen refinement trajectories and demonstrate that training on these trajectories substantially improves specification synthesis by 62-106% relative and transfers gains to general reasoning abilities.

Vision foundation models (VFMs) and Bird's Eye View (BEV) representation have advanced visual perception substantially, yet their internal spatial representations assume the rectilinear geometry of pinhole cameras. Fisheye cameras, widely deployed on production autonomous vehicles for their surround-view coverage, exhibit severe radial distortion that renders these representations geometrically inconsistent. At the same time, the scarcity of large-scale fisheye annotations makes retraining foundation models from scratch impractical. We present \ours, a lightweight framework that adapts frozen VFMs to fisheye geometry through two components: a frozen DINOv2 backbone with Low-Rank Adaptation (LoRA) that transfers rich self-supervised features to fisheye without task-specific pretraining, and Fisheye Rotary Position Embedding (FishRoPE), which reparameterizes the attention mechanism in the spherical coordinates of the fisheye projection so that both self-attention and cross-attention operate on angular separation rather than pixel distance. FishRoPE is architecture-agnostic, introduces negligible computational overhead, and naturally reduces to the standard formulation under pinhole geometry. We evaluate \ours on WoodScape 2D detection (54.3 mAP) and SynWoodScapes BEV segmentation (65.1 mIoU), where it achieves state-of-the-art results on both benchmarks.

Large-scale LLM training is increasingly susceptible to hardware defects stemming from manufacturing escapes and silicon aging. These defects manifest as Silent Data Corruption (SDC) that perturb gradients and parameters throughout the training process. We present LLM-PRISM, a methodology to characterize LLM pre-training resilience to hardware faults. LLM-PRISM couples RTL-level GPU fault simulation with a stochastic injection engine embedded in Megatron-LM. Through 7,664 training runs across FP16, BF16, and FP8 regimes, we analyze how fault type, rate, and numeric format govern resilience. We find that while LLMs resist low-frequency faults, impact is highly non-uniform; critical datapaths and specific precision formats can induce catastrophic divergence even at moderate fault rates. This study provides the first hardware-grounded, pre-training characterization of SDC resilience.

Terminology substitution errors in clinical notes, where one medical term is replaced by a linguistically valid but clinically different term, pose a persistent challenge for automated error detection in healthcare. We introduce BLUEmed, a multi-agent debate framework augmented with hybrid Retrieval-Augmented Generation (RAG) that combines evidence-grounded reasoning with multi-perspective verification for clinical error detection. BLUEmed decomposes each clinical note into focused sub-queries, retrieves source-partitioned evidence through dense, sparse, and online retrieval, and assigns two domain expert agents distinct knowledge bases to produce independent analyses; when the experts disagree, a structured counter-argumentation round and cross-source adjudication resolve the conflict, followed by a cascading safety layer that filters common false-positive patterns. We evaluate BLUEmed on a clinical terminology substitution detection benchmark under both zero-shot and few-shot prompting with multiple backbone models spanning proprietary and open-source families. Experimental results show that BLUEmed achieves the best accuracy (69.13%), ROC-AUC (74.45%), and PR-AUC (72.44%) under few-shot prompting, outperforming both single-agent RAG and debate-only baselines. Further analyses across six backbone models and two prompting strategies confirm that retrieval augmentation and structured debate are complementary, and that the framework benefits most from models with sufficient instruction-following and clinical language understanding.

Mapping parallel threads onto non-box-shaped domains is a known challenge in GPU computing that, if done efficiently, can prevent severe performance penalties from allocating unnecessary computational resources. Currently, achieving this optimal efficiency requires significant analytical human time and effort to manually derive bespoke mapping functions for each specific geometry. This work introduces a novel approach leveraging the symbolic reasoning capabilities of Large Language Models (LLMs) to automate this mathematical derivation process entirely through in-context learning. Focusing on state-of-the-art open-weights models, we conducted a rigorous comparative analysis across spatial domains of increasing complexity. Our results demonstrate that modern local LLMs successfully infer exact O(1) and O(log N) mapping equations for complex 2D/3D dense domains and 2D fractals, vastly outperforming traditional symbolic regression methods which systematically failed at this discrete task. Crucially, we profile the energetic viability of this approach on high-performance infrastructure, drawing a clear distinction between the code-generation and execution phases. While the one-time inference of the equation incurs a high energy penalty -- particularly for reasoning-focused models like DeepSeek-R1 -- this is a single upfront investment. Once integrated, the generated analytical kernels eliminate block waste entirely, yielding massive repeated energy and time savings (e.g., up to 4833x speedup and 2890x energy reduction) during actual GPU workloads. Finally, we identify a current "reasoning ceiling" when these models face highly recursive 3D fractals tested in this work (e.g., the Menger Sponge). This limitation establishes a clear benchmark for the maturity of open-weight architectures, charting a viable and sovereign path toward fully automated, energy-efficient GPU resource optimization.

Accurate estimation of cancer risk from longitudinal electronic health records (EHRs) could support earlier detection and improved care, but modeling such complex patient trajectories remains challenging. We present TrajOnco, a training-free, multi-agent large language model (LLM) framework designed for scalable multi-cancer early detection. Using a chain-of-agents architecture with long-term memory, TrajOnco performs temporal reasoning over sequential clinical events to generate patient-level summaries, evidence-linked rationales, and predicted risk scores. We evaluated TrajOnco on de-identified Truveta EHR data across 15 cancer types using matched case-control cohorts, predicting risk of cancer diagnosis at 1 year. In zero-shot evaluation, TrajOnco achieved AUROCs of 0.64-0.80, performing comparably to supervised machine learning in a lung cancer benchmark while demonstrating better temporal reasoning than single-agent LLMs. The multi-agent design also enabled effective temporal reasoning with smaller-capacity models such as GPT-4.1-mini. The fidelity of TrajOnco's output was validated through human evaluation. Furthermore, TrajOnco's interpretable reasoning outputs can be aggregated to reveal population-level risk patterns that align with established clinical knowledge. These findings highlight the potential of multi-agent LLMs to execute interpretable temporal reasoning over longitudinal EHRs, advancing both scalable multi-cancer early detection and clinical insight generation.

From adversarial robustness to multi-agent learning, many machine learning tasks can be cast as finite-sum min-max optimization or, more generally, as variational inequality problems (VIPs). Owing to their simplicity and scalability, stochastic gradient methods with constant step size are widely used, despite the fact that they converge only up to a constant term. Among the many heuristics adopted in practice, two classical techniques have recently attracted attention to mitigate this issue: \emph{Random Reshuffling} of data and \emph{Richardson--Romberg extrapolation} across iterates. Random Reshuffling sharpens the mean-squared error (MSE) of the estimated solution, while Richardson-Romberg extrapolation acts orthogonally, providing a second-order reduction in its bias. In this work, we show that their composition is strictly better than both, not only maintaining the enhanced MSE guarantees but also yielding an even greater cubic refinement in the bias. To the best of our knowledge, our work provides the first theoretical guarantees for such a synergy in structured non-monotone VIPs. Our analysis proceeds in two steps: (i) we smooth the discrete noise induced by reshuffling and leverage tools from continuous-state Markov chain theory to establish a novel law of large numbers and a central limit theorem for its iterates; and (ii) we employ spectral tensor techniques to prove that extrapolation debiases and sharpens the asymptotic behavior even under the biased gradient oracle induced by reshuffling. Finally, extensive experiments validate our theory, consistently demonstrating substantial speedups in practice.

This study proposes Structural Gating and Effect-aligned Discovery for Temporal Causal Discovery (SGED-TCD), a novel and general framework for lag-resolved causal discovery in complex multivariate time series. SGED-TCD combines explicit structural gating, stability-oriented learning, perturbation-effect alignment, and unified graph extraction to improve the interpretability, robustness, and functional consistency of inferred causal graphs. To evaluate its effectiveness in a representative real-world setting, we apply SGED-TCD to teleconnection-driven compound heatwave--air-pollution extremes in eastern and northern China. Using large-scale climate indices, regional circulation and boundary-layer variables, and compound extreme indicators, the framework reconstructs weighted causal networks with explicit dominant lags and relative causal importance. The inferred networks reveal clear regional and seasonal heterogeneity: warm-season extremes in Eastern China are mainly linked to low-latitude oceanic variability through circulation, radiation, and ventilation pathways, whereas cold-season extremes in Northern China are more strongly governed by high-latitude circulation variability associated with boundary-layer suppression and persistent stagnation. These results show that SGED-TCD can recover physically interpretable, hierarchical, and lag-resolved causal pathways in a challenging climate--environment system. More broadly, the proposed framework is not restricted to the present application and provides a general basis for temporal causal discovery in other complex domains.

Media narratives wield tremendous power in shaping public opinion, yet computational approaches struggle to capture the nuanced storytelling structures that communication theory emphasizes as central to how meaning is constructed. Existing approaches either miss subtle narrative patterns through coarse-grained analysis or require domain-specific taxonomies that limit scalability. To bridge this gap, we present a framework for inducing rich narrative schemas by jointly modeling events and characters via structured clustering. Our approach produces explainable narrative schemas that align with established framing theory while scaling to large corpora without exhaustive manual annotation.

Audio-driven human video generation has achieved remarkable success in monologue scenarios, largely driven by advancements in powerful video generation foundation models. Moving beyond monologues, authentic human communication is inherently a full-duplex interactive process, requiring virtual agents not only to articulate their own speech but also to react naturally to incoming conversational audio. Most existing methods simply extend conventional audio-driven paradigms to listening scenarios. However, relying on strict frame-to-frame alignment renders the model's response to long-range conversational dynamics rigid, whereas directly introducing global attention catastrophically degrades lip synchronization. Recognizing the unique temporal Scale Discrepancy between talking and listening behaviors, we introduce a multi-head Gaussian kernel to explicitly inject this physical intuition into the model as a progressive temporal inductive bias. Building upon this, we construct a full-duplex interactive virtual agent capable of simultaneously processing dual-stream audio inputs for both talking and listening. Furthermore, we introduce a rigorously cleaned Talking-Listening dataset VoxHear featuring perfectly decoupled speech and background audio tracks. Extensive experiments demonstrate that our approach successfully fuses strong temporal alignment with deep contextual semantics, setting a new state-of-the-art for generating highly natural and responsive full-duplex interactive digital humans. The project page is available at https://warmcongee.github.io/beyond-monologue/ .

Accurate battery health prognosis using State of Health (SOH) estimation is essential for the reliability of multi-scale battery energy storage, yet existing methods are limited in generalizability across diverse battery chemistries and operating conditions. The inability of standard neural networks to capture the complex, high-dimensional physics of battery degradation is a major contributor to these limitations. To address this, a physics-informed neural network with the Quantum Feature Mapping(QFM) technique (QPINN) is proposed. QPINN projects raw battery sensor data into a high-dimensional Hilbert space, creating a highly expressive feature set that effectively captures subtle, non-linear degradation patterns using Nyström method. These quantum-enhanced features are then processed by a physics-informed network that enforces physical constraints. The proposed method achieves an average SOH estimation accuracy of 99.46\% across different datasets, substantially outperforming state-of-the-art baselines, with reductions in MAPE and RMSE of up to 65\% and 62\%, respectively. This method was validated on a large-scale, multi-chemistry dataset of 310,705 samples from 387 cells, and further showed notable adaptability in cross-validation settings, successfully transferring from one chemistry to another without relying on target-domain SOH labels.

Low-light image enhancement (LLIE) aims to restore natural visibility, color fidelity, and structural detail under severe illumination degradation. State-of-the-art (SOTA) LLIE techniques often rely on large models and multi-stage training, limiting practicality for edge deployment. Moreover, their dependence on a single color space introduces instability and visible exposure or color artifacts. To address these, we propose Multinex, an ultra-lightweight structured framework that integrates multiple fine-grained representations within a principled Retinex residual formulation. It decomposes an image into illumination and color prior stacks derived from distinct analytic representations, and learns to fuse these representations into luminance and reflectance adjustments required to correct exposure. By prioritizing enhancement over reconstruction and exploiting lightweight neural operations, Multinex significantly reduces computational cost, exemplified by its lightweight (45K parameters) and nano (0.7K parameters) versions. Extensive benchmarks show that all lightweight variants significantly outperform their corresponding lightweight SOTA models, and reach comparable performance to heavy models. Paper page available at https://albrateanu.github.io/multinex.

Stateful tool-using LLM agents treat the context window as working memory, yet today's agent harnesses manage residency and durability as best-effort, causing recurring failures: lost state after compaction, bypassed flushes on reset, and destructive writeback. We present \textsc{ClawVM}, a virtual memory layer that manages state as typed pages with minimum-fidelity invariants, multi-resolution representations under a token budget, and validated writeback at every lifecycle boundary. Because the harness already assembles prompts, mediates tools, and observes lifecycle events, it is the natural enforcement point; placing the contract there makes residency and durability deterministic and auditable. Across synthetic workloads, 12 real-session traces, and adversarial stress tests, \textsc{ClawVM} eliminates all policy-controllable faults whenever the minimum-fidelity set fits within the token budget, confirmed by an offline oracle, and adds median <50 microseconds of policy-engine overhead per turn.

Large Language Models (LLMs) have been recently proposed for supporting domain modeling tasks mostly related to the completion of partial models by recommending additional model elements. However, there are many more modeling tasks, one of them being the instantiation of domain models to represent concrete domain objects. While there is considerable work supporting the generation of structurally valid instantiations, there are still open challenges to incorporating real-world semantics by having realistic values contained in instances and ensuring the generation of semantically diverse models. Only then will such generated models become human-understandable and helpful in educational or data-driven research contexts. To tackle these challenges, this paper presents an approach that employs LLMs and two prompting strategies in combination with existing model validation tools for instantiating semantically realistic and diverse domain models expressed as UML class diagrams. We have applied our approach to models used in education and available in the literature from different domains and evaluated the generated instances in terms of syntactic correctness, model conformance, semantic correctness, and diversity of the generated values. The results show that the generated instances are mostly syntactically correct, that they conform to the domain model, and that there are only a few semantic errors. Moreover, the generated instance values are semantically diverse, i.e., concrete realistic examples in line with the domain and the combination of the values within one model are semantically coherent.

Understanding the reasons behind past code changes is critical for many software engineering tasks, including refactoring and reviewing code, diagnosing bugs, and implementing new features. Unfortunately, locating and reconstructing this rationale can be difficult for developers because the information is often fragmented, inconsistently documented, and scattered across different artifacts such as commit messages, issue reports, and PRs. In this paper, we address this challenge in two steps. First, we conduct an empirical study of 63 commits from five open-source Java projects to analyze how rationale components (e.g., a change's goal, need, and alternative) are distributed across artifacts. We find that the rationale is highly fragmented: commit messages and pull requests primarily capture goals, while needs and alternatives are more often found in issues and PRs. Other components are scarce but found in artifacts other than commit messages. No single artifact type captures all components, underscoring the need for cross-document reasoning and synthesis. Second, we introduce ARGUS, an LLM-based approach that identifies sentences expressing goal, need, and alternative across a commit's artifacts and creates concise rationale summaries to support code comprehension and maintenance tasks. We evaluated ARGUS on the 63 commits and compared its performance against baseline variants. The best-performing version achieved 51.4% precision and 93.2% recall for rationale identification, while producing rationale summaries rated as accurate. A user study with 12 Java developers further showed that these summaries were perceived as useful and helpful for tasks such as code review, documentation, and debugging. Our results highlight the need for multi-document reasoning in capturing rationale and demonstrate the potential of ARGUS to help developers understand and maintain software systems.

We study the operation of community water systems, where pumps and valves must be scheduled to reliably meet water demands while minimizing energy consumption. While existing optimization-based methods are effective under well-modeled environments, real-world community scenarios exhibit highly dynamic contexts-such as human activities, weather variations, etc-that significantly affect water demand patterns and operational targets across different zones. Traditional optimization approaches struggle to aggregate and adapt to such heterogeneous and rapidly evolving contextual information in real time. While Large Language Model (LLM) agents offer strong capabilities for understanding heterogeneous community context, they are not suitable for directly producing reliable real-time control actions. To address these challenges, we propose a bi-level AI-agent-based framework, WaterAdmin, which integrates LLM-based community context abstraction at the upper level with optimization-based operational control at the lower level. This design leverages the complementary strengths of both paradigms to enable adaptive and reliable operation. We implement WaterAdmin on the hydraulic simulation platform EPANET and demonstrate superior performance in maintaining pressure reliability and reducing energy consumption under highly dynamic community contexts.

\textbf{VeriTrans} is a reliability-first ML system that compiles natural-language requirements into solver-ready logic with validator-gated reliability. The pipeline integrates an instruction-tuned NL$\!\to\!$PL translator, round-trip reconstruction (PL$\!\to\!$NL) used as a high-precision acceptance gate, and canonical PL$\!\to\!$CNF compilation, all executed via fixed API configuration (temperature$=0$; fine-tuning runs use seed$=42$) and per-item artifact logging (prompts, outputs, hashes) to support auditability and replay-driven debugging. On \textbf{SatBench} (2{,}100 specifications), VeriTrans achieves 94.46\% SAT/UNSAT correctness and 87.73\% median round-trip similarity. Compact fine-tuning on 100--150 curated examples improves fidelity by about 1--1.5\,pp without increasing latency (mean 25.8\,s/spec on our 201-spec runtime subset). A thresholded acceptance policy on the round-trip score exposes a reliability--coverage knob: at $τ{=}75$, roughly 68\% of items are retained with $\sim$94\% correctness on the accepted set. Validator overhead contributes $<15\%$ of end-to-end runtime, and all prompts/responses and timing metadata are logged to enable replay-driven debugging and regression testing. By separating learned translation from symbolic verification and enforcing deterministic, validator-gated acceptance, VeriTrans turns NL$\!\to\!$logic front-ends into auditable, reproducible components for reliability-critical workflows.

Employee attrition presents a major challenge for organizations, increasing costs and reducing productivity. Predicting attrition accurately enables proactive retention strategies, but existing machine learning models often struggle to capture complex feature interactions in tabular HR datasets. While tree-based models such as XGBoost and LightGBM perform well on structured data, traditional encoding techniques like one-hot encoding can introduce sparsity and fail to preserve semantic relationships between categorical features. This study explores a hybrid approach by integrating SAINT (Self-Attention and Intersample Attention Transformer)-generated embeddings with tree-based models to enhance employee attrition prediction. SAINT leverages self-attention mechanisms to model intricate feature interactions. In this study, we explore SAINT both as a standalone classifier and as a feature extractor for tree-based models. We evaluate the performance, generalizability, and interpretability of standalone models (SAINT, XGBoost, LightGBM) and hybrid models that combine SAINT embeddings with tree-based classifiers. Experimental results show that standalone tree-based models outperform both the standalone SAINT model and the hybrid approaches in predictive accuracy and generalization. Contrary to expectations, the hybrid models did not improve performance. One possible explanation is that tree-based models struggle to utilize dense, high-dimensional embeddings effectively. Additionally, the hybrid approach significantly reduced interpretability, making model decisions harder to explain. These findings suggest that transformer-based embeddings, while capturing feature relationships, do not necessarily enhance tree-based classifiers. Future research should explore alternative fusion strategies for integrating deep learning with structured data.

Large language models (LLMs) demonstrate strong performance in math reasoning benchmarks, but their performance varies inconsistently across problems with varying levels of difficulty. This paper describes Adaptive Multi-Expert Reasoning (AMR), a framework that focuses on problem complexity by reasoning with dynamically adapted strategies. An agile routing system that focuses on problem text predicts problems' difficulty and uncertainty and guides a reconfigurable sampling mechanism to manage the breadth of generation. Three specialized experts create candidate responses, which are modified during multiple correction and finalization phases. A neural verifier assesses the correctness of responses, while a clustering-based aggregation technique identifies the final candidate answer based on a combination of consensus and answer quality. When evaluated on the GSM8K dataset, AMR achieved 75.28% accuracy while only using the original training data. This result outperformed the majority of comparable 7B models that were trained on synthetic data. This showcases that models using difficulty-based routing and uncertainty-driven aggregation are efficient and effective in improving math reasoning models' robustness.

Young children demonstrate early abilities to understand their physical world, estimating depth, motion, object coherence, interactions, and many other aspects of physical scene understanding. Children are both data-efficient and flexible cognitive systems, creating competence despite extremely limited training data, while generalizing to myriad untrained tasks -- a major challenge even for today's best AI systems. Here we introduce a novel computational hypothesis for these abilities, the Zero-shot Visual World Model (ZWM). ZWM is based on three principles: a sparse temporally-factored predictor that decouples appearance from dynamics; zero-shot estimation through approximate causal inference; and composition of inferences to build more complex abilities. We show that ZWM can be learned from the first-person experience of a single child, rapidly generating competence across multiple physical understanding benchmarks. It also broadly recapitulates behavioral signatures of child development and builds brain-like internal representations. Our work presents a blueprint for efficient and flexible learning from human-scale data, advancing both a computational account for children's early physical understanding and a path toward data-efficient AI systems.

We present the progress of the GPT family from GPT-3 through GPT-3.5, GPT-4, GPT-4 Turbo, GPT-4o, GPT-4.1, and the GPT-5 family. Our work is comparative rather than merely historical. We investigates how the family evolved in technical framing, user interaction, modality, deployment architecture, and governance viewpoint. The work focuses on five recurring themes: technical progression, capability changes, deployment shifts, persistent limitations, and downstream consequences. In term of research design, we consider official technical reports, system cards, API and model documentation, product announcements, release notes, and peer-reviewed secondary studies. A primary assertion is that later GPT generations should not be interpreted only as larger or more accurate language models. Instead, the family evolves from a scaled few-shot text predictor into a set of aligned, multimodal, tool-oriented, long-context, and increasingly workflow-integrated systems. This development complicates simple model-to-model comparison because product routing, tool access, safety tuning, and interface design become part of the effective system. Across generations, several limitations remain unchanged: hallucination, prompt sensitivity, benchmark fragility, uneven behavior across domains and populations, and incomplete public transparency about architecture and training. However, the family has evolved software development, educational practice, information work, interface design, and discussions of frontier-model governance. We infer that the transition from GPT-3 to GPT-5 is best understood not only as an improvement in model capability, but also as a broader reformulation of what a deployable AI system is, how it is evaluated, and where responsibility should be located when such systems are used at scale.

Accurate weather nowcasting remains one of the central challenges in atmospheric science, with critical implications for climate resilience, energy security, and disaster preparedness. Since it is not feasible to deploy observation stations everywhere, some regions lack dense observational networks, resulting in unreliable short-term wind predictions across those unobserved areas. Here we present a deep graph self-supervised framework that extends nowcasting capability into such unobserved regions without requiring new sensors. Our approach introduces "virtual nodes" into a diffusion and contrastive-based graph neural network, enabling the model to learn wind condition (i.e., speed, direction and gusts) in places with no direct measurements. Using high-temporal resolution weather station data across the Netherlands, we demonstrate that this approach reduces nowcast mean absolute error (MAE) of wind speed, gusts, and direction in unobserved regions by more than 30% - 46% compared with interpolation and regression methods. By enabling localized nowcasts where no measurements exist, this method opens new pathways for renewable energy integration, agricultural planning, and early-warning systems in data-sparse regions.

Large language models remain vulnerable to jailbreak attacks -- inputs designed to bypass safety mechanisms and elicit harmful responses -- despite advances in alignment and instruction tuning. We propose Head-Masked Nullspace Steering (HMNS), a circuit-level intervention that (i) identifies attention heads most causally responsible for a model's default behavior, (ii) suppresses their write paths via targeted column masking, and (iii) injects a perturbation constrained to the orthogonal complement of the muted subspace. HMNS operates in a closed-loop detection-intervention cycle, re-identifying causal heads and reapplying interventions across multiple decoding attempts. Across multiple jailbreak benchmarks, strong safety defenses, and widely used language models, HMNS attains state-of-the-art attack success rates with fewer queries than prior methods. Ablations confirm that nullspace-constrained injection, residual norm scaling, and iterative re-identification are key to its effectiveness. To our knowledge, this is the first jailbreak method to leverage geometry-aware, interpretability-informed interventions, highlighting a new paradigm for controlled model steering and adversarial safety circumvention.

Background: Large Language Models (LLMs) are transforming artificial intelligence applications in healthcare due to their ability to understand, generate, and summarize complex medical text. They offer valuable support to clinicians, researchers, and patients, yet their deployment in high-stakes clinical environments raises critical concerns regarding accuracy, reliability, and patient safety. Despite substantial attention in recent years, standardized benchmarking of LLMs for medical applications has been limited. Objective: This study addresses the need for a standardized comparative evaluation of LLMs in medical settings. Method: We evaluate multiple models, including ChatGPT, LLaMA, Grok, Gemini, and ChatDoctor, on core medical tasks such as patient note summarization and medical question answering, using the open-access datasets, MedMCQA, PubMedQA, and Asclepius, and assess performance through a combination of linguistic and task-specific metrics. Results: The results indicate that domain-specific models, such as ChatDoctor, excel in contextual reliability, producing medically accurate and semantically aligned text, whereas general-purpose models like Grok and LLaMA perform better in structured question-answering tasks, demonstrating higher quantitative accuracy. This highlights the complementary strengths of domain-specific and general-purpose LLMs depending on the medical task. Conclusion: Our findings suggest that LLMs can meaningfully support medical professionals and enhance clinical decision-making; however, their safe and effective deployment requires adherence to ethical standards, contextual accuracy, and human oversight in relevant cases. These results underscore the importance of task-specific evaluation and cautious integration of LLMs into healthcare workflows.

In CT angiography, the accurate segmentation of abdominal aortic aneurysms (AAAs) is difficult due to large anatomical variability, low-contrast vessel boundaries, and the close proximity of organs whose intensities resemble vascular structures, often leading to false positives. To address these challenges, we propose an anatomy-aware segmentation framework that integrates organ exclusion masks derived from TotalSegmentator into the training process. These masks encode explicit anatomical priors by identifying non-vascular organsand penalizing aneurysm predictions within these regions, thereby guiding the U-Net to focus on the aorta and its pathological dilation while suppressing anatomically implausible predictions. Despite being trained on a relatively small dataset, the anatomy-aware model achieves high accuracy, substantially reduces false positives, and improves boundary consistency compared to a standard U-Net baseline. The results demonstrate that incorporating anatomical knowledge through exclusion masks provides an efficient mechanism to enhance robustness and generalization, enabling reliable AAA segmentation even with limited training data.

Artificial Intelligence (AI) models, encompassing both traditional machine learning (ML) and more advanced approaches such as deep learning and large language models (LLMs), play a central role in modern applications. AI model lifecycle management involves the end-to-end process of managing these models, from data collection and preparation to model building, evaluation, deployment, and continuous monitoring. This process is inherently complex, as it requires the coordination of diverse services that manage AI artifacts such as datasets, dataflows, and models, all orchestrated to operate seamlessly. In this context, it is essential to isolate applications from the complexity of interacting with heterogeneous services, datasets, and AI platforms. In this paper, we introduce Gypscie, a cross-platform AI artifact management system. By providing a unified view of all AI artifacts, the Gypscie platform simplifies the development and deployment of AI applications. This unified view is realized through a knowledge graph that captures application semantics and a rule-based query language that supports reasoning over data and models. Model lifecycle activities are represented as high-level dataflows that can be scheduled across multiple platforms, such as servers, cloud platforms, or supercomputers. Finally, Gypscie records provenance information about the artifacts it produces, thereby enabling explainability. Our qualitative comparison with representative AI systems shows that Gypscie supports a broader range of functionalities across the AI artifact lifecycle. Our experimental evaluation demonstrates that Gypscie can successfully optimize and schedule dataflows on AI platforms from an abstract specification.

Cooperative perception allows connected vehicles and roadside infrastructure to share sensor observations, creating a fused scene representation beyond the capability of any single platform. However, most cooperative 3D object detectors use a uniform fusion strategy for all object classes, which limits their ability to handle the different geometric structures and point-sampling patterns of small and large objects. This problem is further reinforced by narrow evaluation protocols that often emphasize a single dominant class or only a few cooperation settings, leaving robust multi-class detection across diverse vehicle-to-everything interactions insufficiently explored. To address this gap, we propose a class-adaptive cooperative perception architecture for multi-class 3D object detection from LiDAR data. The model integrates four components: multi-scale window attention with learned scale routing for spatially adaptive feature extraction, a class-specific fusion module that separates small and large objects into attentive fusion pathways, bird's-eye-view enhancement through parallel dilated convolution and channel recalibration for richer contextual representation, and class-balanced objective weighting to reduce bias toward frequent categories. Experiments on the V2X-Real benchmark cover vehicle-centric, infrastructure-centric, vehicle-to-vehicle, infrastructure-to-infrastructure, and vehicle-to-infrastructure settings under identical backbone and training configurations. The proposed method consistently improves mean detection performance over strong intermediate-fusion baselines, with the largest gains on trucks, clear improvements on pedestrians, and competitive results on cars. These results show that aligning feature extraction and fusion with class-dependent geometry and point density leads to more balanced cooperative perception in realistic vehicle-to-everything deployments.

LLM-based coding agents are increasingly used to generate code, tests, and documentation. Still, their outputs can be plausible yet misaligned with developer intent and provide limited evidence for review in evolving projects. This limits our understanding of how to structure LLM pair-programming workflows so that artifacts remain reliable, auditable, and maintainable over time. To address this gap, this doctoral research proposes a systematic study of multi-agent LLM pair programming that externalizes intent and uses development tools for iterative validation. The plan includes three studies: translating informal problem statements into standards aligned requirements and formal specifications; refining tests and implementations using automated feedback, such as solver-backed counterexamples; and supporting maintenance tasks, including refactoring, API migrations, and documentation updates, while preserving validated behavior. The expected outcome is a clearer understanding of when multi-agent workflows increase trust, along with practical guidance for building reliable programming assistants for real-world development.

Large Vision-Language Models (LVLMs) rely on attention-based retrieval of safety instructions to maintain alignment during generation. Existing attacks typically optimize image perturbations to maximize harmful output likelihood, but suffer from slow convergence due to gradient conflict between adversarial objectives and the model's safety-retrieval mechanism. We propose Attention-Guided Visual Jailbreaking, which circumvents rather than overpowers safety alignment by directly manipulating attention patterns. Our method introduces two simple auxiliary objectives: (1) suppressing attention to alignment-relevant prefix tokens and (2) anchoring generation on adversarial image features. This simple yet effective push-pull formulation reduces gradient conflict by 45% and achieves 94.4% attack success rate on Qwen-VL (vs. 68.8% baseline) with 40% fewer iterations. At tighter perturbation budgets ($ε=8/255$), we maintain 59.0% ASR compared to 45.7% for standard methods. Mechanistic analysis reveals a failure mode we term safety blindness: successful attacks suppress system-prompt attention by 80%, causing models to generate harmful content not by overriding safety rules, but by failing to retrieve them.

Composed Image Retrieval (CIR) retrieves target images using a reference image paired with modification text. Despite rapid advances, all existing methods and datasets operate at the image level -- a single reference image plus modification text in, a single target image out -- while real e-commerce users reason about products shown from multiple viewpoints. We term this mismatch View Incompleteness and formally define a new Multi-View CIR task that generalizes standard CIR from image-level to product-level retrieval. To support this task, we construct FashionMV, the first large-scale multi-view fashion dataset for product-level CIR, comprising 127K products, 472K multi-view images, and over 220K CIR triplets, built through a fully automated pipeline leveraging large multimodal models. We further propose ProCIR (Product-level Composed Image Retrieval), a modeling framework built upon a multimodal large language model that employs three complementary mechanisms -- two-stage dialogue, caption-based alignment, and chain-of-thought guidance -- together with an optional supervised fine-tuning (SFT) stage that injects structured product knowledge prior to contrastive training. Systematic ablation across 16 configurations on three fashion benchmarks reveals that: (1) alignment is the single most critical mechanism; (2) the two-stage dialogue architecture is a prerequisite for effective alignment; and (3) SFT and chain-of-thought serve as partially redundant knowledge injection paths. Our best 0.8B-parameter model outperforms all baselines, including general-purpose embedding models 10x its size. The dataset, model, and code are publicly available at https://github.com/yuandaxia2001/FashionMV.

Modern HPC file systems can contain billions of files and hundreds of petabytes of data, making even simple questions increasingly intractable to answer. Traditional file system utilities such as find and du fail to scale to these sizes. While external indexing tools like GUFI and Brindexer improve query performance, they remain batch-oriented and unsuitable for heterogeneous, rapidly evolving environments. We present Icicle, a scalable framework for continuous file system metadata indexing and monitoring. Icicle maintains a unified, up-to-date, and queryable view of file system state while supporting both periodic snapshot-based ingestion for bulk metadata updates and event-based ingestion for real-time synchronization from production systems such as Lustre and IBM Storage Scale. Built on Apache Kafka and Apache Flink, Icicle provides high-throughput, fault-tolerant, and horizontally scalable ingestion of metadata events into two complementary search indexes, enabling both individual file discovery and aggregate summary statistics by user, group, and directory. This architecture enables efficient support for both coarse-grained administrative queries and interactive analytics over billions of objects. Our experimental evaluation on production-scale HPC datasets demonstrates order-of-magnitude throughput improvements over existing monitoring and indexing approaches, with tunable options for balancing consistency, latency, and metadata freshness.

Large Language Models (LLMs) have shown promising performance in time series modeling tasks, but do they truly understand time series data? While multiple benchmarks have been proposed to answer this fundamental question, most are manually curated and focus on narrow domains or specific skill sets. To address this limitation, we propose scalable methods for creating comprehensive time series reasoning benchmarks that combine the flexibility of templates with the creativity of LLM agents. We first develop TimeSeriesExam, a multiple-choice benchmark using synthetic time series to evaluate LLMs across five core reasoning categories: pattern recognitionnoise understandingsimilarity analysisanomaly detection, and causality. Then, with TimeSeriesExamAgent, we scale our approach by automatically generating benchmarks from real-world datasets spanning healthcare, finance and weather domains. Through multi-dimensional quality evaluation, we demonstrate that our automatically generated benchmarks achieve diversity comparable to manually curated alternatives. However, our experiments reveal that LLM performance remains limited in both abstract time series reasoning and domain-specific applications, highlighting ongoing challenges in enabling effective time series understanding in these models. TimeSeriesExamAgent is available at https://github.com/magwiazda/TimeSeriesExamAgent.

AI is increasingly deployed in multi-agent systems; however, most research considers only the behavior of individual models. We experimentally show that multi-agent "AI organizations" are simultaneously more effective at achieving business goals, but less aligned, than individual AI agents. We examine 12 tasks across two practical settings: an AI consultancy providing solutions to business problems and an AI software team developing software products. Across all settings, AI Organizations composed of aligned models produce solutions with higher utility but greater misalignment compared to a single aligned model. Our work demonstrates the importance of considering interacting systems of AI agents when doing both capabilities and safety research.

This essay identifies a failure mode of AI chat systems that we term attribution laundering: the model performs substantive cognitive work and then rhetorically credits the user for having generated the resulting insights. Unlike transparent versions of glad handing sycophancy, attribution laundering is systematically occluded to the person it affects and self-reinforcing -- eroding users' ability to accurately assess their own cognitive contributions over time. We trace the mechanisms at both individual and societal scales, from the chat interface that discourages scrutiny to the institutional pressures that reward adoption over accountability. The document itself is an artifact of the process it describes, and is color-coded accordingly -- though the views expressed are the authors' own, not those of any affiliated institution, and the boundary between the human author's views and Claude's is, as the essay argues, difficult to draw.

Autonomous language-model agents increasingly rely on installable skills and tools to complete user tasks. Static skill auditing can expose capability surface before deployment, but it cannot determine whether a particular invocation is unsafe under the current user request and runtime context. We therefore study skill invocation auditing as a continuous-risk estimation problem: given a user request, candidate skill, and runtime context, predict a score that supports ranking and triage before a hard intervention is applied. We introduce STARS, which combines a static capability prior, a request-conditioned invocation risk model, and a calibrated risk-fusion policy. To evaluate this setting, we construct SIA-Bench, a benchmark of 3,000 invocation records with group-safe splits, lineage metadata, runtime context, canonical action labels, and derived continuous-risk targets. On a held-out split of indirect prompt injection attacks, calibrated fusion reaches 0.439 high-risk AUPRC, improving over 0.405 for the contextual scorer and 0.380 for the strongest static baseline, while the contextual scorer remains better calibrated with 0.289 expected calibration error. On the locked in-distribution test split, gains are smaller and static priors remain useful. The resulting claim is therefore narrower: request-conditioned auditing is most valuable as an invocation-time risk-scoring and triage layer rather than as a replacement for static screening. Code is available at https://github.com/123zgj123/STARS.

Cross-modal retrieval between audio recordings and symbolic music representations (MIDI) remains challenging because continuous waveforms and discrete event sequences encode different aspects of the same performance. We study descriptor injection, the augmentation of modality-specific encoders with hand-crafted domain features, as a bridge across this gap. In a three-phase campaign covering 13 descriptor-mechanism combinations, 6 architectural families, and 3 training schedules, the best configuration reaches a mean S of 84.0 percent across five independent seeds, improving the descriptor-free baseline by 8.8 percentage points. Causal ablation shows that the audio descriptor A4, based on octave-band energy dynamics, drives the gain in the top dual models, while the MIDI descriptor D4 has only a weak inference-time effect despite improving training dynamics. We also introduce reverse cross-attention, where descriptor tokens query encoder features, reducing attention operations relative to the standard formulation while remaining competitive. CKA analysis shows that descriptors substantially increase audio-MIDI transformer layer alignment, indicating representational convergence rather than simple feature concatenation. Perturbation analysis identifies high-frequency octave bands as the dominant discriminative signal. All experiments use MAESTRO v3.0.0 with an evaluation protocol controlling for composer and piece similarity.

We prove that in a coupled Kuramoto oscillator network at stable equilibrium, the physical phase displacement under weak output nudging is the gradient of the loss with respect to natural frequencies, with equality as the nudging strength beta tends to zero. Prior oscillator equilibrium propagation work explicitly set aside natural frequency as a learnable parameter; we show that on sparse layered architectures, frequency learning outperforms coupling-weight learning among converged seeds (96.0% vs. 83.3% at matched parameter counts, p = 1.8e-12). The approximately 50% convergence failure rate under random initialization is a loss-landscape property, not a gradient error; topology-aware spectral seeding eliminates it in all settings tested (46/100 to 100/100 seeds on the primary task; 50/50 on a second task, K-only training, and a larger architecture).

In what way could a data breach involving government-issued IDs such as passports, driver's licenses, etc., rival a random voluntary disclosure on a nondescript social-media platform? At first glance, the former appears more significant, and that is a valid assessment. The disclosed data could contain an individual's date of birth and address; for all intents and purposes, a leak of that data would be disastrous. Given the threat, the latter scenario involving an innocuous online post seems comparatively harmless--or does it? From that post and others like it, a forensic linguist could stylometrically uncover equivalent pieces of information, estimating an age range for the author (adolescent or adult) and narrowing down their geographical location (specific country). While not an exact science--the determinations are statistical--stylometry can reveal comparable, though noticeably diluted, information about an individual. To prevent an ID from being breached, simply sharing it as little as possible suffices. Preventing the leakage of personal information from written text requires a more complex solution: adversarial stylometry. In this paper, we explore how performing homoglyph substitution--the replacement of characters with visually similar alternatives (e.g., "h" $\texttt{[U+0068]}$ $\rightarrow$ "h" $\texttt{[U+04BB]}$)--on text can degrade stylometric systems.

Existing tool-use benchmarks for LLM agents are overwhelmingly linear: our analysis of six benchmarks shows 55 to 100% of instances are simple chains of 2 to 5 steps. We introduce The Amazing Agent Race (AAR), a benchmark featuring directed acyclic graph (DAG) puzzles (or "legs") with fork-merge tool chains. We release 1,400 instances across two variants: sequential (800 legs) and compositional (600 DAG legs). Agents must navigate Wikipedia, execute multi-step tool chains, and aggregate results into a verifiable answer. Legs are procedurally generated from Wikipedia seeds across four difficulty levels with live-API validation. Three complementary metrics (finish-line accuracy, pit-stop visit rate, and roadblock completion rate) separately diagnose navigation, tool-use, and arithmetic failures. Evaluating three agent frameworks on 1,400 legs, the best achieves only 37.2% accuracy. Navigation errors dominate (27 to 52% of trials) while tool-use errors remain below 17%, and agent architecture matters as much as model scale (Claude Code matches Codex CLI at 37% with 6x fewer tokens). The compositional structure of AAR reveals that agents fail not at calling tools but at navigating to the right pages, a blind spot invisible to linear benchmarks. The project page can be accessed at: https://minnesotanlp.github.io/the-amazing-agent-race

Reinforcement learning agent-based simulation (RL-ABS) has become an important tool for electricity market mechanism analysis and evaluation. In the modeling of monotone, bounded, multi-segment stepwise bids, existing methods typically let the policy network first output an unconstrained action and then convert it into a feasible bid curve satisfying monotonicity and boundedness through post-processing mappings such as sorting, clipping, or projection. However, such post-processing mappings often fail to satisfy continuous differentiability, injectivity, and invertibility at boundaries or kinks, thereby causing gradient distortion and leading to spurious convergence in simulation results. Meanwhile, most existing studies conduct mechanism analysis and evaluation mainly on the basis of training-curve convergence, without rigorously assessing the distance between the simulation outcomes and Nash equilibrium, which severely undermines the credibility of the results. To address these issues, this paper proposes...

We provide an approach that closely estimates an organization's cyber resources directly from vulnerability timestamps, using a non-stationary queueing framework. Traditional attack-surface metrics operate on static snapshots, ignoring the core attack-defense dynamics within information systems, which exhibit bursty, heavy-tailed, and capacity-constrained behavior. Our approach to modeling such dynamics is based on a queueing abstraction of attack surfaces. We utilize a segmentation method to identify piecewise-stationary regimes via Gaussian mixture modeling (GMM) of queue length distributions. We fit segment-specific arrival, service, and resource parameters through the minimization of Kullback--Leibler divergence (KL) between the empirical and estimated distributions. Applied to both large-scale software supply chain data and multi-year private logistics enterprise cyber-ticket workflows, the model estimates organizational resources, measured in the time-varying active personnel and output rate per personnel, solely from bug report and fix timings for software supply chains, and discovery and patch timestamps in the enterprise setting. Our results provide 91--96\% accuracy in resource estimation, making the dynamic queueing framework a compelling approach for understanding attack surface dynamics. Further, our framework exposes resource bottlenecks, establishing a foundation for predictive workforce planning, patch-race modeling, and proactive cyber-risk management.

Complex systems such as aircraft engines, turbines, and industrial machinery often operate under dynamically changing conditions. These varying operating conditions can substantially influence degradation behavior and make prognostic modeling more challenging, as accurate prediction requires explicit consideration of operational effects. To address this issue, this paper proposes a novel multi-head attention-based fusion neural network. The proposed framework explicitly models and integrates three signal components: (1) the monotonic degradation trend, which reflects the underlying deterioration of the system; (2) discrete operating states, identified through clustering and encoded into dense embeddings; and (3) residual random noise, which captures unexplained variation in sensor measurements. The core strength of the framework lies in its architecture, which combines BiLSTM networks with attention mechanisms to better capture complex temporal dependencies. The attention mechanism allows the model to adaptively weight different time steps and sensor signals, improving its ability to extract prognostically relevant information. In addition, a fusion module is designed to integrate the outputs from the degradation-trend branch and the operating-state embeddings, enabling the model to capture their interactions more effectively. The proposed method is validated using a dataset from the NASA repository, and the results demonstrate its effectiveness.

Agentic code tasks such as fault localization and patch generation require processing long codebases under tight memory constraints, where the Key-Value (KV) cache becomes the primary inference bottleneck. Existing compression methods rely exclusively on attention signals to estimate token importance, systematically discarding structurally critical tokens such as call sites, branch conditions, and assignments that are essential for code understanding. We present CodeComp, a training-free KV cache compression framework that incorporates static program analysis into LLM inference via Code Property Graph priors extracted by Joern. Across bug localization and code generation benchmarks, CodeComp consistently outperforms attention-only compression baselines under equal memory budgets, recovering the majority of full-context accuracy under aggressive KV cache compression, while matching the patch generation quality of uncompressed full-context inference and integrating seamlessly into SGLang-based agentic coding pipelines without model modification.

3D medical image analysis is of great importance in disease diagnosis and treatment. Recently, multimodal large language models (MLLMs) have exhibited robust perceptual capacity, strong cross-modal alignment, and promising generalizability. Therefore, they have great potential to improve the performance of medical report generation (MRG) and medical visual question answering (MVQA), which serve as two important tasks in clinical scenarios. However, due to the scarcity of 3D medical images, existing 3D medical MLLMs suffer from insufficiently pretrained vision encoder and inability to extract customized image features for different kinds of tasks. In this paper, we propose to first transfer a 2D MLLM, which is well trained with 2D natural images, to support 3D medical volumetric inputs while reusing all of its pre-trained parameters. To enable the vision encoder to extract tailored image features for various tasks, we then design a Text-Guided Hierarchical MoE (TGH-MoE) framework, which can distinguish tasks under the guidance of the text prompt. Furthermore, we propose a two-stage training strategy to learn both task-shared and task-specific image features. As demonstrated empirically, our method outperforms existing 3D medical MLLMs in both MRG and MVQA tasks. Our code will be released once this paper is accepted.

Current multimodal models often suffer from shallow reasoning, leading to errors caused by incomplete or inconsistent thought processes. To address this limitation, we propose Self-Verification and Self-Rectification (SVSR), a unified framework that explicitly integrates self-verification and self-rectification into the model's reasoning pipeline, substantially improving robustness and reliability in complex visual understanding and multimodal reasoning tasks. SVSR is built on a novel three-stage training paradigm. First, we construct a high-quality unified preference dataset by refining reasoning traces from pre-trained vision-language models, incorporating both forward and backward reasoning to embed self-reflective signals. Second, we perform cold-start supervised fine-tuning on this dataset to learn structured, multi-step reasoning behaviors. Third, we apply a Semi-online Direct Preference Optimization (Semi-online DPO) process, continuously augmenting the training corpus with high-quality, model-generated reasoning traces filtered by a powerful teacher VLM. This pipeline enables the model to learn, elicit, and refine its ability to self-verify and self-rectify. Extensive experiments across diverse benchmarks demonstrate that SVSR improves reasoning accuracy and enables stronger generalization to unseen tasks and question types. Notably, once trained with explicit self-reflective reasoning, the model also exhibits improved implicit reasoning ability, outperforming strong baselines even when no explicit reasoning traces are provided. These results highlight the potential of SVSR for building more dependable, introspective, and cognitively aligned multimodal systems.

Machine Learning (ML) systems are increasingly used to support decision-making processes that affect individuals. However, these systems often rely on biased data, which can lead to unfair outcomes against specific groups. With the growing adoption of Automated Machine Learning (AutoML), the risk of intensifying discriminatory behaviours increases, as most frameworks primarily focus on model selection to maximise predictive performance. Previous research on fairness in AutoML had largely followed this trend, integrating fairness awareness only in the model selection or hyperparameter tuning, while neglecting other critical stages of the ML pipeline. This paper aims to study the impact of integrating fairness directly into the optimisation component of an AutoML framework that constructs complete ML pipelines, from data selection and transformations to model selection and tuning. As selecting appropriate fairness metrics remains a key challenge, our work incorporates complementary fairness metrics to capture different dimensions of fairness during the optimisation. Their integration within AutoML resulted in measurable differences compared to a baseline focused solely on predictive performance. Despite a 9.4% decrease in predictive power, the average fairness improved by 14.5%, accompanied by a 35.7% reduction in data usage. Furthermore, fairness integration produced complete yet simpler final solutions, suggesting that model complexity is not always required to achieve balanced and fair ML solutions.

Rendering large-scale, unbounded scenes on AR/VR-class devices is constrained by the computation, bandwidth, and storage cost of 3D Gaussian Splatting (3DGS). We propose a low-power, low-cost 3DGS hardware accelerator that renders full-HD images in real time, together with a hardware-friendly compression pipeline that combines iterative Gaussian pruning and fine-tuning, progressive spherical harmonics (SH) degree reduction, and vector quantization of all SH coefficients and colors. The scheme achieves a $51.6\times$ model-size reduction with a 0.743 dB PSNR loss. The accelerator uses a frame-level pipeline that integrates point-based culling and projection with tile-based sorting and rasterization, skips zero-Jacobian matrix multiplications (reducing processing elements by 63\% and computation by 53\%), and adopts comparison-free tile-based sorting with deterministic latency. Implemented in a TSMC 28-nm process at 800 MHz, the design occupies $0.66~\text{mm}^2$ with 1.1438 M gates and 120 kB SRAM, consumes 0.219 W, and delivers 1219 Mpixels/J at 267.5 Mpixels/s, enabling 1080p at 129 FPS. Overall, it is $5.98\times$ smaller in area, $5.94\times$ higher throughput, and delivers $7.5\times$ higher energy efficiency than prior 3DGS accelerators.

Multimodal Large Reasoning Models (MLRMs) have achieved remarkable strides in visual reasoning through test time compute scaling, yet long chain reasoning remains prone to hallucinations. We identify a concerning phenomenon termed the Reasoning Vision Truth Disconnect (RVTD): hallucinations are strongly correlated with cognitive bifurcation points that often exhibit high entropy states. We attribute this vulnerability to a breakdown in visual semantic anchoring, localized within the network's intermediate layers; specifically, during these high uncertainty transitions, the model fails to query visual evidence, reverting instead to language priors. Consequently, we advocate a shift from solely outcome level supervision to augmenting it with fine grained internal attention guidance. To this end, we propose V-STAR (Visual Structural Training with Attention Reinforcement), a lightweight, holistic training paradigm designed to internalize visually aware reasoning capabilities. Central to our approach is the Hierarchical Visual Attention Reward (HVAR), integrated within the GRPO framework. Upon detecting high entropy states, this mechanism dynamically incentivizes visual attention across critical intermediate layers, thereby anchoring the reasoning process back to the visual input. Furthermore, we introduce the Forced Reflection Mechanism (FRM), a trajectory editing strategy that disrupts cognitive inertia by triggering reflection around high entropy cognitive bifurcation points and encouraging verification of subsequent steps against the visual input, thereby translating external debiasing interventions into an intrinsic capability for hallucination mitigation.

Language models can be used to identify relationships between financial entities in text. However, while structured output mechanisms exist, prompting-based pipelines still incur autoregressive decoding costs and decouple graph construction from downstream optimization. We propose \emph{Relational Probing}, which replaces the standard language-model head with a relation head that induces a relational graph directly from language-model hidden states and is trained jointly with the downstream task model for stock-trend prediction. This approach both learns semantic representations and preserves the strict structure of the induced relational graph. It enables language-model outputs to go beyond text, allowing them to be reshaped into task-specific formats for downstream models. To enhance reproducibility, we provide an operational definition of small language models (SLMs): models that can be fine-tuned end-to-end on a single 24GB GPU under specified batch-size and sequence-length settings. Experiments use Qwen3 backbones (0.6B/1.7B/4B) as upstream SLMs and compare against a co-occurrence baseline. Relational Probing yields consistent performance improvements at competitive inference cost.

Asymmetric Tensor PCA (ATPCA) is a prototypical model for studying the trade-offs between sample complexity, computation, and memory. Existing algorithms for this problem typically require at least $d^{\left\lceil\overline{k}/2\right\rceil}$ state memory cost to recover the signal, where $d$ is the vector dimension and $\overline{k}$ is the tensor order. We focus on the setting where $\overline{k} \geq 4$ is even and consider (stochastic) gradient descent-based algorithms under a limited memory budget, which permits only mild over-parameterization of the model. We propose a matrix-parameterized method (in $d^{2}$ state memory cost) using a novel three-phase alternating-update algorithm to address the problem and demonstrate how mild over-parameterization facilitates learning in two key aspects: (i) it improves sample efficiency, allowing our method to achieve \emph{near-optimal} $d^{\overline{k}-2}$ sample complexity in our limited memory setting; and (ii) it enhances adaptivity to problem structure, a previously unrecognized phenomenon, where the required sample size naturally decreases as consecutive vectors become more aligned, and in the symmetric limit attains $d^{\overline{k}/2}$, matching the \emph{best} known polynomial-time complexity. To our knowledge, this is the \emph{first} tractable algorithm for ATPCA with $d^{\overline{k}}$-independent memory costs.

Neural networks (NNs) are central to modern machine learning and achieve state-of-the-art results in many applications. However, the relationship between loss geometry and generalization is still not well understood. The local geometry of the loss function near a critical point is well-approximated by its quadratic form, obtained through a second-order Taylor expansion. The coefficients of the quadratic term correspond to the Hessian matrix, whose eigenspectrum allows us to evaluate the sharpness of the loss at the critical point. Extensive research suggests flat critical points generalize better, while sharp ones lead to higher generalization error. However, sharpness requires the Hessian eigenspectrum, but general matrix characteristic equations have no closed-form solution. Therefore, most existing studies on evaluating loss sharpness rely on numerical approximation methods. Existing closed-form analyses of the eigenspectrum are primarily limited to simplified architectures, such as linear or ReLU-activated networks; consequently, theoretical analysis of smooth nonlinear multilayer neural networks remains limited. Against this background, this study focuses on nonlinear, smooth multilayer neural networks and derives a closed-form upper bound for the maximum eigenvalue of the Hessian with respect to the cross-entropy loss by leveraging the Wolkowicz-Styan bound. Specifically, the derived upper bound is expressed as a function of the affine transformation parameters, hidden layer dimensions, and the degree of orthogonality among the training samples. The primary contribution of this paper is an analytical characterization of loss sharpness in smooth nonlinear multilayer neural networks via a closed-form expression, avoiding explicit numerical eigenspectrum computation. We hope that this work provides a small yet meaningful step toward unraveling the mysteries of deep learning.

As Vision-Language Models (VLMs) become integral to educational decision-making, ensuring their fairness is paramount. However, current text-centric evaluations neglect the visual modality, leaving an unregulated channel for latent social biases. To bridge this gap, we present Edu-MMBias, a systematic auditing framework grounded in the tri-component model of attitudes from social psychology. This framework diagnoses bias across three hierarchical dimensions: cognitive, affective, and behavioral. Utilizing a specialized generative pipeline that incorporates a self-correct mechanism and human-in-the-loop verification, we synthesize contamination-resistant student profiles to conduct a holistic stress test on state-of-the-art VLMs. Our extensive audit reveals critical, counter-intuitive patterns: models exhibit a compensatory class bias favoring lower-status narratives while simultaneously harboring deep-seated health and racial stereotypes. Crucially, we find that visual inputs act as a safety backdoor, triggering a resurgence of biases that bypass text-based alignment safeguards and revealing a systematic misalignment between latent cognition and final decision-making. The contributions of this paper are available at: https://anonymous.4open.science/r/EduMMBias-63B2.

Investigating the impact of fatigue on human physiological function and motor behavior is crucial for developing biomechanics and medical applications aimed at mitigating fatigue, reducing injury risk, and creating sophisticated ergonomic designs, as well as for producing physically-plausible 3D animation sequences. While the former has a prominent position in state-of-the-art literature, fatigue-driven motion generation is still an underexplored area. In this study, we present FatigueFusion, a deep-learning architecture for the fusion of fatigue features within a latent representation space, enabling the creation of a variation of novel fatigued movements, intermediate fatigued states, and progressively fatigued motions. Unlike existing approaches that focus on imitating the effects of fatigue accumulation in motion patterns, our framework incorporates algorithmic and data-driven modules to impose subject-specific temporal and spatial fatigue features on nonfatigued motions, while leveraging PINN-based techniques to simulate fatigue intensity. Since all motion modulation tasks are taking place in latent space, FatigueFusion offers an end-to-end architecture that operates directly on non-fatigued joint angle sequences and control parameters, allowing seamless integration into any motion synthesis pipeline, without relying on fatigue input data. Overall, our framework can be employed for various fatigue-driven synthesis tasks, such as fatigue profile transfer and fusion, while it also provides a solution for accurate rendering of the human fatigue state in both animation and simulation pipelines.

Large Language Models (LLMs) can generate factually inaccurate content even if they have corresponding knowledge, which critically undermines their reliability. Existing approaches attempt to mitigate this by incorporating uncertainty in QA prompt during training, but these numerical scores lack the semantic richness for LLM to properly understand its internal states of trustworthiness and honestness, leading to insufficient factuality alignment. We introduce FAITH (Factuality Alignment through Integrating Trustworthiness and Honestness), a post-training framework for factuality alignment that integrates natural-language uncertainty signals with external knowledge. Specifically, we augment training datasets by computing confidence scores and semantic entropy from LLM outputs and mapping them into a knowledge state quadrant that describes the model's internal knowledge possession (trustworthiness) and answering behaviors (honestness) in natural language. Based on this enhanced data, we design a reward function that considers both correctness and uncertainty signals, and fine-tune the LLM using the Proximal Policy Optimization (PPO) algorithm. To further mitigate weakly grounded responses, we design a retrieval-augmented module that retrieves relevant external passages, improving the consistency between internal and external knowledge representations. Extensive experiments on four knowledge-intensive benchmarks demonstrate that FAITH enhances the factual accuracy and truthfulness of LLMs.

The rapid adoption of Large Language Models (LLMs) has made GPU inference efficiency an increasingly critical system concern. The runtime of LLM workloads is largely dominated by tile-based kernels, particularly General Matrix Multiplications (GEMMs). Although these kernels are highly optimized, their performance remains sensitive to a large space of runtime parameters, such as tile sizes and pipeline stages. The interaction between these parameters and hardware resources leads to a non-convex optimization landscape. Existing approaches to parameter configuration -- including search-based auto-tuning, heuristic rules, and learned cost models -- face a fundamental trade-off between performance optimality and runtime efficiency. In this paper, we present WaveTune, a wave-aware framework for runtime kernel auto-tuning. First, we introduce a unified mapping method to handle input diversity and decompose the configuration space to manage high dimensionality. Second, we develop an analytical wave-aware bilinear model that accurately predicts kernel latency. Third, we design a sparse sampling scheme based on wave structures and a lightweight dual-table retrieval mechanism to minimize runtime overhead. As a result, WaveTune enables precise and efficient runtime configuration for GPU kernels. Across three representative kernels and five GPU architectures, WaveTune consistently achieves near-optimal kernel performance, delivering up to 1.83x kernel-level speedup and up to 1.33x end-to-end TTFT reduction, while reducing runtime decision overhead by five orders of magnitude compared to exhaustive search. These results demonstrate that WaveTune effectively eliminates the traditional trade-off between configuration latency and execution optimality, providing a practical and robust solution for high-performance LLM inference.

The emergence of programmable switches has brought in-network computing (INC) into the spotlight in recent years. By offloading computation directly onto the data transmission process, INC improves network utilization, reduces latency to sub-RTT levels, saves link bandwidth, and maintains throughput. However, INC disrupts the transparency of traditional networks, forcing developers to consider network exceptions like packet loss and out-of-order. If not properly handled, these exceptions can lead to violations of application properties, such as cache consistency and lock exclusion. Usual testing cannot exhaustively cover these exceptions, raising doubts about the correctness of INC systems and hindering their deployment in the industry. This paper presents INCGuard, the first general-purpose tool for verifying INC systems. INCGuard provides a high-level specification language and saves developers 67.2% lines of code on average. To help better understand the behavior of the system, INCGuard offers configurable network environments. INCGuard enables developers to express INC-specific correctness properties. INCGuard translates developer-specified systems into state transition representations, performs model checking to detect potential design risks, and reports violation traces to developers. We propose optimizations for INC-specific scenarios to address the challenge of state space explosion. We modeled seven INC systems and identified their risks with INCGuard in seconds. We further reproduce them in real systems to confirm the validity of our verification result.

Wireless sensing, traditionally relying on signal processing (SP) techniques, has recently shifted toward data-driven deep learning (DL) to achieve performance breakthroughs. However, existing deep wireless sensing models are typically end-to-end and task-specific, lacking reusability and interpretability. We propose RF-LEGO, a modular co-design framework that transforms interpretable SP algorithms into trainable, physics-grounded DL modules through deep unrolling. By replacing hand-tuned parameters with learnable ones while preserving core processing structures and mathematical operators, RF-LEGO ensures modularity, cascadability, and structure-aligned interpretability. Specifically, we introduce three deep-unrolled modules for critical RF sensing tasks: frequency transform, spatial angle estimation, and signal detection. Extensive experiments using real-world data for Wi-Fi, millimeter-wave, UWB, and 6G sensing demonstrate that RF-LEGO significantly outperforms existing SP and DL baselines, both standalone and when integrated into multiple downstream tasks. RF-LEGO pioneers a novel SP-DL co-design paradigm for wireless sensing via deep unrolling, shedding light on efficient and interpretable deep wireless sensing solutions. Our code is available at https://github.com/aiot-lab/RF-LEGO.

Current evaluations of autonomous coding agents assume an unrealistic, infinite-resource environment. However, real-world software engineering is a resource-bound competition. As we scale toward large agent swarms, ignoring compute and time costs risks catastrophic budget exhaustion. To shift the focus from isolated accuracy to cost-aware problem-solving, we introduce USACOArena, an interactive ACM-ICPC-style arena driven by a strict "credit" economy. Every generated token, local test, and elapsed second depletes a fixed budget, forcing agents to make strategic trade-offs. Our comprehensive profiling reveals that frontier single agents and swarms currently fail to optimally balance accuracy with these constraints, exhibiting divergent, path-dependent behaviors. Ultimately, USACOArena provides an essential dynamic training ground for developing highly efficient, resource-aware agent architectures.

Disaggregation maps parts of an AI workload to different types of GPUs, offering a path to utilize modern heterogeneous GPU clusters. However, existing solutions operate at a coarse granularity and are tightly coupled to specific model architectures, leaving much room for performance improvement. This paper presents Tessera, the first kernel disaggregation system to improve performance and cost efficiency on heterogeneous GPUs for large model inference. Our key insight is that kernels within a single application exhibit diverse resource demands, making them the most suitable granularity for aligning computation with hardware capabilities. Tessera integrates offline analysis with online adaptation by extracting precise inter-kernel dependencies from PTX to ensure correctness, overlapping communication with computation through a pipelined execution model, and employing workload-aware scheduling with lightweight runtime adaptation. Extensive evaluations across five heterogeneous GPUs and four model architectures, scaling up to 16 GPUs, show that Tessera improves serving throughput and cost efficiency by up to 2.3x and 1.6x, respectively, compared to existing disaggregation methods, while generalizing to model architectures where prior approaches do not apply. Surprisingly, a heterogeneous GPU pair under Tessera can even exceed the throughput of two homogeneous high-end GPUs at a lower cost.

Byzantine-robust distributed optimization relies on robust aggregation rules to mitigate the influence of malicious Byzantine workers. Despite the proliferation of such rules, a unified convergence analysis framework that accommodates general data heterogeneity is lacking. In this work, we provide a thorough convergence theory of Byzantine-robust distributed stochastic gradient descent (SGD), analyzing variants both with and without local momentum. We establish the convergence rates for nonconvex smooth objectives and those satisfying the Polyak-Lojasiewicz condition under a general data heterogeneity assumption. Our analysis reveals that while stochasticity and data heterogeneity introduce unavoidable error floors, local momentum provably reduces the error component induced by stochasticity. Furthermore, we derive matching lower bounds to demonstrate that the upper bounds obtained in our analysis are tight and characterize the fundamental limits of Byzantine resilience under stochasticity and data heterogeneity. Empirical results support our theoretical findings.

We study the piecewise-stationary restless multi-armed bandit (PS-RMAB) problem, where each arm evolves as a Markov chain but \emph{mean rewards may change across unknown segments}. To address the resulting exploration--detection delay trade-off, we propose a modular framework that integrates arbitrary RMAB base algorithms with change detection and a novel diminishing exploration mechanism. This design enables flexible plug-and-play use of existing solvers and detectors, while efficiently adapting to mean changes without prior knowledge of their number. To evaluate performance, we introduce a refined regret notion that measures the \emph{excess regret due to exploration and detection}, benchmarked against an oracle that restarts the base algorithm at the true change points. Under this metric, we prove a regret bound of $\tilde{O}(\sqrt{LMKT})$, where $L$ denotes the maximum mixing time of the Markov chains across all arms and segments, $M$ the number of segments, $K$ the number of arms, and $T$ the horizon. Simulations confirm that our framework achieves regret close to that of the segment oracle and consistently outperforms base solvers that do not incorporate any mechanism to handle environmental changes.

Toxicity and harassment are widespread in the video-gaming context. Especially in competitive online multiplayer scenarios, gamers oftentimes send harmful messages to other players (teammates or opponents) whose consequences span from mild annoyance to withdrawal and depression. Abundant prior work tackled these problems, e.g., pointing out the negative effects of toxic interactions. However, few works proposed countermeasures specifically developed and tested on textual messages sent during a match -- i.e., when the "harassment" actually occurs. We posit that such a scarcity stems from the lack of high-quality datasets that can be used to devise "automated" detectors based on natural-language processing (NLP) and machine learning (ML), and which can -- ideally -- mitigate the harm of toxic comments during a gaming session. This work provides a foundation for addressing the problem of toxicity and harassment in video games. First, through a systematic literature review (n=1,039), we provide evidence that only few works proposed ML/NLP-based detectors of toxicity/harassment during live matches. Then, we partner-up with 8 expert League of Legend (LoL) players and create a fine-grained labelled dataset, L2DTnH, containing 1.4k toxic and 13.8k non-toxic messages exchanged during LoL matches. We use L2DTnH to develop a detector that we then empirically show outperforms general-purpose and state-of-the-art toxicity detectors reliant on NLP. To further demonstrate the practicality of our resources, we test our detector on game-related data beyond that included in L2DTnH; and we develop a Web-browser extension that flags toxic content in Webpages -- without querying third-party servers owned by AI companies. We publicly release all of our resources. Our contributions pave the way for more applied research devoted to fighting the spread of toxicity and harassment in video games.

Magnetic point-group analysis classifies altermagnets but returns only a binary symmetry verdict, leaving spin-splitting energy (SSE) inaccessible without spin-polarized density functional theory (DFT). This binary ceiling is not fundamental. Sublattice symmetry breaking is promoted here to a continuous, DFT-free scalar -- the Motif Symmetry-Breaking Index (MSBI) -- that quantifies $\mathcal{PT}$-symmetry breaking between antiparallel magnetic motifs directly from crystal coordinates. SHAP analysis of an XGBoost surrogate trained on 3,851 DFT-labeled binary structures identifies three dominant descriptors: MSBI (symmetry-breaking axis), motif packing fraction MPF (superexchange axis), and the $p/d$ electron ratio (covalency axis), each mapping onto a directly tunable experimental handle. A controlled VO--CrSb comparison within the same P$6_3$/mmc host lattice demonstrates that composition alone boosts SSE sevenfold. Bayesian optimization over this three-axis space, followed by independent DFT validation, recovers $α$-NiS (SSE $= 0.823$\,eV) as cross-validation against an independent symmetry-based prediction and identifies three previously unrecognized high-SSE candidates -- square-planar FeS (1.297\,eV), octahedral CoS (1.103\,eV), and FeAs (1.089\,eV) -- all matching or exceeding CrSb. Square-planar Fe--S is proposed as a transferable coordination motif for giant altermagnetic spin splitting, advancing altermagnet design from symmetry classification to continuous quantitative optimization.

This manuscript presents PoreDiT, a novel generative model designed for high-efficiency digital rock reconstruction at gigavoxel scales. Addressing the significant challenges in digital rock physics (DRP), particularly the trade-off between resolution and field-of-view (FOV), and the computational bottlenecks associated with traditional deep learning architectures, PoreDiT leverages a three-dimensional (3D) Swin Transformer to break through these limitations. By directly predicting the binary probability field of pore spaces instead of grayscale intensities, the model preserves key topological features critical for pore-scale fluid flow and transport simulations. This approach enhances computational efficiency, enabling the generation of ultra-large-scale ($1024^3$ voxels) digital rock samples on consumer-grade hardware. Furthermore, PoreDiT achieves physical fidelity comparable to previous state-of-the-art methods, including accurate porosity, pore-scale permeability, and Euler characteristics. The model's ability to scale efficiently opens new avenues for large-domain hydrodynamic simulations and provides practical solutions for researchers in pore-scale fluid mechanics, reservoir characterization, and carbon sequestration.

Trajectory prediction remains a critical yet challenging component in autonomous driving systems, requiring sophisticated reasoning capabilities while meeting strict real-time deployment constraints. While knowledge distillation has demonstrated effectiveness in model compression, existing approaches often fail to preserve complex decision-making capabilities, particularly in dynamic multi-agent scenarios. This paper introduces MAVEN-T, a teacher-student framework that achieves state-of-the-art trajectory prediction through complementary architectural co-design and progressive distillation. The teacher employs hybrid attention mechanisms for maximum representational capacity, while the student uses efficient architectures optimized for deployment. Knowledge transfer is performed via multi-granular distillation with adaptive curriculum learning that dynamically adjusts complexity based on performance. Importantly, the framework incorporates reinforcement learning to overcome the imitation ceiling of traditional distillation, enabling the student to verify, refine, and optimize teacher knowledge through dynamic environmental interaction, potentially achieving more robust decision-making than the teacher itself. Extensive experiments on NGSIM and highD datasets demonstrate 6.2x parameter compression and 3.7x inference speedup while maintaining state-of-the-art accuracy, establishing a new paradigm for deploying sophisticated reasoning models under resource constraints.

Multi-vector models dominate Visual Document Retrieval (VDR) due to their fine-grained matching capabilities, but their high storage and computational costs present a major barrier to practical deployment. In this paper, we propose ColChunk, a plug-and-play framework that introduces multimodal late chunking to construct efficient, contextualized multi-vectors. Unlike existing pruning or fixed-token approaches, ColChunk employs hierarchical clustering on patch-level embeddings, fused with a 2D position prior to ensure spatial-semantic coherence. This adaptive grouping allows for a content-aware representation that preserves global context while drastically reducing the vector count. Evaluations across 24 VDR datasets demonstrate ColChunk achieves over a 90% reduction in storage requirements while simultaneously delivering a 9-point average improvement in nDCG@5 across representative single-vector models. ColChunk provides a practical solution for balancing retrieval accuracy and efficiency in visual document systems.

Intelligent operation of thermal energy networks aims to improve energy efficiency, reliability, and operational flexibility through data-driven control, predictive optimization, and early fault detection. Achieving these goals relies on sufficient observability, requiring continuous and well-distributed monitoring of thermal and hydraulic states. However, district heating systems are typically sparsely instrumented and frequently affected by sensor faults, limiting monitoring. Virtual sensing offers a cost-effective means to enhance observability, yet its development and validation remain limited in practice. Existing data-driven methods generally assume dense synchronized data, while analytical models rely on simplified hydraulic and thermal assumptions that may not adequately capture the behavior of heterogeneous network topologies. Consequently, modeling the coupled nonlinear dependencies between pressure, flow, and temperature under realistic operating conditions remains challenging. In addition, the lack of publicly available benchmark datasets hinders systematic comparison of virtual sensing approaches. To address these challenges, we propose a heterogeneous spatial-temporal graph neural network (HSTGNN) for constructing virtual smart heat meters. The model incorporates the functional relationships inherent in district heating networks and employs dedicated branches to learn graph structures and temporal dynamics for flow, temperature, and pressure measurements, thereby enabling the joint modeling of cross-variable and spatial correlations. To support further research, we introduce a controlled laboratory dataset collected at the Aalborg Smart Water Infrastructure Laboratory, providing synchronized high-resolution measurements representative of real operating conditions. Extensive experiments demonstrate that the proposed approach significantly outperforms existing baselines.

Reasoning on Temporal Knowledge Graphs (TKGs) is essential for predicting future events and time-aware facts. While existing methods are effective at capturing relational dynamics, their performance is limited by a closed-world assumption, which fails to account for emerging entities not present in the training. Notably, these entities continuously join the network without historical interactions. Empirical study reveals that emerging entities are widespread in TKGs, comprising roughly 25\% of all entities. The absence of historical interactions of these entities leads to significant performance degradation in reasoning tasks. Whereas, we observe that entities with semantic similarities often exhibit comparable interaction histories, suggesting the presence of transferable temporal patterns. Inspired by this insight, we propose TransFIR (Transferable Inductive Reasoning), a novel framework that leverages historical interaction sequences from semantically similar known entities to support inductive reasoning. Specifically, we propose a codebook-based classifier that categorizes emerging entities into latent semantic clusters, allowing them to adopt reasoning patterns from similar entities. Experimental results demonstrate that TransFIR outperforms all baselines in reasoning on emerging entities, achieving an average improvement of 28.6% in Mean Reciprocal Rank (MRR) across multiple datasets. The implementations are available at https://github.com/zhaodazhuang2333/TransFIR.

The advancement of large language models (LLMs) has enhanced tabular question answering (Tabular QA), yet they struggle with open-domain queries exhibiting underspecified or uncertain expressions. To address this, we introduce the ODUTQA-MDC task and the first comprehensive benchmark to tackle it. This benchmark includes: (1) a large-scale ODUTQA dataset with 209 tables and 25,105 QA pairs; (2) a fine-grained labeling scheme for detailed evaluation; and (3) a dynamic clarification interface that simulates user feedback for interactive assessment. We also propose MAIC-TQA, a multi-agent framework that excels at detecting ambiguities, clarifying them through dialogue, and refining answers. Experiments validate our benchmark and framework, establishing them as a key resource for advancing conversational, underspecification-aware Tabular QA research.

While modern transformer neural networks achieve grandmaster-level performance in chess and other reasoning tasks, their internal computation process remains largely opaque. Focusing on Leela Chess Zero (LC0), we introduce a sparse decomposition framework to interpret its internal computation by decomposing its MLP and attention modules with sparse replacement layers, which capture the primary computation process of LC0. We conduct a detailed case study showing that these pathways expose rich, interpretable tactical considerations that are empirically verifiable. We further introduce three quantitative metrics and show that LC0 exhibits parallel reasoning behavior consistent with the inductive bias of its policy head architecture. To the best of our knowledge, this is the first work to decompose the internal computation of a transformer on both MLP and attention modules for interpretability. Combining sparse replacement layers and causal interventions in LC0 provides a comprehensive understanding of advanced tactical reasoning, offering critical insights into the underlying mechanisms of superhuman systems. Our code is available at https://github.com/JacklE0niden/Leela-SAEs.

Band gap engineering of oxide semiconductors through doping is critical for photocatalysis and optoelectronics, yet the combinatorial space of dopant elements, substitution sites, and co-doping combinations far exceeds typical density functional theory (DFT) budgets. We screen doped candidates across five oxide hosts (ZnO, TiO2, SrTiO3, SnO2, MgO), culminating in a 529-candidate ZnO co-doping campaign, and identify Cu-containing co-doped ZnO systems as consistently achieving visible-light-range band gaps (1.0-1.8 eV), with Y2Cu2 co-doped ZnO as the optimal candidate (1.84 eV). A three-tier validation funnel (PBE, PBE+U, ionic relaxation) reveals that no single level of theory suffices: V-doped ZnO shifts from near-metallic to wide-gap upon Hubbard U correction, while Cu-doped SrTiO3 enters the visible-light window only after correcting for d-electron localization. To make this screening tractable, we introduce a multi-fidelity screening strategy that replaces 81% of DFT evaluations with computationally inexpensive surrogate predictions, reducing a 529-candidate closed-loop Quantum ESPRESSO campaign from an estimated 440 to 62 CPU-hours while finding the global optimum in 100% of 50 independent trials (p = 5.0e-8 versus random screening, Wilcoxon signed-rank). Cross-host analysis of the dopant-host interaction matrix reveals that dopant performance is governed by just two latent chemical dimensions, enabling prediction of rankings in unseen hosts. All 583 DFT calculations, screening code, and stability proofs are released as an open benchmark.

The Mixture-of-Experts (MoE) architecture has emerged as a promising approach to mitigate the rising computational costs of large language models (LLMs) by selectively activating parameters. However, its high memory requirements and sub-optimal parameter efficiency pose significant challenges for efficient deployment. Although CPU-offloaded MoE inference systems have been proposed in the literature, they offer limited efficiency, particularly for large batch sizes. In this work, we propose SpecMoE, a memory-efficient MoE inference system based on our self-assisted speculative decoding algorithm. SpecMoE demonstrates the effectiveness of applying speculative decoding to MoE inference without requiring additional model training or fine-tuning. Our system improves inference throughput by up to $4.30\times$, while significantly reducing bandwidth requirements of both memory and interconnect on memory-constrained systems.

Large language models are increasingly used as writing tools and pedagogical resources in English for Academic Purposes, but it remains unclear whether they encode culturally differentiated representations when generating academic text. This study tests whether Gemma-3-4b-it encodes nationality-discriminative information in hidden states when generating research article introductions conditioned by British and Chinese academic personas. A corpus of 270 texts was generated from 45 prompt templates crossed with six persona conditions in a 2 x 3 design. Logistic regression probes were trained on hidden-state activations across all 35 layers, with shuffled-label baselines, a surface-text skyline classifier, cross-family tests, and sentence-level baselines used as controls. Probe-selected token positions were annotated for structural, lexical, and stance features using the Stanza NLP pipeline. The nationality probe reached 0.968 cross-validated accuracy at Layer 18, with perfect held-out classification. Nationality encoding followed a non-monotonic trajectory across layers, with structural effects strongest in the middle to upper network and lexical-domain effects peaking earlier. At high-signal token positions, British-associated patterns showed more postmodification, hedging, boosting, passive voice, and evaluative or process-oriented vocabulary, while Chinese-associated patterns showed more premodification, nominal predicates, and sociocultural or internationalisation vocabulary. However, sentence-level analysis found no significant nationality differences in the full generated surface text. The findings extend probing methodology to a sociolinguistic attribute and have practical implications for EAP and language pedagogy.

Generative listwise reranking leverages global context for superior retrieval but is plagued by intrinsic position bias, where models exhibit structural sensitivity to input order independent of relevance. Existing mitigations present a dilemma: inference-time aggregation incurs prohibitive latency, while training-based methods often fail to eradicate ingrained priors, particularly in compact models. To resolve this dilemma, we propose CapCal (Content-Agnostic Probability Calibration), a training-free framework that mechanically decouples positional bias from ranking decisions. By estimating the bias distribution via content-free placeholders, CapCal rectifies output logits through an entropy-adaptive contrastive mechanism. Evaluations across 10 benchmarks confirm that CapCal achieves superior performance among training-free methods while preserving single-pass efficiency. Notably, it unlocks the latent potential of lightweight models (e.g., 0.6B), delivering absolute NDCG gains exceeding 10 points and outperforming both permutation-based aggregation and data-augmentation baselines.

Graph attention networks (GATs) provide one of the best frameworks for learning node representations in relational data; but, existing variants such as Graph Attention Network (GAT) mainly operate on static graphs and rely on implicit temporal aggregation when applied to sequential data. In this paper, we introduce Electroencephalography-temporal Graph Attention Network (EEG-tGAT), a temporally augmented formulation of GATv2 that is tailored for affordance classification from interaction sequences. The proposed model incorporates temporal attention to modulate the contribution of different time segments and temporal dropout to regularize learning across temporally correlated observations. The design reflects the assumption that temporal dimensions in affordance data are not semantically uniform and that discriminative information may be unevenly distributed across time. Experimental results on affordance datasets show that EEG-tGAT achieves improved classification performance compared to GATv2. The observed gains helps to conclude that explicitly encoding temporal importance and enforcing temporal robustness introduce inductive biases that are much better aligned with the structure of affordance-driven interaction data. These findings show us that modest architectural changes to graph attention models can help one obtain consistent benefits when temporal relationships play a nontrivial role in the task.

Capturing user intent across heterogeneous behavioral domains stands as a fundamental challenge in session-based recommender systems. Yet, existing multi-domain approaches frequently fail to isolate the distinct contribution of cross-domain interactions from those arising within individual domains, limiting their ability to build rich and transferable user representations. In this work, we propose MOSAIC, a Multi-Domain Orthogonal Session Adaptive Intent Capture framework that explicitly factorizes user preferences into three orthogonal components: domain-specific, domain-common, and cross-sequence-exclusive representations. Our approach employs a triple-encoder architecture, where each encoder is dedicated to one preference type, enforced through domain masking objectives and adversarial training via a gradient reversal layer. Representational alignment and mutual independence constraints are jointly optimized to ensure clean preference separation. Additionally, a dynamic gating mechanism modulates the relative contribution of each component at every timestep, yielding a unified and temporally adaptive session-level user representation. We conduct extensive experiments on two large-scale real-world benchmarks spanning multiple domains and interaction types. The ablation study validates that each component domain-specific encoding, domain-common modeling, cross-sequence representation, and dynamic gating contributes meaningfully to the overall performance. Experimental results demonstrate that MOSAIC consistently outperforms state-of-the-art baselines in recommendation accuracy, while simultaneously providing interpretable insights into the interplay between domain-specific and cross-domain preference signals. These findings highlight the potential of orthogonal preference decomposition as a principled strategy for next-generation multi-domain recommender systems.

Multi-output Gaussian Processes provide principled uncertainty-aware learning of vector-valued fields but are difficult to deploy in large-scale, distributed, and streaming settings due to their computational and centralized nature. This paper proposes a Consensus-based Recursive Multi-Output Gaussian Process (CRMGP) framework that combines recursive inference on shared basis vectors with neighbour-to-neighbour information-consensus updates. The resulting method supports parallel, fully distributed learning with bounded per-step computation while preserving inter-output correlations and calibrated uncertainty. Experiments on synthetic wind fields and real LiDAR data demonstrate that CRMGP achieves competitive predictive performance and reliable uncertainty calibration, offering a scalable alternative to centralized Gaussian process models for multi-agent sensing applications.

Researchers have explored different ways to improve large language models (LLMs)' capabilities via dummy token insertion in contexts. However, existing works focus solely on the dummy tokens themselves, but fail to leverage the inherent sentence-level structure of natural language. This is a critical oversight, as LLMs acquire linguistic capabilities through exposure to human-generated texts, which are inherently structured at the sentence level. Motivated by this gap, we propose an approach that inserts delimiters at sentence boundaries in LLM inputs, which not only integrates dummy tokens into the context, but also facilitates LLMs with sentence-by-sentence processing behavior during reasoning. Two concrete methods: (1). In-context learning and (2). Supervised fine-tuning are experimented using 7B models to 600B Deepseek-V3. Our results demonstrate consistent improvements across various tasks, with notable gains of up to 7.7\% on GSM8k and 12.5\% on DROP. Furthermore, the fine-tuned LLMs can incorporate sentence awareness evidenced by their internal representations. Our work establishes a simple yet effective technique for enhancing LLM's capabilities, offering promising directions for cognitive-inspired LLM enhancement paradigm.

Image Manipulation Localization (IML) aims to identify edited regions in an image. However, with the increasing use of modern image editing and generative models, many manipulations no longer exhibit obvious low-level artifacts. Instead, they often involve subtle but meaning-altering edits to an object's attributes, state, or relationships while remaining highly consistent with the surrounding content. This makes conventional IML methods less effective because they mainly rely on artifact detection rather than semantic sensitivity. To address this issue, we introduce Semantic Manipulation Localization (SML), a new task that focuses on localizing subtle semantic edits that significantly change image interpretation. We further construct a dedicated fine-grained benchmark for SML using a semantics-driven manipulation pipeline with pixel-level annotations. Based on this task, we propose TRACE (Targeted Reasoning of Attributed Cognitive Edits), an end-to-end framework that models semantic sensitivity through three progressively coupled components: semantic anchoring, semantic perturbation sensing, and semantic-constrained reasoning. Specifically, TRACE first identifies semantically meaningful regions that support image understanding, then injects perturbation-sensitive frequency cues to capture subtle edits under strong visual consistency, and finally verifies candidate regions through joint reasoning over semantic content and semantic scope. Extensive experiments show that TRACE consistently outperforms existing IML methods on our benchmark and produces more complete, compact, and semantically coherent localization results. These results demonstrate the necessity of moving beyond artifact-based localization and provide a new direction for image forensics in complex semantic editing scenarios.

The rapid advancement of AIGC-based video generation has underscored the critical need for comprehensive evaluation frameworks that go beyond traditional generation quality metrics to encompass aesthetic appeal. However, existing benchmarks remain largely focused on technical fidelity, leaving a significant gap in holistic assessment-particularly with respect to perceptual and artistic qualities. To address this limitation, we introduce VGA-Bench, a unified benchmark for joint evaluation of video generation quality and aesthetic quality. VGA-Bench is built upon a principled three-tier taxonomy: Aesthetic Quality, Aesthetic Tagging, and Generation Quality, each decomposed into multiple fine-grained sub-dimensions to enable systematic assessment. Guided by this taxonomy, we design 1,016 diverse prompts and generate a large-scale dataset of over 60,000 videos using 12 video generation models, ensuring broad coverage across content, style, and artifacts. To enable scalable and automated evaluation, we annotate a subset of the dataset via human labeling and develop three dedicated multi-task neural assessors: VAQA-Net for aesthetic quality prediction, VTag-Net for automatic aesthetic tagging, and VGQA-Net for generation and basic quality attributes. Extensive experiments demonstrate that our models achieve reliable alignment with human judgments, offering both accuracy and efficiency. We release VGA-Bench as a public benchmark to foster research in AIGC evaluation, with applications in content moderation, model debugging, and generative model optimization.

Metamorphic testing (MT) is a widely recognized technique for alleviating the oracle problem in software testing. However, its adoption is hindered by the difficulty of constructing effective metamorphic relations (MRs), which often require domain-specific or hard-to-obtain knowledge. In this work, we propose a novel approach that leverages the functional coupling between methods, which is readily available in source code, to automatically construct MRs and generate metamorphic test cases (MTCs). Our technique, MR-Coupler, identifies functionally coupled method pairs, employs large language models to generate candidate MTCs, and validates them through test amplification and mutation analysis. In particular, we leverage three functional coupling features to avoid expensive enumeration of possible method pairs, and a novel validation mechanism to reduce false alarms. Our evaluation of MR-Coupler on 100 human-written MTCs and 50 real-world bugs shows that it generates valid MTCs for over 90% of tasks, improves valid MTC generation by 64.90%, and reduces false alarms by 36.56% compared to baselines. Furthermore, the MTCs generated by MR-Coupler detect 44% of the real bugs. Our results highlight the effectiveness of leveraging functional coupling for automated MR construction and the potential of MR-Coupler to facilitate the adoption of MT in practice. We also released the tool and experimental data to support future research.

Dysarthric speech severity assessment typically requires trained clinicians or supervised models built from labelled pathological speech, limiting scalability across languages and clinical settings. We present a training-free method that quantifies dysarthria severity by measuring degradation in phonological feature subspaces within frozen HuBERT representations. No supervised severity model is trained; feature directions are estimated from healthy control speech using a pretrained forced aligner. For each speaker, we extract phone-level embeddings via Montreal Forced Aligner, compute d-prime scores along phonological contrast directions (nasality, voicing, stridency, sonorance, manner, and four vowel features) derived exclusively from healthy controls, and construct a 12-dimensional phonological profile.Evaluating 890 speakers across 10 corpora, 5 languages (English, Spanish, Dutch, Mandarin, French), and 3 primary aetiologies (Parkinson's disease, cerebral palsy, ALS), we find that all five consonant d-prime features correlate significantly with clinical severity (random-effects meta-analysis rho = -0.50 to -0.56, p < 2e-4; pooled Spearman rho = -0.47 to -0.55 with bootstrap 95% CIs not crossing zero). The effect replicates within individual corpora, survives FDR correction, and remains robust to leave-one-corpus-out removal and alignment quality controls. Nasality d-prime decreases monotonically from control to severe in 6 of 7 severity-graded corpora. Mann-Whitney U tests confirm that all 12 features distinguish controls from severely dysarthric speakers (p < 0.001).The method requires no dysarthric training data and applies to any language with an existing MFA acoustic model (currently 29 languages). We release the full pipeline and phone feature configurations for six languages.

Photoplethysmography (PPG)-based blood pressure (BP) estimation is a challenging task, particularly on resource-constrained wearable devices. However, fully on-board processing is desirable to ensure user data confidentiality. Recent deep neural networks (DNNs) have achieved high BP estimation accuracy by reconstructing BP waveforms or directly regressing BP values, but their large memory, computation, and energy requirements hinder deployment on wearables. This work introduces a fully automated DNN design pipeline that combines hardware-aware neural architecture search (NAS), pruning, and mixed-precision search (MPS) to generate accurate yet compact BP prediction models optimized for ultra-low-power multicore systems-on-chip (SoCs). Starting from state-of-the-art baseline models on four public datasets, our optimized networks achieve up to 7.99% lower error with a 7.5x parameter reduction, or up to 83x fewer parameters with negligible accuracy loss. All models fit within 512 kB of memory on our target SoC (GreenWaves' GAP8), requiring less than 55 kB and achieving an average inference latency of 142 ms and energy consumption of 7.25 mJ. Patient-specific fine-tuning further improves accuracy by up to 64%, enabling fully autonomous, low-cost BP monitoring on wearables.

Major depressive disorder (MDD) is a prevalent mental disorder associated with complex neurobiological changes that cannot be fully captured using a single imaging modality. The use of multimodal magnetic resonance imaging (MRI) provides a more comprehensive understanding of brain changes by combining structural and functional data. Despite this, the effective integration of these modalities remains challenging. In this study, we propose a dual cross-attention-based multimodal fusion framework that explicitly models bidirectional interactions between structural MRI (sMRI) and resting-state functional MRI (rs-fMRI) representations. The proposed approach is tested on the large-scale REST-meta-MDD dataset using both structural and functional brain atlas configurations. Numerous experiments conducted under a 10-fold stratified cross-validation demonstrated that the proposed fusion algorithm achieves robust and competitive performance across all atlas types. The proposed method consistently outperforms conventional feature-level concatenation for functional atlases, while maintaining comparable performance for structural atlases. The most effective dual cross-attention multimodal model obtained 84.71% accuracy, 86.42% sensitivity, 82.89% specificity, 84.34% precision, and 85.37% F1-score. These findings emphasize the importance of explicitly modeling cross-modal interactions for multimodal neuroimaging-based MDD classification.

The generation of high-fidelity synthetic data is a cornerstone of modern machine learning, yet Large Language Models (LLMs) frequently suffer from hallucinations, logical inconsistencies, and mode collapse when tasked with structured generation. Existing approaches, such as prompting or retrieval-augmented generation, lack the mechanisms to balance linguistic expressivity with formal guarantees regarding validity and coverage. To address this, we propose CircuitSynth, a novel neuro-symbolic framework that decouples semantic reasoning from surface realization. By distilling the reasoning capabilities of a Teacher LLM into a Probabilistic Sentential Decision Diagram (PSDD), CircuitSynth creates a tractable semantic prior that structurally enforces hard logical constraints. Furthermore, we introduce a convex optimization mechanism to rigorously satisfy soft distributional goals. Empirical evaluations across diverse benchmarks demonstrate that CircuitSynth achieves 100% Schema Validity even in complex logic puzzles where unconstrained baselines fail (12.4%) while significantly outperforming state-of-the-art methods in rare-combination coverage.

Graph Convolutional Networks (GCNs) are widely adopted for tasks involving relational or graph-structured data and can be formulated as two-stage sparse-dense matrix multiplication (SpMM) during inference. However, existing accelerators often struggle with the irregular workloads induced by power-law node degree distributions. In this work, we propose FlexVector, a vector-processor-based architecture that efficiently accelerates SpMM for GCN inference. To address irregular computation patterns, FlexVector adopts a row-wise, product-based dataflow that regularizes SpMM execution and exposes vector parallelism through full-row access to vector registers, eliminating the need for multi-banked register file designs. Building on this dataflow, it introduces software-managed, flexible vector register files (VRFs) that adapt to irregular data access patterns, without sacrificing memory access efficiency. To further exploit these architectural capabilities, we develop a graph-aware preprocessing and node partitioning strategy that restructures irregular graph workloads to better match the row-wise dataflow and VRF capacity. This hardware-software co-design reduces memory traffic, leading to significant performance and energy efficiency gains on real-world GCN workloads. Experimental results on five real-world GCN datasets show that the VRF-centric FlexVector achieves a 3.78x speedup and 40.5% lower energy at comparable area cost relative to a state-of-the-art cache-centric baseline with buffers of the same size.

Large Language Models (LLMs) have become a key foundation for enabling personalized smart home experiences. While existing studies have explored how smart home assistants understand user queries to control devices in real time, their ability to perform memory-driven device control remains challenging from both evaluation and methodological perspectives. In terms of evaluation, existing benchmarks either focus on immediate device control or general open-domain memory retrieval tasks, and therefore cannot effectively evaluate a model's ability to perform memory-driven device control. Methodologically, while memory-driven device control can be approached using Reinforcement Learning, conventional RL methods generally rely on outcome-based supervision (i.e., whether the final task is achieved). This lack of intermediate feedback can lead to sub-optimal performance or local failures in fine-grained memory management tasks (adding, updating, deleting, and utilizing). To address these issues, we first release MemHomeLife, built from anonymized real-world long-term user interaction logs. To enable more fine-grained evaluation of different memory-related subtasks, we further construct MemHome, the first benchmark designed to systematically evaluate memory-driven device control in smart home scenarios.

AI-generated image detectors suffer significant performance degradation under real-world image corruptions such as JPEG compression, Gaussian blur, and resolution downsampling. We observe that state-of-the-art methods, including B-Free, treat degradation robustness as a byproduct of data augmentation rather than an explicit training objective. In this work, we propose Degradation-Consistent Paired Training (DCPT), a simple yet effective training strategy that explicitly enforces robustness through paired consistency constraints. For each training image, we construct a clean view and a degraded view, then impose two constraints: a feature consistency loss that minimizes the cosine distance between clean and degraded representations, and a prediction consistency loss based on symmetric KL divergence that aligns output distributions across views. DCPT adds zero additional parameters and zero inference overhead. Experiments on the Synthbuster benchmark (9 generators, 8 degradation conditions) demonstrate that DCPT improves the degraded-condition average accuracy by 9.1 percentage points compared to an identical baseline without paired training, while sacrificing only 0.9% clean accuracy. The improvement is most pronounced under JPEG compression (+15.7% to +17.9%). Ablation further reveals that adding architectural components leads to overfitting on limited training data, confirming that training objective improvement is more effective than architectural augmentation for degradation robustness.

The rapid development of large language models (LLMs) has extended text generation tasks into the literary domain. However, AI-generated literary creations has raised increasingly prominent issues of creative authenticity and ethics in literary world, making the detection of LLM-generated literary texts essential and urgent. While previous works have made significant progress in detecting AI-generated text, it has yet to address classical Chinese poetry. Due to the unique linguistic features of classical Chinese poetry, such as strict metrical regularity, a shared system of poetic imagery, and flexible syntax, distinguishing whether a poem is authored by AI presents a substantial challenge. To address these issues, we introduce ChangAn, a benchmark for detecting LLM-generated classical Chinese poetry that containing total 30,664 poems, 10,276 are human-written poems and 20,388 poems are generated by four popular LLMs. Based on ChangAn, we conducted a systematic evaluation of 12 AI detectors, investigating their performance variations across different text granularities and generation strategies. Our findings highlight the limitations of current Chinese text detectors, which fail to serve as reliable tools for detecting LLM-generated classical Chinese poetry. These results validate the effectiveness and necessity of our proposed ChangAn benchmark. Our dataset and code are available at https://github.com/VelikayaScarlet/ChangAn.

As the foundational architecture of modern machine learning, Transformers have driven remarkable progress across diverse AI domains. Despite their transformative impact, a persistent challenge across various Transformers is Attention Sink (AS), in which a disproportionate amount of attention is focused on a small subset of specific yet uninformative tokens. AS complicates interpretability, significantly affecting the training and inference dynamics, and exacerbates issues such as hallucinations. In recent years, substantial research has been dedicated to understanding and harnessing AS. However, a comprehensive survey that systematically consolidates AS-related research and offers guidance for future advancements remains lacking. To address this gap, we present the first survey on AS, structured around three key dimensions that define the current research landscape: Fundamental Utilization, Mechanistic Interpretation, and Strategic Mitigation. Our work provides a pivotal contribution by clarifying key concepts and guiding researchers through the evolution and trends of the field. We envision this survey as a definitive resource, empowering researchers and practitioners to effectively manage AS within the current Transformer paradigm, while simultaneously inspiring innovative advancements for the next generation of Transformers. The paper list of this work is available at https://github.com/ZunhaiSu/Awesome-Attention-Sink.

Anthropogenic methane (CH4) point sources drive near-term climate forcing, safety hazards, and system inefficiencies. Space-based imaging spectroscopy is emerging as a tool for identifying emissions globally, but existing approaches largely rely on manual plume identification. Here we present the Methane Analysis and Plume Localization with EMIT (MAPL-EMIT) model, an end-to-end vision transformer framework that leverages the complete radiance spectrum from the Earth Surface Mineral Dust Source Investigation (EMIT) instrument to jointly retrieve methane enhancements across all pixels within a scene. This approach brings together spectral and spatial context to significantly lower detection limits. MAPL-EMIT simultaneously supports enhancement quantification, plume delineation, and source localization, even for multiple overlapping plumes. The model was trained on 3.6 million physics-based synthetic plumes injected into global EMIT radiance data. Synthetic evaluation confirms the model's ability to identify plumes with high recall and precision and to capture weaker plumes relative to existing matched-filter approaches. On real-world benchmarks, MAPL-EMIT captures 79% of known hand-annotated NASA L2B plume complexes across a test set of 1084 EMIT granules, while capturing twice as many plausible plumes than identified by human analysts. Further validation against coincident airborne data, top-emitting landfills, and controlled release experiments confirms the model's ability to identify previously uncaptured sources. By incorporating model-generated metrics such as spectral fit scores and estimated noise levels, the framework can further limit false-positive rates. Overall, MAPL-EMIT enables high-throughput implementation on the full EMIT catalog, shifting methane monitoring from labor-intensive workflows to a rapid, scalable paradigm for global plume mapping at the facility scale.

The growing complexity of cyber-physical systems (CPSs) calls for early prototyping tools that combine accuracy, speed, and usability. Virtual Platforms (VPs) provide fast functional simulation, but hybrid co-emulation solutions, in which key digital components are deployed on FPGA, become necessary when accurate timing modelling is required and RTL simulation is too costly. However, existing hybrid emulation tools are mostly proprietary, and rely on vendor-specific FPGA features. To address this gap, we introduce an open-source framework that connects SystemC-based VPs with FPGA emulation, enabling full-system co-emulation of digital and non-digital components. The FPGA accelerates the execution of main digital subsystems, while a wrapper coordinates timing and communication with the VP through JTAG, maintaining synchronization with simulated peripherals. Evaluations using a RISC-V SoC, with an example in the biosignals processing domain, show up to 2500x speedup compared to RTL simulation, while maintaining less than 2x total simulation time relative to pure FPGA emulation.

Large language models (LLMs) have shown remarkable performance in various domains, but they are constrained by massive computational and storage costs. Quantization, an effective technique for compressing models to fit resource-limited devices while preserving generative quality, encompasses two primary methods: quantization aware training (QAT) and post-training quantization (PTQ). QAT involves additional retraining or fine-tuning, thus inevitably resulting in high training cost and making it unsuitable for LLMs. Consequently, PTQ has become the research hotspot in recent quantization methods. However, existing PTQ methods usually rely on various complex computation procedures and suffer from considerable performance degradation under low-bit quantization settings. To alleviate the above issues, we propose a simple and effective post-training quantization paradigm for LLMs, named SEPTQ. Specifically, SEPTQ first calculates the importance score for each element in the weight matrix and determines the quantization locations in a static global manner. Then it utilizes the mask matrix which represents the important locations to quantize and update the associated weights column-by-column until the appropriate quantized weight matrix is obtained. Compared with previous methods, SEPTQ simplifies the post-training quantization procedure into only two steps, and considers the effectiveness and efficiency simultaneously. Experimental results on various datasets across a suite of models ranging from millions to billions in different quantization bit-levels demonstrate that SEPTQ significantly outperforms other strong baselines, especially in low-bit quantization scenarios.

We present EL-DRUIN, an ontological reasoning system for geopolitical intelligence analysis that combines formal ontology, finite semigroup algebra, and Lie algebra approximation to forecast long-run relationship trajectories. Current LLM-based political analysis systems operate as summarisation engines, producing outputs bounded by textual pattern matching. EL-DRUIN departs from this paradigm by modelling geopolitical relationships as states in a finite set of named Dynamic Patterns, composing patterns via a semigroup operation whose structure constants are defined by an explicit composition table, and embedding each pattern as a vector in an 8-dimensional semantic Lie algebra space. Forward simulation iterates this semigroup operation, yielding reachable pattern sets at each discrete timestep; convergence to idempotent absorbing states (fixed points of the composition) constitutes the predicted long-run attractor. Bayesian posterior weights combine ontology-derived confidence priors with a Lie similarity term measuring the cosine similarity between the vector sum of composing patterns and the target pattern vector, providing interpretable, calibrated probabilities that are not self-reported by a language model. Bifurcation points -- steps at which two candidate attractors have near-equal posterior mass -- are detected and exposed to downstream analysis. We demonstrate the framework on six geopolitical scenarios including US-China technology decoupling and the Taiwan Strait military coercion trajectory. The architecture is publicly available as an open-source system with a Streamlit frontend exposing full computation traces, Bayesian posterior breakdowns, and 8D ontological state vectors.

Real-world image pairs often exhibit both severe degradations and large viewpoint changes, making image restoration and geometric matching mutually interfering tasks when treated independently. In this work, we propose MatRes, a zero-shot test-time adaptation framework that jointly improves restoration quality and correspondence estimation using only a single low-quality and high-quality image pair. By enforcing conditional similarity at corresponding locations, MatRes updates only lightweight modules while keeping all pretrained components frozen, requiring no offline training or additional supervision. Extensive experiments across diverse combinations show that MatRes yields significant gains in both restoration and geometric alignment compared to using either restoration or matching models alone. MatRes offers a practical and widely applicable solution for real-world scenarios where users commonly capture multiple images of a scene with varying viewpoints and quality, effectively addressing the often-overlooked mutual interference between matching and restoration.

Supervised Fine-Tuning (SFT) is the standard approach for adapting large language models (LLMs) to downstream tasks. However, we observe a persistent failure mode: even after convergence, models often fail to correctly reproduce a subset of their own supervised training data. We refer to this behavior as the Incomplete Learning Phenomenon(ILP). This paper presents the first systematic study of ILP in LLM fine-tuning. We formalize ILP as post-training failure to internalize supervised instances and demonstrate its prevalence across multiple model families, domains, and datasets. Through controlled analyses, we identify five recurrent sources of incomplete learning: (1) missing prerequisite knowledge in the pre-trained model, (2) conflicts between SFT supervision and pre-training knowledge, (3) internal inconsistencies within SFT data, (4) left-side forgetting during sequential fine-tuning, and (5) insufficient optimization for rare or complex patterns. We introduce a diagnostic-first framework that maps unlearned samples to these causes using observable training and inference signals, and study several targeted mitigation strategies as causal interventions. Experiments on Qwen, LLaMA, and OLMo2 show that incomplete learning is widespread and heterogeneous, and that improvements in aggregate metrics can mask persistent unlearned subsets. The findings highlight the need for fine-grained diagnosis of what supervised fine-tuning fails to learn, and why.

Student engagement is crucial for improving learning outcomes in group activities. Highly engaged students perform better both individually and contribute to overall group success. However, most existing automated engagement recognition methods are designed for online classrooms or estimate engagement at the individual level. Addressing this gap, we propose DualEngage, a novel two-stream framework for group-level engagement recognition from in-classroom videos. It models engagement as a joint function of both individual and group-level behaviors. The primary stream models person-level motion dynamics by detecting and tracking students, extracting dense optical flow with the Recurrent All-Pairs Field Transforms network, encoding temporal motion patterns using a transformer encoder, and finally aggregating per-student representations through attention pooling into a unified representation. The secondary stream captures scene-level spatiotemporal information from the full video clip, leveraging a pretrained three-dimensional Residual Network. The two-stream representations are combined via softmax-gated fusion, which dynamically weights each stream's contribution based on the joint context of both features. DualEngage learns a joint representation of individual actions with overarching group dynamics. We evaluate the proposed approach using fivefold cross-validation on the Classroom Group Engagement Dataset developed by Ocean University of China, achieving an average classification accuracy of 0.9621+/-0.0161 with a macro-averaged F1 of 0.9530+/-0.0204. To understand the contribution of each branch, we further conduct an ablation study comparing single-stream variants against the two-stream model. This work is among the first in classroom engagement recognition to adopt a dual-stream design that explicitly leverages motion cues as an estimator.

Text-to-CAD code generation is a long-horizon task that translates textual instructions into long sequences of interdependent operations. Existing methods typically decode text directly into executable code (e.g., bpy) without explicitly modeling assembly hierarchy or geometric constraints, which enlarges the search space, accumulates local errors, and often causes cascading failures in complex assemblies. To address this issue, we propose a hierarchical and geometry-aware graph as an intermediate representation. The graph models multi-level parts and components as nodes and encodes explicit geometric constraints as edges. Instead of mapping text directly to code, our framework first predicts structure and constraints, then conditions action sequencing and code generation, thereby improving geometric fidelity and constraint satisfaction. We further introduce a structure-aware progressive curriculum learning strategy that constructs graded tasks through controlled structural edits, explores the model's capability boundary, and synthesizes boundary examples for iterative training. In addition, we build a 12K dataset with instructions, decomposition graphs, action sequences, and bpy code, together with graph- and constraint-oriented evaluation metrics. Extensive experiments show that our method consistently outperforms existing approaches in both geometric fidelity and accurate satisfaction of geometric constraints.

Transformer-based diffusion models have demonstrated remarkable performance at generating high-quality samples. However, our theoretical understanding of the reasons for this success remains limited. For instance, existing models are typically trained by minimizing a denoising objective, which is equivalent to fitting the score function of the training data. However, we do not know why transformer-based models can match the score function for denoising, or why gradient-based methods converge to the optimal denoising model despite the non-convex loss landscape. To the best of our knowledge, this paper provides the first convergence analysis for training transformer-based diffusion models. More specifically, we consider the population Denoising Diffusion Probabilistic Model (DDPM) objective for denoising data that follow a multi-token Gaussian mixture distribution. We theoretically quantify the required number of tokens per data point and training iterations for the global convergence towards the Bayes optimal risk of the denoising objective, thereby achieving a desired score matching error. A deeper investigation reveals that the self-attention module of the trained transformer implements a mean denoising mechanism that enables the trained model to approximate the oracle Minimum Mean Squared Error (MMSE) estimator of the injected noise in the diffusion steps. Numerical experiments validate these findings.

Long-horizon Flexible Job-Shop Scheduling~(FJSP) presents a formidable combinatorial challenge due to complex, interdependent decisions spanning extended time horizons. While learning-based Rolling Horizon Optimization~(RHO) has emerged as a promising paradigm to accelerate solving by identifying and fixing invariant operations, its effectiveness is hindered by the structural complexity of FJSP. Existing methods often fail to capture intricate graph-structured dependencies and ignore the asymmetric costs of prediction errors, in which misclassifying critical-path operations is significantly more detrimental than misclassifying non-critical ones. Furthermore, dynamic shifts in predictive confidence during the rolling process make static pruning thresholds inadequate. To address these limitations, we propose Graph-RHO, a novel critical-path-aware graph-based RHO framework. First, we introduce a topology-aware heterogeneous graph network that encodes subproblems as operation-machine graphs with multi-relational edges, leveraging edge-feature-aware message passing to predict operation stability. Second, we incorporate a critical-path-aware mechanism that injects inductive biases during training to distinguish highly sensitive bottleneck operations from robust ones. Third, we devise an adaptive thresholding strategy that dynamically calibrates decision boundaries based on online uncertainty estimation to align model predictions with the solver's search space. Extensive experiments on standard benchmarks demonstrate that \mbox{Graph-RHO} establishes a new state of the art in solution quality and computational efficiency. Remarkably, it exhibits exceptional zero-shot generalization, reducing solve time by over 30\% on large-scale instances (2000 operations) while achieving superior solution quality. Our code is available \href{https://github.com/IntelliSensing/Graph-RHO}{here}.

Recent advancements in the Generative Reward Model (GRM) have demonstrated its potential to enhance the reasoning abilities of LLMs through Chain-of-Thought (CoT) prompting. Despite these gains, existing implementations of GRM suffer from two critical limitations. First, CoT prompting is applied indiscriminately to all inputs regardless of their inherent complexity. This introduces unnecessary computational costs for tasks amenable to fast, direct inference. Second, existing approaches primarily rely on voting-based mechanisms to evaluate CoT outputs, which often lack granularity and precision in assessing reasoning quality. In this paper, we propose E-GRM, an efficient generative reward modeling framework grounded in model-internal uncertainty. E-GRM leverages the convergence behavior of parallel model generations to estimate uncertainty and selectively trigger CoT reasoning only when needed, without relying on handcrafted features or task-dependent signals. To improve reward fidelity, we introduce a lightweight discriminative scorer trained with a hybrid regression--ranking objective to provide fine-grained evaluation of reasoning paths. Experiments on multiple reasoning benchmarks show that E-GRM substantially reduces inference cost while consistently improving answer accuracy, demonstrating that model-internal uncertainty is an effective and general signal for efficient reasoning-aware reward modeling.

End-to-end full-duplex Speech Language Models (SLMs) require precise turn-taking for natural interaction. However, optimizing temporal dynamics via standard raw-token reinforcement learning (RL) degrades semantic quality, causing severe generative collapse and repetition. We propose ASPIRin, an interactivity-optimized RL framework that explicitly decouples when to speak from what to say. Using Action Space Projection, ASPIRin maps the text vocabulary into a coarse-grained binary state (active speech vs. inactive silence). By applying Group Relative Policy Optimization (GRPO) with rule-based rewards, it balances user interruption and response latency. Empirical evaluations show ASPIRin optimizes interactivity across turn-taking, backchanneling, and pause handling. Crucially, isolating timing from token selection preserves semantic coherence and reduces the portion of duplicate n-grams by over 50% compared to standard GRPO, effectively eliminating degenerative repetition.

Social media research on mental health has focused predominantly on detecting and diagnosing conditions at the individual level. In this work, we shift attention to \emph{intergroup} behavior, examining how two prominent neurodivergent communities, ADHD and autism, adjust their language when engaging with each other on Reddit. Grounded in Communication Accommodation Theory (CAT), we first establish that each community maintains a distinct linguistic profile as measured by Language Inquiry and Word Count Lexicon (LIWC). We then show that these profiles shift in opposite directions when users cross community boundaries: features that are elevated in one group's home community decrease when its members post in the other group's space, and vice versa, consistent with convergent accommodation. The involvement of topic-independent summary variables (Authentic, Clout) in these shifts provides partial evidence against a purely topical explanation. Finally, in an exploratory longitudinal analysis around the moment of public diagnosis disclosure, we find that its effects on linguistic style are small and, in some cases, directionally opposite to cross-community accommodation, providing initial evidence that situational audience adaptation and longer-term identity processes may involve different mechanisms. Our findings contribute to understanding intergroup communication dynamics among neurodivergent populations online and carry implications for community moderation and clinical perspectives on these conditions.

We study reward poisoning attacks in reinforcement learning (RL), where an adversary manipulates rewards within constrained budgets to force the target RL agent to adopt a policy that aligns with the attacker's objectives. Prior works on reward poisoning mainly focused on sufficient conditions to design a successful attacker, while only a few studies discussed the infeasibility of targeted attacks. This paper provides the first precise necessity and sufficiency characterization of the attackability of a linear MDP under reward poisoning attacks. Our characterization draws a bright line between the vulnerable RL instances, and the intrinsically robust ones which cannot be attacked without large costs even running vanilla non-robust RL algorithms. Our theory extends beyond linear MDPs -- by approximating deep RL environments as linear MDPs, we show that our theoretical framework effectively distinguishes the attackability and efficiently attacks the vulnerable ones, demonstrating both the theoretical and practical significance of our characterization.

This study addresses a key limitation in deep learning Automatic Modulation Classification (AMC) models, which perform well at high signal-to-noise ratios (SNRs) but degrade under noisy conditions due to conventional feature extraction suppressing both discriminative structure and interference. The goal was to develop a feature-preserving denoising method that mitigates the loss of modulation class separation. A deep learning AMC model was proposed, incorporating a cross-channel self-attention block to capture dependencies between in-phase and quadrature components, along with dual-path deep residual shrinkage denoising blocks to suppress noise. Experiments using the RML2018.01a dataset employed stratified sampling across 24 modulation types and 26 SNR levels. Results showed that denoising depth strongly influences robustness at low and moderate SNRs. Compared to benchmark models PET-CGDNN, MCLDNN, and DAE, the proposed model achieved notable accuracy improvements across -8 dB to +2 dB SNR, with increases of 3%, 2.3%, and 14%, respectively. Cross-validation confirmed the model's robustness, yielding a mean accuracy of 62.6%, macro precision of 65.8%, macro-recall of 62.6%, and macro-F1 score of 62.9%. The architecture advances interference-aware AMC by formalizing baseband modeling as orthogonal subproblems and introducing cross-channel attention as a generalized complex interaction operator, with ablations confirming the critical role of feature-preserving denoising for robustness at low-to-medium SNR.

The F10.7 and F30 solar indices are the solar radio fluxes measured at wavelengths of 10.7 cm and 30 cm, respectively, which are key indicators of solar activity. F10.7 is valuable for explaining the impact of solar ultraviolet (UV) radiation on the upper atmosphere of Earth, while F30 is more sensitive and could improve the reaction of thermospheric density to solar stimulation. In this study, we present a new deep learning model, named the Solar Index Network, or SINet for short, to predict daily values of the F10.7 and F30 solar indices. The SINet model is designed to make medium-term predictions of the index values (1-60 days in advance). The observed data used for SINet training were taken from the National Oceanic and Atmospheric Administration (NOAA) as well as Toyokawa and Nobeyama facilities. Our experimental results show that SINet performs better than five closely related statistical and deep learning methods for the prediction of F10.7. Furthermore, to our knowledge, this is the first time deep learning has been used to predict the F30 solar index.

Through systematic experiments on long-context generation, we observe a damaging failure mode in which decoding can collapse into persistent repetition loops. We find that this degeneration is driven by collapsed attention patterns, where a subset of heads locks onto a narrow suffix of the history, and is further stabilized by inference-time KV cache reuse. Crucially, since many existing KV cache policies rely on attention-based importance, this collapse can produce spuriously high scores for repetitive tokens, causing cache management to inadvertently amplify repetition. To study this phenomenon in a controlled and reproducible manner, we introduce LoopBench, a benchmark with explicit loop-inducing conditions and loop-oriented metrics that quantify repetition severity and generation instability beyond downstream task scores. Building on these insights, we propose LoopGuard, a lightweight, plug-in KV cache guard that detects loop onset online and disrupts the feedback cycle by pruning repetitive tail spans under a fixed cache budget. Experiments on LoopBench show that LoopGuard reduces loop incidence by over 90 percentage points, while restoring output diversity and reducing token waste.

In this study, we developed a computational implementation for a model of metaphor comprehension based on the theory of indeterminate natural transformation (TINT) proposed by Fuyama et al. We simplified the algorithms implementing the model to be closer to the original theory and verified it through data fitting and simulations. The outputs of the algorithms are evaluated with three measures: data-fitting with experimental data, the systematicity of the metaphor comprehension result, and the novelty of the comprehension (i.e. the correspondence of the associative structure of the source and target of the metaphor). The improved algorithm outperformed the existing ones in all the three measures.

This paper reports the first documented instance of a language model achieving a perfect score on an officially disclosed Law School Admission Test (LSAT). Controlled experiments on eight reasoning models show that varying the prompt, shuffling answer choices, and sampling multiple responses have no meaningful effect as drivers of performance. Ablating the thinking phase that models generate before answering, however, lowers frontier accuracy by up to 8 percentage points, predominantly in logical reasoning. Distilled models produce full thinking traces in the same format yet plateau far below frontier performance. A pilot process reward model fine-tuned via QLoRA on official LSAT explanations narrows this gap through Best-of-5 selection, with gains again predominantly in logical reasoning. The gatekeeper of elite legal education since 1948, the LSAT has not merely been passed but answered without a single error by models that reason. The upper bound of the cognitive capacities it has tested is no longer exclusive to human cognition.

While modern generative models such as diffusion-based architectures have enabled impressive creative capabilities, they also raise important safety and ethical risks. These concerns have led to growing interest in concept erasure, the process of removing unwanted concepts from model representations. Existing approaches often achieve strong erasure performance but rely on iterative optimization and may inadvertently distort unrelated concepts. In this work, we present a simple yet principled alternative: a linear transformation framework that achieves concept erasure analytically, without any training. Our method adapts a pretrained model through two sequential, closed-form steps: first, computing a proxy projection of the target concept, and second, applying a constrained transformation within the left null space of known concept directions. This design yields a deterministic and geometrically interpretable procedure for safe, efficient, and theory-grounded concept removal. Across a wide range of experiments, including object and style erasure on multiple Stable Diffusion variants and the flow-matching model (FLUX), our approach matches or surpasses the performance of state-of-the-art methods while preserving non-target concepts more faithfully. Requiring only a few seconds to apply, it offers a lightweight and drop-in tool for controlled model editing, advancing the goal of safer and more responsible generative models.

Theory of Mind (ToM), the ability to attribute mental states to others, is a hallmark of social intelligence. While large language models (LLMs) demonstrate promising performance on standard ToM benchmarks, we observe that they often fail to generalize to complex task-specific scenarios, relying heavily on prompt scaffolding to mimic reasoning. The critical misalignment between the internal knowledge and external behavior raises a fundamental question: Do LLMs truly possess intrinsic cognition, and can they externalize this internal knowledge into stable, high-quality behaviors? To answer this, we introduce CoSToM (Causal-oriented Steering for ToM alignment), a framework that transitions from mechanistic interpretation to active intervention. First, we employ causal tracing to map the internal distribution of ToM features, empirically uncovering the internal layers' characteristics in encoding fundamental ToM semantics. Building on this insight, we implement a lightweight alignment framework via targeted activation steering within these ToM-critical layers. Experiments demonstrate that CoSToM significantly enhances human-like social reasoning capabilities and downstream dialogue quality.

Long video summarization presents significant challenges for current multimodal large language models (MLLMs), particularly in maintaining temporal fidelity over extended durations and producing summaries that are both semantically and temporally grounded. In this work, we present LVSum, a human-annotated benchmark designed specifically for evaluating long video summarization with fine-grained temporal alignment. LVSum comprises diverse long-form videos across 13 domains, each paired with human-generated summaries containing precise temporal references. We conduct a comprehensive evaluation of both proprietary and open-source MLLMs on LVSum, assessing performance using newly introduced LLM-based metrics for content relevance and modality coherence, alongside standard evaluation metrics. Our experiments reveal systematic gaps in temporal understanding among existing MLLMs and offer insights that establish a new foundation for advancing temporal reasoning in long video summarization.

With the growing availability of open-sourced adapters trained on the same diffusion backbone for diverse scenes and objects, combining these pretrained weights enables low-cost customized generation. However, most existing model merging methods are designed for classification or text generation, and when applied to image generation, they suffer from content drift due to error accumulation across multiple diffusion steps. For image-oriented methods, training-based approaches are computationally expensive and unsuitable for edge deployment, while training-free ones use uniform fusion strategies that ignore inter-adapter differences, leading to detail degradation. We find that since different adapters are specialized for generating different types of content, the contribution of each diffusion step carries different significance for each adapter. Accordingly, we propose a frequency-domain importance-driven dynamic LoRA switch method. Furthermore, we observe that maintaining semantic consistency across adapters effectively mitigates detail loss; thus, we design an automatic Generation Alignment mechanism to align generation intents at the semantic level. Experiments demonstrate that our FREE-Switch (Frequency-based Efficient and Dynamic LoRA Switch) framework efficiently combines adapters for different objects and styles, substantially reducing the training cost of high-quality customized generation.

Weird generalization is a phenomenon in which models fine-tuned on data from a narrow domain (e.g. insecure code) develop surprising traits that manifest even outside that domain (e.g. broad misalignment)-a phenomenon that prior work has highlighted as a critical safety concern. Here, we present an extended replication study of key weird generalization results across an expanded suite of models and datasets. We confirm that surprising (and dangerous) traits can emerge under certain circumstances, but we find that weird generalization is exceptionally brittle: it emerges only for specific models on specific datasets, and it vanishes under simple training-time, prompt-based interventions. We find that the most effective interventions provide prompt context that makes the generalized behavior the expected behavior. However, we show that even very generic interventions that do not anticipate specific generalized traits can still be effective in mitigating weird generalization's effects. Our findings thus help clarify the nature of the safety threat that weird generalization poses and point toward an easily implemented set of solutions.

Self-supervised music foundation models underperform on key detection, which requires pitch-sensitive representations. In this work, we present the first systematic study showing that the design of self-supervised pretraining directly impacts pitch sensitivity, and demonstrate that masked contrastive embeddings uniquely enable state-of-the-art (SOTA) performance in key detection in the supervised setting. First, we discover that linear evaluation after masking-based contrastive pretraining on Mel spectrograms leads to competitive performance on music key detection out of the box. This leads us to train shallow but wide multi-layer perceptrons (MLPs) on features extracted from our base model, leading to SOTA performance without the need for sophisticated data augmentation policies. We further analyze robustness and show empirically that the learned representations naturally encode common augmentations. Our study establishes self-supervised pretraining as an effective approach for pitch-sensitive MIR tasks and provides insights for designing and probing music foundation models.

Solar eruptions, including flares and coronal mass ejections (CMEs), have a significant impact on Earth. Some flares are associated with CMEs, and some flares are not. The association between flares and CMEs is not always obvious. In this study, we propose a new deep learning method, specifically a hybrid neural network (HNN) that combines a vision transformer with long short-term memory, to predict associations between flares and CMEs. HNN finds spatio-temporal patterns in the time series of line-of-sight magnetograms of solar active regions (ARs) collected by the Helioseismic and Magnetic Imager on board the Solar Dynamics Observatory and uses the patterns to predict whether a flare projected to occur within the next 24 hours will be eruptive (i.e., CME-associated) or confined (i.e., not CME-associated). Our experimental results demonstrate the good performance of the HNN method. Furthermore, the results show that magnetic flux cancellation in polarity inversion line regions may well play a role in triggering flare-associated CMEs, a finding consistent with literature.

Recent studies demonstrate that tool-calling capability enables large language models (LLMs) to interact with external environments for long-horizon financial tasks. While existing benchmarks have begun evaluating financial tool calling, they focus on limited scenarios and rely on call-level metrics that fail to capture trajectory-level reasoning quality. To address this gap, we introduce FinTrace, a benchmark comprising 800 expert-annotated trajectories spanning 34 real-world financial task categories across multiple difficulty levels. FinTrace employs a rubric-based evaluation protocol with nine metrics organized along four axes -- action correctness, execution efficiency, process quality, and output quality -- enabling fine-grained assessment of LLM tool-calling behavior. Our evaluation of 13 LLMs reveals that while frontier models achieve strong tool selection, all models struggle with information utilization and final answer quality, exposing a critical gap between invoking the right tools and reasoning effectively over their outputs. To move beyond diagnosis, we construct FinTrace-Training, the first trajectory-level preference dataset for financial tool-calling, containing 8,196 curated trajectories with tool-augmented contexts and preference pairs. We fine-tune Qwen-3.5-9B using supervised fine-tuning followed by direct preference optimization (DPO) and show that training on FinTrace-Training consistently improves intermediate reasoning metrics, with DPO more effectively suppressing failure modes. However, end-to-end answer quality remains a bottleneck, indicating that trajectory-level improvements do not yet fully propagate to final output quality.

This paper provides a comprehensive evaluation of demographic and linguistic biases in omnimodal language models that process text, images, audio, and video within a single framework. Although these models are being widely deployed, their performance across different demographic groups and modalities is not well studied. Four omnimodal models are evaluated on tasks that include demographic attribute estimation, identity verification, activity recognition, multilingual speech transcription, and language identification. Accuracy differences are measured across age, gender, skin tone, language, and country of origin. The results show that image and video understanding tasks generally exhibit better performance with smaller demographic disparities. In contrast, audio understanding tasks exhibit significantly lower performance and substantial bias, including large accuracy differences across age groups, genders, and languages, and frequent prediction collapse toward narrow categories. These findings highlight the importance of evaluating fairness across all supported modalities as omnimodal language models are increasingly used in real-world applications.

Automatic sleep staging is a multimodal learning problem involving heterogeneous physiological signals such as EEG and EOG, which often suffer from domain shifts across institutions, devices, and populations. In practice, these data are also affected by noisy annotations, yet label-noise-robust multi-source domain generalization remains underexplored. We present the first benchmark for Noisy Labels in Multi-Source Domain-Generalized Sleep Staging (NL-DGSS) and show that existing noisy-label learning methods degrade substantially when domain shifts and label noise coexist. To address this challenge, we propose FF-TRUST, a domain-invariant multimodal sleep staging framework with Joint Time-Frequency Early Learning Regularization (JTF-ELR). By jointly exploiting temporal and spectral consistency together with confidence-diversity regularization, FF-TRUST improves robustness under noisy supervision. Experiments on five public datasets demonstrate consistent state-of-the-art performance under diverse symmetric and asymmetric noise settings. The benchmark and code will be made publicly available at https://github.com/KNWang970918/FF-TRUST.git.

Large language models (LLMs) have recently emerged as promising tools for augmenting Security Operations Center (SOC) workflows, with vendors increasingly marketing autonomous AI solutions for SOCs. However, there remains a limited empirical understanding of how such tools are used, perceived, and adopted by real-world security practitioners. To address this gap, we conduct a mixed-methods analysis of discussions in cybersecurity-focused forums to learn how a diverse group of practitioners use and perceive modern LLM tools for security operations. More specifically, we analyzed 892 posts between December 2022 and September 2025 from three cybersecurity-focused forums on Reddit, and, using a combination of qualitative coding and statistical analysis, examined how security practitioners discuss LLM tools across three dimensions: (1) their stated tools and use cases, (2) the perceived pros and cons of each tool across a set of critical factors, and (3) their adoption of such tools and the expected impacts on the cybersecurity industry and individual analysts. Overall, our findings reveal nuanced patterns in LLM tools adoption, highlighting independent use of LLMs for low-risk, productivity-oriented tasks, alongside active interest around enterprise-grade, security-focused LLM platforms. Although practitioners report meaningful gains in efficiency and effectiveness in LLM-assisted workflows, persistent issues with reliability, verification overheads, and security risks sharply constrain the autonomy granted to LLM tools. Based on these results, we also provide recommendations for developing and adopting LLM tools to ensure the security of organizations and the safety of cybersecurity practitioners.

This paper introduces an LLM agent that automates power grid static analysis by converting natural language into MATPOWER scripts. The framework utilizes DeepSeek-OCR to build an enhanced vector database from MATPOWER manuals. To ensure reliability, it devises a three-tier error-correction system: a static pre-check, a dynamic feedback loop, and a semantic validator. Operating via the Model Context Protocol, the tool enables asynchronous execution and automatically debugging in MATLAB. Experimental results demonstrate that the system achieves a 82.38% accuracy regarding the code fidelity, effectively eliminating hallucinations even in complex analysis tasks.

Deep neural networks (DNNs) have showcased remarkable performance across various tasks and are widely deployed on AI accelerators fabricated in advanced technology nodes for efficiency. As aging effects become more pronounced, timing and voltage guardbands are increasingly applied. Aging-aware adaptive voltage scaling (AVS), which adjusts supply voltage based on on-chip aging scenarios, has emerged as a promising solution to avoid excessive guardbanding. However, conventional AVS techniques overlook the inherent resilience of DNNs and frequently raise the supply voltage unnecessarily, thereby exacerbating aging and increasing power consumption. To enable reliable and efficient AI inference with AVS, in this paper, we develop an accurate aging prediction framework that incorporates historical effects and iterative extrapolation for full-lifetime modeling. Building on this framework, we propose a fault-tolerant voltage scaling policy that exploits DNN resilience and defers voltage increases accordingly. Experiments show that our framework mitigates the pessimism of maximum-voltage baselines, reducing predicted threshold voltage shift (ΔVth) by 19.4% for PMOS and 19.1% for NMOS, respectively. Furthermore, evaluation on representative DNN workloads demonstrates that our optimization reduces aging degradation by up to 45.8% (NMOS) and 30.6% (PMOS) while achieving 14.0% average lifetime power savings compared to resilience-agnostic methods.

Recently, synthetic palmprints have been increasingly used as substitutes for real data to train recognition models. To be effective, such synthetic data must reflect the diversity of real palmprints, including both style variation and geometric variation. However, existing palmprint generation methods mainly focus on style translation, while geometric variation is either ignored or approximated by simple handcrafted augmentations. In this work, we propose FlowPalm, an optical-flow-driven palmprint generation framework capable of simulating the complex non-rigid deformations observed in real palms. Specifically, FlowPalm estimates optical flows between real palmprint pairs to capture the statistical patterns of geometric deformations. Building on these priors, we design a progressive sampling process that gradually introduces the geometric deformations during diffusion while maintaining identity consistency. Extensive experiments on six benchmark datasets demonstrate that FlowPalm significantly outperforms state-of-the-art palmprint generation approaches in downstream recognition tasks. Project page: https://yuchenzou.github.io/FlowPalm/

Deep Neural Networks (DNNs) are widely used by engineers to solve difficult problems that require predictive modeling from data. However, these models are often massive, with millions or billions of parameters, and require substantial computational power, RAM, and storage. This becomes a limitation in practical scenarios where strict size and resource constraints must be respected. In this paper, we present a novel concept-based pruning technique for DNNs that guides pruning decisions using human-interpretable concepts, such as features, colors, and classes. This is particularly important in a software engineering context, as DNNs are integrated into systems and must be pruned according to specific system requirements. Our concept-based pruning solution analyzes neuron activations to identify important neurons from a system requirements viewpoint and uses this information to guide the DNN pruning. We assess our solution using the VGG-19 network and a dataset of 26'384 RGB images, focusing on its ability to produce small, effective pruned DNNs and on the computational complexity and performance of these pruned DNNs. We also analyzed the pruning efficiency of our solution and compared alternative configurations. Our results show that concept-based pruning efficiently generates much smaller, effective pruned DNNs. Pruning greatly improves the computational efficiency and performance of DNNs, properties that are particularly useful for practical applications with stringent memory and computational time constraints. Finally, alternative configuration options enable engineers to identify trade-offs adapted to different practical situations.

Reproducibility must validate architectural robustness, not just numerical accuracy. We evaluate ColBERT-v2 and ConstBERT across five dimensions, finding that while ConstBERT reproduces within 0.05% MRR@10 on MS-MARCO, both models show a drop of 86-97% on long, narrative queries (TREC ToT 2025). Ablations prove this failure is architectural: performance plateaus at 20 words because the MaxSim operator's uniform token weighting cannot distinguish signal from filler noise. Furthermore, undocumented backend parameters create an 8-point gap due to ConstBERT's sparse centroid coverage, and fine-tuning with 3x more data actually degrades performance by up to 29%. We conclude that architectural constraints in multi-vector retrieval cannot be overcome by adaptation alone. Code: https://github.com/utshabkg/multi-vector-reproducibility.

Self-supervised representation learning is central to modern machine learning because it extracts structured latent features from unlabeled data and enables robust transfer across tasks and domains. However, it can suffer from representation collapse, a widely observed failure mode in which embeddings lose discriminative structure and distinct inputs become indistinguishable. To understand the mechanisms that drive collapse and the ingredients that prevent it, we introduce a minimal embedding-only model whose gradient-flow dynamics and fixed points can be analyzed in closed form, using a classification-representation setting as a concrete playground where collapse is directly quantified through the contraction of label-embedding geometry. We illustrate that the model does not collapse when the data are perfectly classifiable, while a small fraction of frustrated samples that cannot be classified consistently induces collapse through an additional slow time scale that follows the early performance gain. Within the same framework, we examine collapse prevention by adding a shared projection head and applying stop-gradient at the level of the training dynamics. We analyze the resulting fixed points and develop a dynamical mean-field style self-consistency description, showing that stop-gradient enables non-collapsed solutions and stabilizes finite class separation under frustration. We further verify empirically that the same qualitative dynamics and collapse-prevention effects appear in a linear teacher-student model, indicating that the minimal theory captures features that persist beyond the pure embedding setting.

In the decentralized distributed learning, achieving fast convergence and low communication cost is essential for scalability and high efficiency. Adaptive gradient methods, such as Adam, have demonstrated strong practical performance in deep learning and centralized distributed settings. However, their convergence properties remain largely unexplored in decentralized settings involving multiple local training steps, such as federated learning. To address this limitation, we propose LoDAdaC, a unified multiple Local Training (MLT) Decentralized framework with Adam-type updates and Compressed communication (CC). LoDAdaC accommodates a broad class of optimizers for its local adaptive updates, including AMSGrad, Adam, and AdaGrad; it is compatible with standard (possibly biased) compressors such as low-bit quantization and sparsification. MLT and CC enable LoDAdaC to achieve multiplied reduction of communication cost, while the technique of adaptive updates enables fast convergence. We rigorously prove the combined advantage through complexity analysis. In addition, experiments on image classification and GPT-style language model training validate our theoretical findings and show that LoDAdaC significantly outperforms existing decentralized algorithms in terms of convergence speed and communication efficiency.

Muon has emerged as a promising optimizer for large-scale foundation model pre-training by exploiting the matrix structure of neural network updates through iterative orthogonalization. However, its practical efficiency is limited by the need for multiple Newton--Schulz (NS) iterations per optimization step, which introduces non-trivial computation and communication overhead. We propose Muon$^2$, an extension of Muon that applies Adam-style adaptive second-moment preconditioning before orthogonalization. Our key insight is that the core challenge of polar approximation in Muon lies in the ill-conditioned momentum matrix, of which the spectrum is substantially improved by Muon$^2$, leading to faster convergence toward a practically sufficient orthogonalization. We further characterize the practical orthogonalization quality via directional alignment, under which Muon$^2$ demonstrates dramatic improvement over Muon at each polar step. Across GPT and LLaMA pre-training experiments from 60M to 1.3B parameters, Muon$^2$ consistently outperforms Muon and recent Muon variants while reducing NS iterations by 40\%. We further introduce Muon$^2$-F, a memory-efficient factorized variant that preserves most of the gains of Muon$^2$ with negligible memory overhead.

We present a family of Kaczmarz-based preference learning algorithms for real-time personalized matchmaking in reciprocal recommender systems. Post-step L2 normalization, common in Kaczmarz-inspired online learners, induces exponential recency bias: the influence of the t-th interaction decays as eta^(n - t), reaching approximately 1e-6 after just 20 swipes at eta = 0.5. We resolve this by replacing the normalization step with a Tikhonov-regularized projection denominator that bounds step size analytically without erasing interaction history. When candidate tag vectors are not pre-normalized, as in realistic deployments where candidates vary in tag density, the Tikhonov denominator ||a||^2 + alpha produces genuinely per-candidate adaptive step sizes, making it structurally distinct from online gradient descent with any fixed learning rate. We further derive a block variant that processes full swipe sessions as a single Gram matrix solve. Population-scale simulation over 6,400 swipes reveals that Block Normalized Kaczmarz (BlockNK), which combines the batch Gram solve with post-session L2 normalization, achieves the highest preference alignment (Align@20 = 0.698), the strongest inter-session direction stability (delta = 0.994), and the flattest degradation profile under label noise across flip ratios p_flip in [0.10, 0.35]. Experiments under cosine similarity subsampling further show that adaptively filtering the candidate pool toward the current preference direction substantially improves asymptotic alignment, at the cost of introducing a feedback loop that may slow recovery from miscalibration. The sequential Tikhonov-Kaczmarz method performs comparably to K-NoNorm under our simulation conditions, suggesting the dominant practical gain over normalized Kaczmarz is the removal of per-step normalization rather than the Tikhonov constant alpha itself.

Microreboot enables fast recovery by restarting only the failing component, but in modern microservices naive restarts are unsafe: dense dependencies mean rebooting one service can disrupt many callers. Autonomous remediation agents compound this by actuating raw infrastructure commands without safety guarantees. We make microreboot practical by separating planning from actuation: a three-agent architecture (diagnosis, planning, verification) proposes typed remediation plans over a seven-action ISA with explicit side-effect semantics, and a small microkernel validates and executes each plan transactionally. Agents are explicitly untrusted; safety derives from the ISA and microkernel. To determine where restart is safe, we infer recovery boundaries online from distributed traces, computing minimal restart groups and ordering constraints. On industrial traces (Alibaba, Meta) and DeathStarBench with fault injection, recovery-group inference runs in 21 ms at P99; typed actuation reduces agent-caused harm by 95% in simulation and achieves 0% harm online. The primary value is safety, not speed: LLM inference overhead increases TTR for services with fast auto-restart.

The rapid adoption of large language models has introduced a new class of AI-generated fake news that coexists with traditional human-written misinformation, raising important questions about how these two forms of deceptive content differ and how reliably they can be distinguished. This study examines linguistic, structural, and emotional differences between human-written and AI-generated fake news and evaluates machine learning and ensemble-based methods for distinguishing these content types. A document-level feature representation is constructed using sentence structure, lexical diversity, punctuation patterns, readability indices, and emotion-based features capturing affective dimensions such as fear, anger, joy, sadness, trust, and anticipation. Multiple classification models, including logistic regression, random forest, support vector machines, extreme gradient boosting, and a neural network, are applied alongside an ensemble framework that aggregates predictions across models. Model performance is assessed using accuracy and area under the receiver operating characteristic curve. The results show strong and consistent classification performance, with readability-based features emerging as the most informative predictors and AI-generated text exhibiting more uniform stylistic patterns. Ensemble learning provides modest but consistent improvements over individual models. These findings indicate that stylistic and structural properties of text provide a robust basis for distinguishing AI-generated misinformation from human-written fake news.

Many applications today use large language models for code generation; however, production systems have strict latency requirements that can be difficult to meet with large models. Small language models with a few billion parameters are resource efficient but may suffer from limited reasoning, hallucinations, or poor retention of longer context. Fine tuning improves task specific accuracy by embedding domain knowledge directly into model weights, reducing reliance on runtime context. We previously implemented a baseline natural language to code generation approach using a retrieval augmented generation pipeline that dynamically selected few shot examples to embed domain specific language context for a large language model. In this study, we evaluate small language models for generating domain specific language from natural language by fine tuning variants of Mistral and other models on a dataset of natural language code pairs. Our results show that the fine-tuned models achieve improved performance and latency on test datasets compared to larger models. We also demonstrate that the trained model can be further fine-tuned for customer specific scenarios without degrading general performance, helping resolve production issues. Load testing followed by production deployment confirmed optimal performance in terms of latency and quality. These findings demonstrate that task specific fine tuning with small language models provides an efficient, faster, and cost-effective alternative to large language models for domain specific language generation.

The rapid adoption of large vision-language models (LVLMs) in recent years has been accompanied by growing fairness concerns due to their propensity to reinforce harmful societal stereotypes. While significant attention has been paid to such fairness concerns in the context of social biases, relatively little prior work has examined the presence of stereotypes in LVLMs related to cultural contexts such as religion, nationality, and socioeconomic status. In this work, we aim to narrow this gap by investigating how cultural contexts depicted in images influence the judgments LVLMs make about a person's moral, ethical, and political values. We conduct a multi-dimensional analysis of such value judgments in five popular LVLMs using counterfactual image sets, which depict the same person across different cultural contexts. Our evaluation framework diagnoses LVLM awareness of cultural value differences through the use of Moral Foundations Theory, lexical analyses, and the sensitivity of generated values to depicted cultural contexts.

Reservoir computing (RC) is a computational framework known for its training efficiency, making it ideal for physical hardware implementations. However, realizing the complex interconnectivity of traditional reservoirs in physical systems remains a significant challenge. This paper proposes a physical RC scheme inspired by the biological vestibular system. To overcome hardware complexity, we introduce a designed uncoupled topology and demonstrate that it achieves performance comparable to fully coupled networks. We theoretically analyze the difference between these topologies by deriving a memory capacity formula for linear reservoirs, identifying specific conditions where both configurations yield equivalent memory. These analytical results are demonstrated to approximately hold for nonlinear reservoir systems. Furthermore, we systematically examine the impact of reservoir size on predictive statistics and memory capacity. Our findings suggest that uncoupled reservoir architectures offer a mathematically sound and practically feasible pathway for efficient physical reservoir computing.

Object binding is a foundational process in visual cognition, during which low-level perceptual features are joined into object representations. Binding has been considered a fundamental challenge for neural networks, and a major milestone on the way to artificial models with flexible visual intelligence. Recently, several investigations have demonstrated evidence that binding mechanisms emerge in pretrained vision models, enabling them to associate portions of an image that contain an object. The question remains: how are these models binding objects together? In this work, we investigate whether vision models rely on the principle of Gestalt continuity to perform object binding, over and above other principles like similarity and proximity. Using synthetic datasets, we demonstrate that binding probes are sensitive to continuity across a wide range of pretrained vision transformers. Next, we uncover particular attention heads that track continuity, and show that these heads generalize across datasets. Finally, we ablate these attention heads, and show that they often contribute to producing representations that encode object binding.

Fine-tuning Large Language Models (LLMs) typically involves either full fine-tuning, which updates all model parameters, or Parameter-Efficient Fine-Tuning (PEFT), which adjusts a small subset of parameters. However, both approaches have inherent limitations: full fine-tuning is computationally expensive, while PEFT often struggles to learn new knowledge and exhibits suboptimal performance. To overcome these issues, we propose a novel hybrid fine-tuning approach that jointly updates both LLMs and PEFT modules using a combination of zeroth-order and first-order optimization methods. To analyze our new algorithm, we develop a theoretical framework centered on the concept of hybrid smoothness condition, which accounts for the heterogeneous nature of the optimization landscape in joint LLM and PEFT training. We derive a rigorous convergence analysis for the convergence of reshuffling-type SGD algorithm under multiple learning rates and demonstrate its effectiveness through extensive empirical studies across various downstream tasks and model architectures. On the practical side, our results demonstrate consistent performance improvement, making the approach a viable solution for large-scale language model fine-tuning.

Conventional field operations spend most of their energy moving the tractor body, not the implement. Yet feasibility studies for novel agricultural vehicles rarely tie mechanics, energy harvest, draft, field geometry, economics, life-cycle CO2, and uncertainty quantification together on a single reproducible code path. This paper builds such a framework and applies it to CableTract, a two-module cable-driven field robot. A stationary Main Unit (winch + motor + battery + harvester module) (MU) and a lighter Anchor module (held by helical screw piles) tension a cable across a strip while a lightweight implement carriage rolls along it. The heavy bodies stay on the headland; only the carriage enters the field. The carriage runs a 10-implement library co-designed for the cable architecture. This co-design is the paper's central analytical lever. The framework is prototype-free. It chains a catenary cable model, a drivetrain efficiency chain, a stochastic draft model fitted to the co-designed library, an hourly solar + wind + battery simulator on six sites, a polygon coverage planner on a 50-field corpus, a contact-pressure compaction model, a discounted cash-flow economics engine with battery replacement and life-cycle CO2, and a global sensitivity analysis on 20 inputs. An operating-envelope sweep and an architectural-variant comparison close the loop. The full implementation is open source. Applied to the codesigned reference, the framework yields energy, compaction advantages and potential off-grid operation.

Healthcare administration accounts for over $1 trillion in annual spending, making it a promising target for LLM-based computer-use agents (CUAs). While clinical applications of LLMs have received significant attention, no benchmark exists for evaluating CUAs on end-to-end administrative workflows. To address this gap, we introduce HealthAdminBench, a benchmark comprising four realistic GUI environments: an EHR, two payer portals, and a fax system, and 135 expert-defined tasks spanning three administrative task types: Prior Authorization, Appeals and Denials Management, and Durable Medical Equipment (DME) Order Processing. Each task is decomposed into fine-grained, verifiable subtasks, yielding 1,698 evaluation points. We evaluate seven agent configurations under multiple prompting and observation settings and find that, despite strong subtask performance, end-to-end reliability remains low: the best-performing agent (Claude Opus 4.6 CUA) achieves only 36.3 percent task success, while GPT-5.4 CUA attains the highest subtask success rate (82.8 percent). These results reveal a substantial gap between current agent capabilities and the demands of real-world administrative workflows. HealthAdminBench provides a rigorous foundation for evaluating progress toward safe and reliable automation of healthcare administrative workflows.

Hybrid approaches that combine data-driven learning with physics-based insight have shown promise for improving the reliability of industrial condition monitoring. This work develops a hybrid condition monitoring framework that integrates primary sensor measurements, lagged temporal features, and physics-informed residuals derived from nominal surrogate models. Two hybrid integration strategies are examined. The first is a feature-level fusion approach that augments the input space with residual and temporal information. The second is a model-level ensemble approach in which machine learning classifiers trained on different feature types are combined at the decision level. Both hybrid approaches of the condition monitoring framework are evaluated on a continuous stirred-tank reactor (CSTR) benchmark using several machine learning models and ensemble configurations. Both feature-level and model-level hybridization improve diagnostic accuracy relative to single-source baselines, with the best model-level ensemble achieving a 2.9\% improvement over the best baseline ensemble. To assess predictive reliability, conformal prediction is applied to quantify coverage, prediction-set size, and abstention behavior. The results show that hybrid integration enhances uncertainty management, producing smaller and well-calibrated prediction sets at matched coverage levels. These findings demonstrate that lightweight physics-informed residuals, temporal augmentation, and ensemble learning can be combined effectively to improve both accuracy and decision reliability in nonlinear industrial systems.

Text-to-image (T2I) models, and their encoded biases, increasingly shape the visual media the public encounters. While researchers have produced a rich body of work on bias measurement, auditing, and mitigation in T2I systems, those methods largely target technical stakeholders, leaving a gap in public legibility. We introduce GLEaN (Generative Likeness Evaluation at N-Scale), a portrait-based explainability pipeline designed to make T2I model biases visually understandable to a broad audience. GLEaN comprises three stages: automated large-scale image generation from identity prompts, facial landmark-based filtering and spatial alignment, and median-pixel composition that distills a model's central tendency into a single representative portrait. The resulting composites require no statistical background to interpret; a viewer can see, at a glance, who a model 'imagines' when prompted with 'a doctor' versus a 'felon.' We demonstrate GLEaN on Stable Diffusion XL across 40 social and occupational identity prompts, producing composites that reproduce documented biases and surface new associations between skin tone and predicted emotion. We find in a between-subjects user study (N = 291) that GLEaN portraits communicate biases as effectively as conventional data tables, but require significantly less viewing time. Because the method relies solely on generated outputs, it can also be replicated on any black-box and closed-weight systems without access to model internals. GLEaN offers a scalable, model-agnostic approach to bias explainability, purpose-built for public comprehension, and is publicly available at https://github.com/cultureiolab/GLEaN.

Subsurface stratigraphy contains important spatio-temporal information about accumulation, deformation, and layer formation in polar ice sheets. In particular, variations in internal ice layer thickness provide valuable constraints for snow mass balance estimation and projections of ice sheet change. Although radar sensors can capture these layered structures as depth-resolved radargrams, convolutional neural networks applied directly to radar images are often sensitive to speckle noise and acquisition artifacts. In addition, purely data-driven methods may underuse physical knowledge, leading to unrealistic thickness estimates under spatial or temporal extrapolation. To address these challenges, we develop K-STEMIT, a novel knowledge-informed, efficient, multi-branch spatio-temporal graph neural network that combines a geometric framework for spatial learning with temporal convolution to capture temporal dynamics, and incorporates physical data synchronized from the Model Atmospheric Regional physical weather model. An adaptive feature fusion strategy is employed to dynamically combine features learned from different branches. Extensive experiments have been conducted to compare K-STEMIT against current state-of-the-art methods in both knowledge-informed and non-knowledge-informed settings, as well as other existing methods. Results show that K-STEMIT consistently achieves the highest accuracy while maintaining near-optimal efficiency. Most notably, incorporating adaptive feature fusion and physical priors reduces the root mean-squared error by 21.01% with negligible additional cost compared to its conventional multi-branch variants. Additionally, our proposed K-STEMIT achieves consistently lower per-year relative MAE, enabling reliable, continuous spatiotemporal assessment of snow accumulation variability across large spatial regions.

Much work has been done on designing fast and accurate sampling for diffusion language models (dLLMs). However, these efforts have largely focused on the tradeoff between speed and quality of individual samples; how to additionally ensure diversity across samples remains less well understood. In this work, we show that diversity can be increased by using softened, tempered versions of familiar confidence-based remasking heuristics, retaining their computational benefits and offering simple implementations. We motivate this approach by introducing an idealized formal model of fork tokens and studying the impact of remasking on the expected entropy at the forks. Empirically, the proposed tempered heuristics close the exploration gap (pass@k) between existing confidence-based and autoregressive sampling, hence outperforming both when controlling for cost (pass@NFE). We further study how the increase in diversity translates to downstream post-training and test-time compute scaling. Overall, our findings demonstrate that simple, efficient, and diverse sampling from dLLMs is possible.

LLM-based agents increasingly coordinate decisions in multi-agent systems, often attaching natural-language reasoning to actions. However, reasoning is neither free nor automatically reliable: it incurs computational cost and, without verification, may degenerate into persuasive cheap talk. We introduce Explanatory Equilibrium as a design principle for explanation-aware multi-agent systems and study a regime in which agents exchange structured reasoning artifacts-auditable claims paired with concise text-while receivers apply bounded verification through probabilistic audits under explicit resource constraints. We contribute (i) a minimal mechanism-level exchange-audit model linking audit intensity, misreporting incentives, and reasoning costs, and (ii) empirical evidence from a finance-inspired LLM setting involving a Trader and a Risk Manager. In ambiguous, borderline proposals, auditable artifacts prevent the cost of silence driven by conservative validation under asymmetric information: without structured claims, approval and welfare collapse. By contrast, structured reasoning unlocks coordination while maintaining consistently low bad-approval rates across audit intensities, audit budgets, and incentive regimes. Our results suggest that scalable, safety-preserving coordination in LLM-based multi-agent systems depends not only on audit strength, but more fundamentally on disciplined externalization of reasoning into partially verifiable artifacts.

Simultaneous Speech Translation (SimulST) requires balancing high translation quality with low latency. Recent work introduced REINA, a method that trains a Read/Write policy based on estimating the information gain of reading more audio. However, we find that information-based policies often lack temporal context, leading the policy to bias itself toward reading most of the audio before starting to write. We improve REINA using two distinct strategies: a supervised alignment network (REINA-SAN) and a timestep-augmented network (REINA-TAN). Our results demonstrate that while both methods significantly outperform the baseline and resolve stability issues, REINA-TAN provides a slightly superior Pareto frontier for streaming efficiency, whereas REINA-SAN offers more robustness against 'read loops'. Applied to Whisper, both methods improve the pareto frontier of streaming efficiency as measured by Normalized Streaming Efficiency (NoSE) scores up to 7.1% over existing competitive baselines.

In the psychological literature the term `general intelligence' describes correlations between abilities and not simply the number of abilities. This paper connects Spearman's $g$-factor from psychometrics, measuring a positive manifold, to the implicit ``$G$-factor'' in claims about artificial general intelligence (AGI) performance on temporally structured benchmarks. By treating LLM benchmark batteries as cognitive test batteries and model releases as subjects, principal component analysis is applied to a models $\times$ benchmarks $\times$ time matrix spanning 39 models (2019--2025) and 14 benchmarks. Preliminary results confirm a strong positive manifold in which all 28 pairwise correlations positive across 8 benchmarks. By analyzing the spectrum of the benchmark correlation through time, PC1 explains 90\% of variance on a 5-benchmark core battery ($n=19$)) reducing to 77\% by 2024. On a four benchmark battery, PC1 is found to peak at 92\% of the variance between 2023--2024 and reduce to 64\% with the arrival of reasoning-specialized models in 2024. This is coincident with a rotation in the G-factor as models outsource `reasoning' to tools. The analysis of partial correlation matrices through time provides evidence for the evolution of specialization beneath the positive manifold of general intelligence (AI-hedgehog) encompassing diverse high dimensional problem solving systems (AI-foxes). In strictly psychometric terms, AI models exhibit general intelligence suppressing specialized intelligences. LLMs invert the ideal of substituting complicated models with parsimonious mechanisms, a `Ptolemaic Succession' of theories, with architectures of increasing hierarchical complication and capability.

We study last-iterate convergence of SGD with greedy step size over smooth quadratics in the interpolation regime, a setting which captures the classical Randomized Kaczmarz algorithm as well as other popular iterative linear system solvers. For these methods, we show that the $t$-th iterate attains an $O(1/t^{3/4})$ convergence rate, addressing a question posed by Attia, Schliserman, Sherman, and Koren, who gave an $O(1/t^{1/2})$ guarantee for this setting. In the proof, we introduce the family of stochastic contraction processes, whose behavior can be described by the evolution of a certain deterministic eigenvalue equation, which we analyze via a careful discrete-to-continuous reduction.

To improve crop genetics, high-throughput, effective and comprehensive phenotyping is a critical prerequisite. While such tasks were traditionally performed manually, recent advances in multimodal foundation models, especially in vision-language models (VLMs), have enabled more automated and robust phenotypic analysis. However, plant science remains a particularly challenging domain for foundation models because it requires domain-specific knowledge, fine-grained visual interpretation, and complex biological and agronomic reasoning. To address this gap, we develop PlantXpert, an evidence-grounded multimodal reasoning benchmark for soybean and cotton phenotyping. Our benchmark provides a structured and reproducible framework for agronomic adaptation of VLMs, and enables controlled comparison between base models and their domain-adapted counterparts. We constructed a dataset comprising 385 digital images and more than 3,000 benchmark samples spanning key plant science domains including disease, pest control, weed management, and yield. The benchmark can assess diverse capabilities including visual expertise, quantitative reasoning, and multi-step agronomic reasoning. A total of 11 state-of-the-art VLMs were evaluated. The results indicate that task-specific fine-tuning leads to substantial improvement in accuracy, with models such as Qwen3-VL-4B and Qwen3-VL-30B achieving up to 78%. At the same time, gains from model scaling diminish beyond a certain capacity, generalization across soybean and cotton remains uneven, and quantitative as well as biologically grounded reasoning continue to pose substantial challenges. These findings suggest that PlantXpert can serve as a foundation for assessing evidence-grounded agronomic reasoning and for advancing multimodal model development in plant science.

Emergency department triage relies heavily on both quantitative vital signs and qualitative clinical notes, yet multimodal machine learning models predicting triage acuity often suffer from modality collapse by over-relying on structured tabular data. This limitation severely hinders demographic generalizability, particularly for pediatric patients where developmental variations in vital signs make unstructured clinical narratives uniquely crucial. To address this gap, we propose a late-fusion multimodal architecture that processes tabular vitals via XGBoost and unstructured clinical text via Bio_ClinicalBERT, combined through a Logistic Regression meta-classifier to predict the 5-level Emergency Severity Index. To explicitly target the external validity problem, we train our model exclusively on adult encounters from the MIMIC-IV and NHAMCS datasets and evaluate its zero-shot generalization on a traditionally overlooked pediatric cohort. Furthermore, we employ symmetric modality dropout during training to prevent the ensemble from overfitting to adult-specific clinical correlations. Our results demonstrate that the multimodal framework significantly outperforms single-modality baselines. Most notably, applying a 30-40% symmetric modality dropout rate yielded steep performance improvements in the unseen pediatric cohort, elevating the Quadratic Weighted Kappa to 0.351. These findings highlight modality dropout as a critical regularization technique for mitigating modality collapse and enhancing cross-demographic generalization in clinical AI.

Polyanskiy proposed a framework for the unsourced multiple access channel (MAC) problem where users employ a common codebook in the finite blocklength regime. However, existing approaches handle channel noise before the joint decoder. In this work, we introduce a decoder compatible diffusion denoiser as a lightweight analysis within joint decoding. The score network is trained on samples drawn from the channel output distribution, making the method easy to integrate with existing code designs. In our theoretical analysis, we derive a diffusion-denoiser random-coding achievable bound that is strictly tighter. Simulations on existing decoders, including FASURA, MSUG-MRA and pilot-based method, show consistent performance gains with at least a $0.5$ $\mathrm{dB}$ improvement in required $\mathrm{E_b/N_0}$ at a fixed error target.

Across multiple language pairings (English $\to$ \{Spanish, French, German, Mandarin, Japanese, Urdu, Cantonese\}), we find reasoning errors in translation. To quantify how often these reasoning errors occur, we leverage an automated annotation protocol for reasoning evaluation wherein the goal is to detect if a reasoning step is any of three error categories: (1) source sentence-misaligned, (2) model hypothesis-misaligned, or (3) reasoning trace-misaligned. We probe the reasoning model with perturbed traces correcting for these identified reasoning errors using an array of weak-to-strong interventions: hedging, removal, re-reasoning after removal, hindsight, and oracle interventions. Experimenting with interventions on the reasoning traces suggests that small corrections to the reasoning have little impact on translation quality, but stronger interventions yield the highest resolution rates, despite translation quality gains being mixed. We find ultimately that reasoning errors in MT can be identified with high precision in Urdu but lower precision in Spanish, but that removing these reasoning errors does not resolve the initial errors significantly, suggesting limited reasoning faithfulness for machine translation.

AI agents are being increasingly deployed across a wide range of real-world applications. In this paper, we propose an agentic AI framework for in-situ process monitoring for defect detection in wire-arc additive manufacturing (WAAM). The autonomous agent leverages a WAAM process monitoring dataset and a trained classification tool to build AI agents and uses a large language model (LLM) for in-situ process monitoring decision-making for defect detection. A processing agent is developed based on welder process signals, such as current and voltage, and a monitoring agent is developed based on acoustic data collected during the process. Both agents are tasked with identifying porosity defects from processing and monitoring signals, respectively. Ground truth X-ray computed tomography (XCT) data are used to develop classification tools for both the processing and monitoring agents. Furthermore, a multi-agent framework is demonstrated in which the processing and monitoring agents are orchestrated together for parallel decision-making on the given task of defect classification. Evaluation metrics are proposed to determine the efficacy of both individual agents, the combined single-agent, and the coordinated multi-agent system. The multi-agent configuration outperforms all individual-agent counterparts, achieving a decision accuracy of 91.6% and an F1 score of 0.821 on decided runs, across 15 independent runs, and a reasoning quality score of 3.74 out of 5. These in-situ process monitoring agents hold significant potential for autonomous real-time process monitoring and control toward building qualified parts for WAAM and other additive manufacturing processes.

Medical vision-language datasets are often limited in size and biased toward negative findings, as clinicians report abnormalities mostly but might omit some positive/neutral findings because they might be considered as irrelevant to the patient's condition. We propose a self-supervised data enrichment method that leverages semantic clustering of report sentences. Then we enrich the findings in the medical reports in the training set by adding positive/neutral observations from different clusters in a self-supervised manner. Our approach yields consistent gains in supervised fine-tuning (5.63%, 3.04%, 7.40%, 5.30%, 7.47% average gains on COMET score, Bert score, Sentence Bleu, CheXbert-F1 and RadGraph-F1 scores respectively). Ablation studies confirm that improvements stem from semantic clustering rather than random augmentation. Furthermore, we introduce a way to incorporate semantic cluster information into the reward design for GRPO training, which leads to further performance gains (2.78%, 3.14%, 12.80% average gains on COMET score, Bert score and Sentence Bleu scores respectively). We share our code at https://anonymous.4open.science/r/SemEnrich-75CF

Accurate volume estimation of objects from visual data is a long-standing challenge in computer vision with significant applications in robotics, logistics, and smart health. Existing methods often rely on complex 3D reconstruction pipelines or struggle with the ambiguity inherent in single-view images. To address these limitations, we introduce a new method that fuses implicit 3D cues from stereo vision with explicit prior knowledge from natural language text. Our approach extracts deep features from a stereo image pair and a descriptive text prompt that contains the object's class and an approximate volume, then integrates them using a simple yet effective projection layer into a unified, multi-modal representation for regression. We conduct extensive experiments on public datasets demonstrating that our text-guided approach significantly outperforms vision-only baselines. Our findings show that leveraging even simple textual priors can effectively guide the volume estimation task, paving the way for more context-aware visual measurement systems. Code: https://gitlab.com/viper-purdue/stereo-typical-estimator.

Recent work has shown that probing model internals can reveal a wealth of information not apparent from the model generations. This poses the risk of unintentional or malicious information leakage, where model users are able to learn information that the model owner assumed was inaccessible. Using vision-language models as a testbed, we present the first systematic comparison of information retained at different "representational levels'' as it is compressed from the rich information encoded in the residual stream through two natural bottlenecks: low-dimensional projections of the residual stream obtained using tuned lens, and the final top-k logits most likely to impact model's answer. We show that even easily accessible bottlenecks defined by the model's top logit values can leak task-irrelevant information present in an image-based query, in some cases revealing as much information as direct projections of the full residual stream.

In the intersection of computer vision and robotic perception, 4D reconstruction of dynamic scenes serve as the critical bridge connecting low-level geometric sensing with high-level semantic understanding. We present DINO\_4D, introducing frozen DINOv3 features as structural priors, injecting semantic awareness into the reconstruction process to effectively suppress semantic drift during dynamic tracking. Experiments on the Point Odyssey and TUM-Dynamics benchmarks demonstrate that our method maintains the linear time complexity $O(T)$ of its predecessors while significantly improving Tracking Accuracy (APD) and Reconstruction Completeness. DINO\_4D establishes a new paradigm for constructing 4D World Models that possess both geometric precision and semantic understanding.

Generative user interfaces (UIs) create new opportunities to adapt interfaces to individual users on demand, but personalization remains difficult because desirable UI properties are subjective, hard to articulate, and costly to infer from sparse feedback. We study this problem through a new dataset in which 20 trained designers each provide pairwise judgments over the same 600 generated UIs, enabling direct analysis of preference divergence. We find substantial disagreement across designers (average kappa = 0.25), and written rationales reveal that even when designers appeal to similar concepts such as hierarchy or cleanliness, designers differ in how they define, prioritize, and apply those concepts. Motivated by these findings, we develop a sample-efficient personalization method that represents a new user in terms of prior designers rather than a fixed rubric of design concepts. In a technical evaluation, our preference model outperforms both a pretrained UI evaluator and a larger multimodal model, and scales better with additional feedback. When used to personalize generation, it also produces interfaces preferred by 12 new designers over baseline approaches, including direct user prompting. Our findings suggest that lightweight preference elicitation can serve as a practical foundation for personalized generative UI systems.

Simulating how organized groups (e.g., corporations) make decisions (e.g., responding to a competitor's move) is essential for understanding real-world dynamics and could benefit relevant applications (e.g., market prediction). In this paper, we formalize this problem as a concrete research platform for group behavior understanding, providing: (1) a task definition with benchmark and evaluation criteria, (2) a structured analytical framework with a corresponding algorithm, and (3) detailed temporal and cross-group analysis. Specifically, we propose Organized Group Behavior Simulation, a task that models organized groups as collective entities from a practical perspective: given a group facing a particular situation (e.g., AI Boom), predict the decision it would take. To support this task, we present GROVE (GRoup Organizational BehaVior Evaluation), a benchmark covering 44 entities with 8,052 real-world context-decision pairs collected from Wikipedia and TechCrunch across 9 domains, with an end-to-end evaluation protocol assessing consistency, initiative, scope, magnitude, and horizon. Beyond straightforward prompting pipelines, we propose a structured analytical framework that converts collective decision-making events into an interpretable, adaptive, and traceable behavioral model, achieving stronger performance than summarization- and retrieval-based baselines. It further introduces an adapter mechanism for time-aware evolution and group-aware transfer, and traceable evidence nodes grounding each decision rule in originating historical events. Our analysis reveals temporal behavioral drift within individual groups, which the time-aware adapter effectively captures for stronger prediction, and structured cross-group similarity that enables knowledge transfer for data-scarce organizations.

We investigate how looped transformers encode human preference in their internal iteration states. Using Ouro-2.6B-Thinking, a 2.6B-parameter looped transformer with iterative refinement, we extract hidden states from each loop iteration and train lightweight evaluator heads (~5M parameters) to predict human preference on the Anthropic HH-RLHF dataset. Our pairwise evaluator achieves 95.2% test accuracy on 8,552 unseen examples, surpassing a full-batch L-BFGS probe (84.5%) while the base model remains completely frozen. Our central finding is that loop states encode preference predominantly relationally: a linear probe on pairwise differences achieves 84.5%, the best nonlinear independent evaluator reaches only 65% test accuracy, and linear independent classification scores 21.75%, below chance and with inverted polarity. Interpreted precisely, the evaluator functions as a model-internal consistency probe, measuring how stably Ouro's own learned value system organizes its representations rather than how well it predicts noisy human annotations. We also document a systematic architecture search that established a genuine 70% ceiling for independent scoring, and show how the 50% argument-swap protocol required to prevent degenerate pairwise solutions deflated pairwise training metrics by about 31 points at peak, creating the false appearance that pairwise and pointwise evaluators shared the same ceiling. Finally, we show that a cosine learning-rate dead zone at epoch 2 accidentally acted as early stopping, preserving the generalization peak before overfitting degraded test accuracy from 95.2% to 62.4% by epoch 5. Cross-epoch flip-test analysis shows that antisymmetry correlation remains stable while strict sign-flip rate mainly tracks scorer bias. We propose the flip test as a mandatory diagnostic for pairwise preference evaluators.

Recent advances in large language models (LLMs) have shown the promise to significantly accelerate the workflow by automating structural modeling and analysis. However, existing studies primarily focus on enabling LLMs to operate a single structural analysis software platform. In practice, structural engineers often rely on multiple finite element analysis (FEA) tools, such as ETABS, SAP2000, and OpenSees, depending on project needs, user preferences, and company constraints. This limitation restricts the practical deployment of LLM-assisted engineering workflows. To address this gap, this study develops LLMs capable of automating frame structural analysis across multiple software platforms. The LLMs adopt a two-stage multi-agent architecture. In Stage 1, a cohort of agents collaboratively interpret user input and perform structured reasoning to infer geometric, material, boundary, and load information required for finite element modeling. The outputs of these agents are compiled into a unified JSON representation. In Stage 2, code translation agents operate in parallel to convert the JSON file into executable scripts across multiple structural analysis platforms. Each agent is prompted with the syntax rules and modeling workflows of its target software. The LLMs are evaluated using 20 representative frame problems across three widely used platforms: ETABS, SAP2000, and OpenSees. Results from ten repeated trials demonstrate consistently reliable performance, achieving accuracy exceeding 90% across all cases.

The goal of domain adaptation is to make predictions for unlabeled samples from a target domain with the help of labeled samples from a different but related source domain. The performance of domain adaptation methods is highly influenced by the choice of source domain and pre-trained feature extractor. However, the selection of source data and pre-trained model is not trivial due to the absence of a labeled validation set for the target domain and the large number of available pre-trained models. In this work, we propose PAS, a novel score designed to estimate the transferability of a source domain set and a pre-trained feature extractor to a target classification task before actually performing domain adaptation. PAS leverages the generalization power of pre-trained models and assesses source-target compatibility based on the pre-trained feature embeddings. We integrate PAS into a framework that indicates the most relevant pre-trained model and source domain among multiple candidates, thus improving target accuracy while reducing the computational overhead. Extensive experiments on image classification benchmarks demonstrate that PAS correlates strongly with actual target accuracy and consistently guides the selection of the best-performing pre-trained model and source domain for adaptation.

Text-to-image diffusion models exhibit strong generative performance but remain highly sensitive to prompt formulation, often requiring extensive manual trial and error to obtain satisfactory results. This motivates the development of automated, model-agnostic prompt optimization methods that can systematically explore the conditioning space beyond conventional text rewriting. This work investigates the use of a Genetic Algorithm (GA) for prompt optimization by directly evolving the token vectors employed by CLIP-based diffusion models. The GA optimizes a fitness function that combines aesthetic quality, measured by the LAION Aesthetic Predictor V2, with prompt-image alignment, assessed via CLIPScore. Experiments on 36 prompts from the Parti Prompts (P2) dataset show that the proposed approach outperforms the baseline methods, including Promptist and random search, achieving up to a 23.93% improvement in fitness. Overall, the method is adaptable to image generation models with tokenized text encoders and provides a modular framework for future extensions, the limitations and prospects of which are discussed.

The pursuit of general-purpose robotics has yielded impressive foundation models, yet simulation-based benchmarking remains a bottleneck due to rapid performance saturation and a lack of true generalization testing. Existing benchmarks often exhibit significant domain overlap between training and evaluation, trivializing success rates and obscuring insights into robustness. We introduce RoboLab, a simulation benchmarking framework designed to address these challenges. Concretely, our framework is designed to answer two questions: (1) to what extent can we understand the performance of a real-world policy by analyzing its behavior in simulation, and (2) which external factors most strongly affect that behavior under controlled perturbations. First, RoboLab enables human-authored and LLM-enabled generation of scenes and tasks in a robot- and policy-agnostic manner within a physically realistic and photorealistic simulation. With this, we propose the RoboLab-120 benchmark, consisting of 120 tasks categorized into three competency axes: visual, procedural, relational competency, across three difficulty levels. Second, we introduce a systematic analysis of real-world policies that quantify both their performance and the sensitivity of their behavior to controlled perturbations, indicating that high-fidelity simulation can serve as a proxy for analyzing performance and its dependence on external factors. Evaluation with RoboLab exposes significant performance gap in current state-of-the-art models. By providing granular metrics and a scalable toolset, RoboLab offers a scalable framework for evaluating the true generalization capabilities of task-generalist robotic policies.

The recent advancement of Large Language Models (LLMs) has established their potential as autonomous interactive agents. However, they often struggle in strategic games of incomplete information, such as bilateral price negotiation. In this paper, we investigate if Reinforcement Learning from Verifiable Rewards (RLVR) can effectively teach LLMs to negotiate. Specifically, we explore the strategic behaviors that emerge during the learning process. We introduce a framework that trains a mid-sized buyer agent against a regulated LLM seller across a wide distribution of real-world products. By grounding reward signals directly in the maximization of economic surplus and strict adherence to private budget constraints, we reveal a novel four-phase strategic evolution. The agent progresses from naive bargaining to using aggressive starting prices, moves through a phase of deadlock, and ultimately develops sophisticated persuasive skills. Our results demonstrate that this verifiable training allows a 30B agent to significantly outperform frontier models over ten times its size in extracting surplus. Furthermore, the trained agent generalizes robustly to stronger counterparties unseen during training and remains effective even when facing hostile, adversarial seller personas.

LLMs have so far failed both to generate consistently compelling stories and to recognize this failure--on the leading creative-writing benchmark (EQ-Bench), LLM judges rank zero-shot AI stories above New Yorker short stories, a gold standard for literary fiction. We argue that existing rubrics overlook a key dimension of compelling human stories: narrative tension. We introduce the 100-Endings metric, which walks through a story sentence by sentence: at each position, a model predicts how the story will end 100 times given only the text so far, and we measure tension as how often predictions fail to match the ground truth. Beyond the mismatch rate, the sentence-level curve yields complementary statistics, such as inflection rate, a geometric measure of how frequently the curve reverses direction, tracking twists and revelations. Unlike rubric-based judges, 100-Endings correctly ranks New Yorker stories far above LLM outputs. Grounded in narratological principles, we design a story-generation pipeline using structural constraints, including analysis of story templates, idea formulation, and narrative scaffolding. Our pipeline significantly increases narrative tension as measured by the 100-Endings metric, while maintaining performance on the EQ-Bench leaderboard.

Reasoning models think in long, unstructured streams with no mechanism for compressing or organizing their own intermediate state. We introduce MEMENTO: a method that teaches models to segment reasoning into blocks, compress each block into a memento, i.e., a dense state summary, and reason forward by attending only to mementos, reducing context, KV cache, and compute. To train MEMENTO models, we release OpenMementos, a public dataset of 228K reasoning traces derived from OpenThoughts-v3, segmented and annotated with intermediate summaries. We show that a two-stage SFT recipe on OpenMementos is effective across different model families (Qwen3, Phi-4, Olmo 3) and scales (8B--32B parameters). Trained models maintain strong accuracy on math, science, and coding benchmarks while achieving ${\sim}2.5\times$ peak KV cache reduction. We extend vLLM to support our inference method, achieving ${\sim}1.75\times$ throughput improvement while also enabling us to perform RL and further improve accuracy. Finally, we identify a dual information stream: information from each reasoning block is carried both by the memento text and by the corresponding KV states, which retain implicit information from the original block. Removing this channel drops accuracy by 15\,pp on AIME24.

Vision-language models (VLMs) are increasingly adapted through domain-specific fine-tuning, yet it remains unclear whether this improves reasoning beyond superficial visual cues, particularly in high-stakes domains like medicine. We evaluate four paired open-source VLMs (LLaVA vs. LLaVA-Med; Gemma vs. MedGemma) across four medical imaging tasks of increasing difficulty: brain tumor, pneumonia, skin cancer, and histopathology classification. We find that performance degrades toward near-random levels as task difficulty increases, indicating limited clinical reasoning. Medical fine-tuning provides no consistent advantage, and models are highly sensitive to prompt formulation, with minor changes causing large swings in accuracy and refusal rates. To test whether closed-form VQA suppresses latent knowledge, we introduce a description-based pipeline where models generate image descriptions that a text-only model (GPT-5.1) uses for diagnosis. This recovers a limited additional signal but remains bounded by task difficulty. Analysis of vision encoder embeddings further shows that failures stem from both weak visual representations and downstream reasoning. Overall, medical VLM performance is fragile, prompt-dependent, and not reliably improved by domain-specific fine-tuning.

Activation steering is a popular white-box control technique that modifies model activations to elicit an abstract change in output behavior. It has also become a standard tool in interpretability (e.g., probing truthfulness, or translating activations into human-readable explanations and safety research (e.g., studying jailbreakability). However, it is unclear whether steered activation states are realizable by any textual prompt. In this work, we cast this question as a surjectivity problem: for a fixed model, does every steered activation admit a pre-image under the model's natural forward pass? Under practical assumptions, we prove that activation steering pushes the residual stream off the manifold of states reachable from discrete prompts. Almost surely, no prompt can reproduce the same internal behavior induced by steering. We also illustrate this finding empirically across three widely used LLMs. Our results establish a formal separation between white-box steerability and black-box prompting. We therefore caution against interpreting the ease and success of activation steering as evidence of prompt-based interpretability or vulnerability, and argue for evaluation protocols that explicitly decouple white-box and black-box interventions.

AI agents hold growing promise for accelerating scientific discovery; yet, a lack of frontier evaluations hinders adoption into real workflows. Expert-written benchmarks have proven effective at measuring AI reasoning, but most at this stage have become saturated and only measure performance on constrained outputs. To help address this gap, we introduce COMPOSITE-STEM, a benchmark of 70 expert-written tasks in physics, biology, chemistry, and mathematics, curated by doctoral-level researchers. Our benchmark combines exact-match grading and criterion-based rubrics with an LLM-as-a-jury grading protocol, allowing more flexible assessment of scientifically meaningful outputs. Using an adapted multimodal Terminus-2 agent harness within the Harbor agentic evaluation framework, we evaluate four frontier models. The top-performing model achieves 21%, demonstrating that COMPOSITE-STEM captures capabilities beyond current agent reach. All tasks are open-sourced with contributor permission to support reproducibility and to promote additional research towards AI's acceleration of scientific progress in these domains.

Existing full-body Gaussian avatar methods primarily optimize global reconstruction quality and often fail to preserve fine-grained facial geometry and expression details. This challenge arises from limited facial representational capacity that causes difficulties in modeling high-frequency pose-dependent deformations. To address this, we propose F3G-Avatar, a full-body, face-aware avatar synthesis method that reconstructs animatable human representations from multi-view RGB video and regressed pose/shape parameters. Starting from a clothed Momentum Human Rig (MHR) template, front/back positional maps are rendered and decoded into 3D Gaussians through a two-branch architecture: a body branch that captures pose-dependent non-rigid deformations and a face-focused deformation branch that refines head geometry and appearance. The predicted Gaussians are fused, posed with linear blend skinning (LBS), and rendered with differentiable Gaussian splatting. Training combines reconstruction and perceptual objectives with a face-specific adversarial loss to enhance realism in close-up views. Experiments demonstrate strong rendering quality, with face-view performance reaching PSNR/SSIM/LPIPS of 26.243/0.964/0.084 on the AvatarReX dataset. Ablations further highlight contributions of the MHR template and the face-focused deformation. F3G-Avatar provides a practical, high-quality pipeline for realistic, animatable full-body avatar synthesis.

Physical implementations of neural computation now extend far beyond silicon hardware, encompassing substrates such as memristive devices, photonic circuits, mechanical metamaterials, microfluidic networks, chemical reaction systems, and living neural tissue. By exploiting intrinsic physical processes such as charge transport, wave interference, elastic deformation, mass transport, and biochemical regulation, these substrates can realize neural inference and adaptation directly in matter. As silicon GPU-centered AI faces growing energy and data-movement constraints, physical neural computation is becoming increasingly relevant as a complementary path beyond conventional digital accelerators. This trend is driven in particular by pervasive intelligence, i.e., the deployment of on-device and edge AI across large numbers of resource-constrained systems. In such settings, co-locating computation with sensing and memory can reduce data shuttling and improve efficiency. Meanwhile, physical neural approaches have emerged across disparate disciplines, yet progress remains fragmented, with limited shared terminology and few principled ways to compare platforms. This survey unifies the field by mapping neural primitives to substrate-specific mechanisms, analyzing architectural and training paradigms, and identifying key engineering constraints including scalability, precision, programmability, and I/O interfacing overhead. To enable cross-domain comparison, we introduce a first-order benchmarking scheme based on standardized static and dynamic tasks and physically interpretable performance dimensions. We show that no single substrate dominates across the considered dimensions; instead, physical neural systems occupy complementary operating regimes, enabling applications ranging from ultrafast signal processing and in-memory inference to embodied control and in-sample biochemical decision making.

Vision language action (VLA) models enable generalist robotic agents but often exhibit language ignorance, relying on visual shortcuts and remaining insensitive to instruction changes. We present Prospective Grounding and Alignment VLA (ProGAL-VLA), which constructs a 3D entity-centric graph (GSM), uses a slow planner to produce symbolic sub-goals, and aligns them with grounded entities via a Grounding Alignment Contrastive (GAC) loss. All actions are conditioned on a verified goal embedding $g_t$, whose attention entropy provides an intrinsic ambiguity signal. On LIBERO-Plus, ProGAL-VLA increases robustness under robot perturbations from 30.3 to 71.5 percent, reduces language ignorance by 3x-4x, and improves entity retrieval from 0.41 to 0.71 Recall@1. On the Custom Ambiguity Benchmark, it reaches AUROC 0.81 (vs. 0.52), AUPR 0.79, and raises clarification on ambiguous inputs from 0.09 to 0.81 without harming unambiguous success. The verification bottleneck increases mutual information of language-actions, the GAC loss imposes an entity-level InfoNCE bound, and attention entropy yields calibrated selective prediction, indicating that explicit verified grounding is an effective path toward instruction-sensitive, ambiguity-aware agents.

We introduce ACCIDENT, a benchmark dataset for traffic accident detection in CCTV footage, designed to evaluate models in supervised (IID and OOD) and zero-shot settings, reflecting both data-rich and data-scarce scenarios. The benchmark consists of a curated set of 2,027 real and 2,211 synthetic clips annotated with the accident time, spatial location, and high-level collision type. We define three core tasks: (i) temporal localization of the accident, (ii) its spatial localization, and (iii) collision type classification. Each task is evaluated using custom metrics that account for the uncertainty and ambiguity inherent in CCTV footage. In addition to the benchmark, we provide a diverse set of baselines, including heuristic, motion-aware, and vision-language approaches, and show that ACCIDENT is challenging. You can access the ACCIDENT at: https://accidentbench.github.io

Mycorrhizal fungi are vital to terrestrial ecosystem functioning. Yet monitoring their biodiversity at landscape scales is often unfeasible due to time and cost constraints. Current predictions suggest that 90\% of mycorrhizal diversity hotspots remain unprotected, opening questions of how to broadly and effectively map underground fungal communities. Here, we show that self-supervised learning (SSL) applied to satellite imagery can predict below-ground ectomycorrhizal fungal richness across diverse environments. Our models explain over half the variance in species richness across ~12,000 field samples spanning Europe and Asia. SSL-derived features prove to be the single most informative predictor, subsuming the majority of information contained in climate, soil, and land cover datasets. Using this approach, we achieve a 10,000-fold increase in spatial resolution over existing techniques, moving from 1km landscape averages to 10m habitat-scale observations with nearly no systematic bias. As satellite observations are dynamic rather than static, this enables temporal monitoring of below-ground biodiversity at landscape scales for the first time. We analyze multi-year trends in predicted fungal richness across UK National Park woodlands, finding that ancient forests may be losing ectomycorrhizal diversity at disproportionate rates. These results establish SSL satellite features as a scalable tool for extending sparse field observations to continuous, high-resolution biodiversity maps for monitoring the invisible half of terrestrial ecosystems.

Visual encoding and decoding models act as gateways to understanding the neural mechanisms underlying human visual perception. Typically, visual encoding models that predict brain activity from stimuli and decoding models that reproduce stimuli from brain activity are treated as distinct tasks, requiring separate models and training procedures. This separation is inefficient and fails to model the consistency between encoding and decoding processes. To address this limitation, we propose NeuroFlow, the first unified framework that jointly models visual encoding and decoding from neural activity within a single flow model. NeuroFlow introduces two key components: (1) NeuroVAE is designed as a variational backbone to model neural variability and establish a compact, semantically structured latent space for bidirectional modeling across visual and neural modalities. (2) Cross-modal Flow Matching (XFM) bypasses the typical paradigm of noise-to-data diffusion guided by a specific modality condition, instead learning a reversibly consistent flow model between visual and neural latent distributions. For the first time, visual encoding and decoding are reformulated as a time-dependent, reversible process within a shared latent space for unified modeling. Empirical results demonstrate that NeuroFlow achieves superior overall performance in visual encoding and decoding tasks with higher computational efficiency compared to any isolated methods. We further analyze principal factors that steer the model toward encoding-decoding consistency and, through brain functional analyses, demonstrate that NeuroFlow captures consistent activation patterns underlying neural variability. NeuroFlow marks a major step toward unified visual encoding and decoding from neural activity, providing mechanistic insights that inform future bidirectional visual brain-computer interfaces.

Computer-use agents that combine GUI interaction with structured API calls via the Model Context Protocol (MCP) show promise for automating software tasks. However, existing approaches lack a principled understanding of how agents should balance these two modalities and how to enable iterative self-improvement across diverse applications. We formulate MCP-GUI interplay as a unified hybrid policy learning problem where the agent learns when each modality provides complementary advantages, and show that distillation and experience augmentation target fundamentally different failure modes - requiring application-aware mechanism selection. Built on this formulation, we propose a self-evolving framework with a fully automatic pipeline that orchestrates automatic environment generation and validation, trajectory collection, gap-driven task synthesis, and quality-filtered training - all without manual intervention. A key innovation is our experience bank, which accumulates LLM-learned rules from trajectory comparison, enabling inference-time improvement without fine-tuning. Systematic \textbf{cross-application analysis} across three desktop applications reveals that the optimal strategy depends on MCP-GUI composition: distillation achieves 77.8\% pass rate on MCP-dominant tasks (+17.8pp), while the experience bank excels on GUI-intensive tasks (+10.0pp).

Existing synthetic tool-use corpora are primarily designed for offline supervised fine-tuning, yet reinforcement learning (RL) requires executable environments that support reward-checkable online rollouts. We propose COVERT, a two-stage pipeline that first generates reliable base tool-use trajectories through self-evolving synthesis with multi-level validation, and then applies oracle-preserving augmentations that systematically increase environmental complexity. These augmentations introduce distractor tools, indirect or ambiguous user queries, and noisy, multi-format, or erroneous tool outputs, while strictly preserving oracle tool calls and final answers as ground truth. This design enables automatic reward computation via reference matching for standard cases and lightweight judge-assisted verification for special behaviors such as error detection, supporting RL optimization of tool-calling policies. On Qwen2.5-Instruct-14B, COVERT-RL improves overall accuracy on BFCL v3 from 56.5 to 59.9 and on ACEBench from 53.0 to 59.3, with minimal regressions on general-ability benchmarks; when stacked on SFT, it further reaches 62.1 and 61.8, confirming additive gains. These results suggest that oracle-preserving synthetic environments offer a practical RL refinement stage, complementary to SFT, for improving tool-use robustness under ambiguity and unreliable tool feedback.

Recurrent claims present a major challenge for automated fact-checking systems designed to combat misinformation, especially in multilingual settings. While tasks such as claim matching and fact-checked claim retrieval aim to address this problem by linking claim pairs, the broader challenge of effectively representing groups of similar claims that can be resolved with the same fact-check via claim clustering remains relatively underexplored. To address this gap, we introduce Claim2Vec, the first multilingual embedding model optimized to represent fact-check claims as vectors in an improved semantic embedding space. We fine-tune a multilingual encoder using contrastive learning with similar multilingual claim pairs. Experiments on the claim clustering task using three datasets, 14 multilingual embedding models, and 7 clustering algorithms demonstrate that Claim2Vec significantly improves clustering performance. Specifically, it enhances both cluster label alignment and the geometric structure of the embedding space across different cluster configurations. Our multilingual analysis shows that clusters containing multiple languages benefit from fine-tuning, demonstrating cross-lingual knowledge transfer.

Enterprise teams building internal coding agents face a gap between prototype performance and production readiness. The root cause is that technical model quality alone is insufficient -- tool design, safety enforcement, state management, and human trust calibration are equally decisive, yet underreported in the literature. We present CodeGen, an internal coding agent at Zup, and show that targeted tool design (e.g., string-replacement edits over full-file rewrites) and layered safety guardrails improved agent reliability more than prompt engineering, while progressive human oversight modes drove organic adoption without mandating trust. These findings suggest that the engineering decisions surrounding the model -- not the model itself -- determine whether a coding agent delivers real value in practice.

Human activity recognition (HAR) has become a key component of intelligent systems for healthcare monitoring, assistive living, smart environments, and human-computer interaction. Although deep learning has substantially improved HAR performance on multivariate sensor data, the resulting models often remain opaque, limiting trust, reliability, and real-world deployment. Explainable artificial intelligence (XAI) has therefore emerged as a critical direction for making HAR systems more transparent and human-centered. This paper presents a comprehensive review of explainable HAR methods across wearable, ambient, physiological, and multimodal sensing settings. We introduce a unified perspective that separates conceptual dimensions of explainability from algorithmic explanation mechanisms, reducing ambiguities in prior surveys. Building on this distinction, we present a mechanism-centric taxonomy of XAI-HAR methods covering major explanation paradigms. The review examines how these methods address the temporal, multimodal, and semantic complexities of HAR, and summarize their interpretability objectives, explanation targets, and limitations. In addition, we discuss current evaluation practices, highlight key challenges in achieving reliable and deployable XAI-HAR, and outline directions toward trustworthy activity recognition systems that better support human understanding and decision-making.

Scientific breakthroughs often emerge from synthesizing prior ideas into novel contributions. While language models (LMs) show promise in scientific discovery, their ability to perform this targeted, literature-grounded synthesis remains underexplored. We introduce insight anticipation, a generation task in which a model predicts a downstream paper's core insight from its foundational parent papers. To evaluate this capability, we develop GiantsBench, a benchmark of 17k examples across eight scientific domains, where each example consists of a set of parent papers paired with the core insight of a downstream paper. We evaluate models using an LM judge that scores similarity between generated and ground-truth insights, and show that these similarity scores correlate with expert human ratings. Finally, we present GIANTS-4B, an LM trained via reinforcement learning (RL) to optimize insight anticipation using these similarity scores as a proxy reward. Despite its smaller open-source architecture, GIANTS-4B outperforms proprietary baselines and generalizes to unseen domains, achieving a 34% relative improvement in similarity score over gemini-3-pro. Human evaluations further show that GIANTS-4B produces insights that are more conceptually clear than those of the base model. In addition, SciJudge-30B, a third-party model trained to compare research abstracts by likely citation impact, predicts that insights generated by GIANTS-4B are more likely to lead to higher citations, preferring them over the base model in 68% of pairwise comparisons. We release our code, benchmark, and model to support future research in automated scientific discovery.

Small language models are attractive for production deployment due to their low cost, fast inference, and ease of specialization. However, adapting them to a specific task remains a challenging engineering loop, driven not by training itself but by surrounding decisions: data curation, failure diagnosis, regression avoidance, and iteration control. We present Pioneer Agent, a closed-loop system that automates this lifecycle. In cold-start mode, given only a natural-language task description, the agent acquires data, constructs evaluation sets, and iteratively trains models by jointly optimizing data, hyperparameters, and learning strategy. In production mode, given a deployed model with labeled failures, it diagnoses error patterns, constructs targeted training data, and retrains under explicit regression constraints. To evaluate this setting, we introduce AdaptFT-Bench, a benchmark of synthetic inference logs with progressively increasing noise, designed to test the full adaptation loop: diagnosis, curriculum synthesis, retraining, and verification. Across eight cold-start benchmarks spanning reasoning, math, code generation, summarization, and classification, Pioneer Agent improves over base models by 1.6-83.8 points. On AdaptFT-Bench, it improves or preserves performance in all seven scenarios, while naive retraining degrades by up to 43 points. On two production-style deployments built from public benchmark tasks, it raises intent classification from 84.9% to 99.3% and Entity F1 from 0.345 to 0.810. Beyond performance gains, the agent often discovers effective training strategies, including chain-of-thought supervision, task-specific optimization, and quality-focused data curation, purely from downstream feedback.

Data collected from the physical world is always a combination of multiple sources: an underlying signal from the physical process of interest and a signal from measurement-dependent artifacts from the sensor or instrument. This secondary signal acts as a confounding factor, limiting our ability to extract information about the physics underlying the phenomena we observe. Furthermore, it complicates the combination of observations in heterogeneous or multi-instrument settings. We propose a deep learning framework that leverages overlapping observations, a dual-encoder architecture, and a counterfactual generation objective to disentangle these factors of variation. The resulting representations explicitly separate intrinsic signals from sensor-specific distortions and noise, and can be used for counterfactual view generation, parameter inference unconfounded by measurement distortions, and instrument-independent similarity search. We demonstrate the effectiveness of our approach on astrophysical galaxy images from the DESI Legacy Imaging Survey (Legacy) and the Hyper Suprime-Cam (HSC) Survey as a representative multi-instrument setting. This framework provides a general recipe for scientific and multi-modal self-supervised pretraining: construct training pairs from overlapping observations of the same physical system, treat sensor- or modality-specific effects as augmentations, and learn invariant representations through counterfactual generation.

The sequential nature of autoregressive next-token prediction imposes a fundamental speed limit on large language models. While continuous flow models offer a path to parallel generation, they traditionally demand expensive iterative integration. Flow Maps bypass this bottleneck by compressing generative trajectories into single-step mappings, theoretically enabling the generation of full text sequences from noise in a single forward pass. However, standard formulations rely on Euclidean regression losses that are geometrically ill-suited for discrete data. In this work, we resolve this conflict with Discrete Flow Maps, a framework that reconciles trajectory compression with the geometry of the probability simplex. We recast standard flow map training for the discrete domain, aligning the training dynamics with the discrete nature of language. Empirically, this strict geometric alignment allows our method to surpass previous state-of-the-art results in discrete flow modeling.

Mixture of Experts (MoEs) are now ubiquitous in large language models, yet the mechanisms behind their "expert specialization" remain poorly understood. We show that, since MoE routers are linear maps, hidden state similarity is both necessary and sufficient to explain expert usage similarity, and specialization is therefore an emergent property of the representation space, not of the routing architecture itself. We confirm this at both token and sequence level across five pre-trained models. We additionally prove that load-balancing loss suppresses shared hidden state directions to maintain routing diversity, which might provide a theoretical explanation for specialization collapse under less diverse data, e.g. small batch. Despite this clean mechanistic account, we find that specialization patterns in pre-trained MoEs resist human interpretation: expert overlap between different models answering the same question is no higher than between entirely different questions ($\sim$60\%); prompt-level routing does not predict rollout-level routing; and deeper layers exhibit near-identical expert activation across semantically unrelated inputs, especially in reasoning models. We conclude that, while the efficiency perspective of MoEs is well understood, understanding expert specialization is at least as hard as understanding LLM hidden state geometry, a long-standing open problem in the literature.

Rare events are central to the evolution of complex many-body systems, characterized as key transitional configurations on the free energy surface (FES). Conventional methods require adequate sampling of rare event transitions to obtain the FES, which is computationally very demanding. Here we introduce the variational free energy surface (VaFES), a dataset-free framework that directly models FESs using tractable-density generative models. Rare events can then be immediately identified from the FES with their configurations generated directly via one-shot sampling of generative models. By extending a coarse-grained collective variable (CV) into its reversible equivalent, VaFES constructs a latent space of intermediate representation in which the CVs explicitly occupy a subset of dimensions. This latent-space construction preserves the physical interpretability and transparent controllability of the CVs by design, while accommodating arbitrary CV formulations. The reversibility makes the system energy exactly accessible, enabling variational optimization of the FES without pre-generated simulation data. A single optimization yields a continuous, differentiable FES together with one-shot generation of rare-event configurations. Our method can reproduce the exact analytical solution for the bistable dimer potential and identify a chignolin native folded state in close alignment with the experimental NMR structure. Our approach thus establishes a scalable, systematic framework for advancing the study of complex statistical systems.

Large language models (LLMs) undergo alignment training to avoid harmful behaviors, yet the resulting safeguards remain brittle: jailbreaks routinely bypass them, and fine-tuning on narrow domains can induce ``emergent misalignment'' that generalizes broadly. Whether this brittleness reflects a fundamental lack of coherent internal organization for harmfulness remains unclear. Here we use targeted weight pruning as a causal intervention to probe the internal organization of harmfulness in LLMs. We find that harmful content generation depends on a compact set of weights that are general across harm types and distinct from benign capabilities. Aligned models exhibit a greater compression of harm generation weights than unaligned counterparts, indicating that alignment reshapes harmful representations internally--despite the brittleness of safety guardrails at the surface level. This compression explains emergent misalignment: if weights of harmful capabilities are compressed, fine-tuning that engages these weights in one domain can trigger broad misalignment. Consistent with this, pruning harm generation weights in a narrow domain substantially reduces emergent misalignment. Notably, LLMs harmful generation capability is dissociated from how they recognize and explain such content. Together, these results reveal a coherent internal structure for harmfulness in LLMs that may serve as a foundation for more principled approaches to safety.

The persistent storage requirements for high-resolution, spatiotemporally evolving fields governed by large-scale and high-dimensional partial differential equations (PDEs) have reached the petabyte-to-exabyte scale. Transient simulations modeling Navier-Stokes equations, magnetohydrodynamics, plasma physics, or binary black hole mergers generate data volumes that are prohibitive for modern high-performance computing (HPC) infrastructures. To address this bottleneck, we introduce ANTIC (Adaptive Neural Temporal in situ Compressor), an end-to-end in situ compression pipeline. ANTIC consists of an adaptive temporal selector tailored to high-dimensional physics that identifies and filters informative snapshots at simulation time, combined with a spatial neural compression module based on continual fine-tuning that learns residual updates between adjacent snapshots using neural fields. By operating in a single streaming pass, ANTIC enables a combined compression of temporal and spatial components and effectively alleviates the need for explicit on-disk storage of entire time-evolved trajectories. Experimental results demonstrate how storage reductions of several orders of magnitude relate to physics accuracy.

Molecular docking is a crucial step in the development of new drugs as it guides the positioning of a small molecule (ligand) within the pocket of a target protein. In the literature, a feasibility study explored the potential of D-Wave quantum annealers for purely geometric molecular docking, neglecting physicochemical interactions between the protein and the ligand and focusing solely on their simplified geometries. To achieve this, the ligands were represented as graphs incorporating their geometric properties and then mapped onto a grid that discretized the three-dimensional space of the protein pocket. The quality of the ligand pose on the protein pocket was evaluated through the isomorphism between the ligand graph and the spatial grid. This paper builds on the previous study by introducing physicochemical interactions between the protein-ligand pair into the QUBO problem to improve the accuracy of the docking results. This paper presents a novel QUBO formulation that includes Coulomb and van der Waals forces, together with components representing H-bond and hydrophobic interactions. We integrate these physical interactions as corrective terms to the previous purely geometric QUBO formulation, and provide experimental results using the D-Wave quantum annealers to demonstrate their impact on the accuracy of the docking results.

Evidence-grounded reasoning requires more than attaching retrieved text to a prediction: a model should make decisions that depend on whether the provided evidence supports the target claim. In practice, this often fails because supervision is weak, evidence is only loosely tied to the claim, and evaluation does not test evidence dependence directly. We introduce case-grounded evidence verification, a general framework in which a model receives a local case context, external evidence, and a structured claim, and must decide whether the evidence supports the claim for that case. Our key contribution is a supervision construction procedure that generates explicit support examples together with semantically controlled non-support examples, including counterfactual wrong-state and topic-related negatives, without manual evidence annotation. We instantiate the framework in radiology and train a standard verifier on the resulting support task. The learned verifier substantially outperforms both case-only and evidence-only baselines, remains strong under correct evidence, and collapses when evidence is removed or swapped, indicating genuine evidence dependence. This behavior transfers across unseen evidence articles and an external case distribution, though performance degrades under evidence-source shift and remains sensitive to backbone choice. Overall, the results suggest that a major bottleneck in evidence grounding is not only model capacity, but the lack of supervision that encodes the causal role of evidence.

Prompt learning is a parameter-efficient approach for vision-language models, yet its robustness under label noise is less investigated. Visual content contains richer and more reliable semantic information, which remains more robust under label noise. However, the prompt itself is highly susceptible to label noise. Motivated by this intuition, we propose VisPrompt, a lightweight and robust vision-guided prompt learning framework for noisy-label settings. Specifically, we exploit a cross-modal attention mechanism to reversely inject visual semantics into prompt representations. This enables the prompt tokens to selectively aggregate visual information relevant to the current sample, thereby improving robustness by anchoring prompt learning to stable instance-level visual evidence and reducing the influence of noisy supervision. To address the instability caused by using the same way of injecting visual information for all samples, despite differences in the quality of their visual cues, we further introduce a lightweight conditional modulation mechanism to adaptively control the strength of visual information injection, which strikes a more robust balance between text-side semantic priors and image-side instance evidence. The proposed framework effectively suppresses the noise-induced disturbances, reduce instability in prompt updates, and alleviate memorization of mislabeled samples. VisPrompt significantly improves robustness while keeping the pretrained VLM backbone frozen and introducing only a small amount of additional trainable parameters. Extensive experiments under synthetic and real-world label noise demonstrate that VisPrompt generally outperforms existing baselines on seven benchmark datasets and achieves stronger robustness. Our code is publicly available at https://github.com/gezbww/Vis_Prompt.

Vision-language models (VLMs) still struggle with visual perception tasks such as spatial understanding and viewpoint recognition. One plausible contributing factor is that natural image datasets provide limited supervision for low-level visual skills. This motivates a practical question: can targeted synthetic supervision, generated from only a task keyword such as Depth Order, address these weaknesses? To investigate this question, we introduce VisionFoundry, a task-aware synthetic data generation pipeline that takes only the task name as input and uses large language models (LLMs) to generate questions, answers, and text-to-image (T2I) prompts, then synthesizes images with T2I models and verifies consistency with a proprietary VLM, requiring no reference images or human annotation. Using VisionFoundry, we construct VisionFoundry-10K, a synthetic visual question answering (VQA) dataset containing 10k image-question-answer triples spanning 10 tasks. Models trained on VisionFoundry-10K achieve substantial improvements on visual perception benchmarks: +7% on MMVP and +10% on CV-Bench-3D, while preserving broader capabilities and showing favorable scaling behavior as data size increases. Our results suggest that limited task-targeted supervision is an important contributor to this bottleneck and that synthetic supervision is a promising path toward more systematic training for VLMs.

Large Vision Language Models (LVLMs) achieve strong multimodal reasoning but frequently exhibit hallucinations and incorrect responses with high certainty, which hinders their usage in high-stakes domains. Existing verbalized confidence calibration methods, largely developed for text-only LLMs, typically optimize a single holistic confidence score using binary answer-level correctness. This design is mismatched to LVLMs: an incorrect prediction may arise from perceptual failures or from reasoning errors given correct perception, and a single confidence conflates these sources while visual uncertainty is often dominated by language priors. To address these issues, we propose VL-Calibration, a reinforcement learning framework that explicitly decouples confidence into visual and reasoning confidence. To supervise visual confidence without ground-truth perception labels, we introduce an intrinsic visual certainty estimation that combines (i) visual grounding measured by KL-divergence under image perturbations and (ii) internal certainty measured by token entropy. We further propose token-level advantage reweighting to focus optimization on tokens based on visual certainty, suppressing ungrounded hallucinations while preserving valid perception. Experiments on thirteen benchmarks show that VL-Calibration effectively improves calibration while boosting visual reasoning accuracy, and it generalizes to out-of-distribution benchmarks across model scales and architectures.

Accurately anticipating how complex, diverse scenes will evolve requires models that represent uncertainty, simulate along extended interaction chains, and efficiently explore many plausible futures. Yet most existing approaches rely on dense video or latent-space prediction, expending substantial capacity on dense appearance rather than on the underlying sparse trajectories of points in the scene. This makes large-scale exploration of future hypotheses costly and limits performance when long-horizon, multi-modal motion is essential. We address this by formulating the prediction of open-set future scene dynamics as step-wise inference over sparse point trajectories. Our autoregressive diffusion model advances these trajectories through short, locally predictable transitions, explicitly modeling the growth of uncertainty over time. This dynamics-centric representation enables fast rollout of thousands of diverse futures from a single image, optionally guided by initial constraints on motion, while maintaining physical plausibility and long-range coherence. We further introduce OWM, a benchmark for open-set motion prediction based on diverse in-the-wild videos, to evaluate accuracy and variability of predicted trajectory distributions under real-world uncertainty. Our method matches or surpasses dense simulators in predictive accuracy while achieving orders-of-magnitude higher sampling speed, making open-set future prediction both scalable and practical. Project page: http://compvis.github.io/myriad.

The transition of Multi-Agent Reinforcement Learning (MARL) policies from simulated cyber wargames to operational Security Operations Centers (SOCs) is fundamentally bottlenecked by the Sim2Real gap. Legacy simulators abstract away network protocol physics, rely on synchronous ticks, and provide clean state vectors rather than authentic, noisy telemetry. To resolve these limitations, we introduce NetForge_RL: a high-fidelity cyber operations simulator that reformulates network defense as an asynchronous, continuous-time Partially Observable Semi-Markov Decision Process (POSMDP). NetForge enforces Zero-Trust Network Access (ZTNA) constraints and requires defenders to process NLP-encoded SIEM telemetry. Crucially, NetForge bridges the Sim2Real gap natively via a dual-mode engine, allowing high-throughput MARL training in a mock hypervisor and zero-shot evaluation against live exploits in a Docker hypervisor. To navigate this continuous-time POSMDP, we propose Continuous-Time Graph MARL (CT-GMARL), utilizing fixed-step Neural Ordinary Differential Equations (ODEs) to process irregularly sampled alerts. We evaluate our framework against discrete baselines (R-MAPPO, QMIX). Empirical results demonstrate that CT-GMARL achieves a converged median Blue reward of 57,135 - a 2.0x improvement over R-MAPPO and 2.1x over QMIX. Critically, CT-GMARL restores 12x more compromised services than the strongest baseline by avoiding the "scorched earth" failure mode of trivially minimizing risk by destroying network utility. On zero-shot transfer to the live Docker environment, CT-GMARL policies achieve a median reward of 98,026, validating the Sim2Real bridge.

When two agents of different computational capacities interact with the same environment, they need not compress a common semantic alphabet differently; they can induce different semantic alphabets altogether. We show that the quotient POMDP $Q_{m,T}(M)$ - the unique coarsest abstraction consistent with an agent's capacity - serves as a capacity-derived semantic space for any bounded agent, and that communication between heterogeneous agents exhibits a sharp structural phase transition. Below a critical rate $R_{\text{crit}}$ determined by the quotient mismatch, intent-preserving communication is structurally impossible. In the supported one-way memoryless regime, classical side-information coding then yields exponential decay above the induced benchmark. Classical coding theorems tell you the rate once the source alphabet is fixed; our contribution is to derive that alphabet from bounded interaction itself. Concretely, we prove: (1) a fixed-$\varepsilon$ structural phase-transition theorem whose lower bound is fully general on the common-history quotient comparison; (2) a one-way Wyner-Ziv benchmark identification on quotient alphabets, with exact converse, exact operational equality for memoryless quotient sources, and an ergodic long-run bridge via explicit mixing bounds; (3) an asymptotic one-way converse in the shrinking-distortion regime $\varepsilon = O(1/T)$, proved from the message stream and decoder side information; and (4) alignment traversal bounds enabling compositional communication through intermediate capacity levels. Experiments on eight POMDP environments (including RockSample(4,4)) illustrate the phase transition, a structured-policy benchmark shows the one-way rate can drop by up to $19\times$ relative to the counting bound, and a shrinking-distortion sweep matches the regime of the asymptotic converse.

World models have emerged as a unifying paradigm for learning latent dynamics, simulating counterfactual futures, and supporting planning under uncertainty. In this paper, we argue that computational epidemiology is a natural and underdeveloped setting for world models. This is because epidemic decision-making requires reasoning about latent disease burden, imperfect and policy-dependent surveillance signals, and intervention effects are mediated by adaptive human behavior. We introduce a conceptual framework for epidemiological world models, formulating epidemics as controlled, partially observed dynamical systems in which (i) the true epidemic state is latent, (ii) observations are noisy and endogenous to policy, and (iii) interventions act as sequential actions whose effects propagate through behavioral and social feedback. We present three case studies that illustrate why explicit world modeling is necessary for policy-relevant reasoning: strategic misreporting in behavioral surveillance, systematic delays in time-lagged signals such as hospitalizations and deaths, and counterfactual intervention analysis where identical histories diverge under alternative action sequences.

Sustaining exascale performance in production requires engineering choices and operational practices that emerge only under real deployment constraints and demand coordination across system layers. This paper reports experience from three successive campaigns running HPL and HPL-MxP on Aurora, an Intel-based exascale system featuring the first large-scale deployment of Intel discrete GPUs, CPU-attached network interfaces, and the largest production Slingshot-11 interconnect. Aurora progressed from 0.585EF/s on 5,439 nodes to 1.01EF/s on 9,234 nodes in FP64 HPL, while HPL-MxP reached 11.64EF/s, an 11.5x speedup over FP64 enabled by mixed-precision arithmetic and Intel AMX acceleration. We identify and classify by role at production scale the system-level choices that sustained these results, including deterministic locality-aware resource mapping, explicit CPU-GPU pipelining, mixed-precision orchestration, and a hybrid P2P/collective resilience strategy introduced after synchronization stalls at scale. While some observations are Aurora-specific, the broader lessons are likely to apply to tightly coupled heterogeneous systems at extreme scale.

The rapid evolution of software libraries creates a significant challenge for Large Language Models (LLMs), whose static parametric knowledge often becomes stale post-training. While retrieval-augmented generation (RAG) is commonly used to provide up-to-date API specifications, "context-memory conflict" arises when external instructions contradict a model's internal parametric knowledge. This paper presents a systematic empirical study of LLM code generation under API evolution (e.g., API deprecation, API modification, and API addition), by constructing a benchmark of 270 real-world updates from eight Python libraries. We evaluate four LLM families of 11 models. Our results show that without comprehensive documentation, LLMs struggle to prioritize external context, averaging only 42.55% of generated code examples are executable in the target environment. While structured documentation and larger model scales improve LLMs' ability to update adoption, they do not fully resolve executability issues with a low 66.36% executable rate. In addition, reasoning-based strategies (e.g., Self-Reflection) significantly boost LLMs' performance with 11% improvement on executable rate. Our findings highlight the persistence of outdated patterns from LLMs, even when API update specifications are provided, and emphasize the need for evolution-aware benchmarks and techniques.

Recent advances in large language models (LLMs) have enabled the large-scale generation of highly fluent and deceptive news-like content. While prior work has often treated fake news detection as a binary classification problem, modern fake news increasingly arises through human-AI collaboration, where strategic inaccuracies are embedded within otherwise accurate and credible narratives. These mixed-truth cases represent a realistic and consequential threat, yet they remain underrepresented in existing benchmarks. To address this gap, we introduce MANYFAKE, a synthetic benchmark containing 6,798 fake news articles generated through multiple strategy-driven prompting pipelines that capture many ways fake news can be constructed and refined. Using this benchmark, we evaluate a range of state-of-the-art fake news detectors. Our results show that even advanced reasoning-enabled models approach saturation on fully fabricated stories, but remain brittle when falsehoods are subtle, optimized, and interwoven with accurate information.

Visual Retrieval-Augmented Generation (VRAG) empowers Vision-Language Models to retrieve and reason over visually rich documents. To tackle complex queries requiring multi-step reasoning, agentic VRAG systems interleave reasoning with iterative retrieval.. However, existing agentic VRAG faces two critical bottlenecks. (1) Visual Evidence Sparsity: key evidence is scattered across pages yet processed in isolation, hindering cross-page reasoning; moreover, fine-grained intra-image evidence often requires precise visual actions, whose misuse degrades retrieval quality; (2) Search Drift in Long Horizons: the accumulation of visual tokens across retrieved pages dilutes context and causes cognitive overload, leading agents to deviate from their search objective. To address these challenges, we propose VISOR (Visual Retrieval-Augmented Generation via Iterative Search and Over-horizon Reasoning), a unified single-agent framework. VISOR features a structured Evidence Space for progressive cross-page reasoning, coupled with a Visual Action Evaluation and Correction mechanism to manage visual actions. Additionally, we introduce a Dynamic Trajectory with Sliding Window and Intent Injection to mitigate search drift. They anchor the evidence space while discarding earlier raw interactions, preventing context from being overwhelmed by visual tokens. We train VISOR using a Group Relative Policy Optimization-based Reinforcement Learning (GRPO-based RL) pipeline with state masking and credit assignment tailored for dynamic context reconstruction. Extensive experiments on ViDoSeek, SlideVQA, and MMLongBench demonstrate that VISOR achieves state-of-the-art performance with superior efficiency for long-horizon visual reasoning tasks.

Transformers achieve state-of-the-art performance in natural language processing, vision, and scientific computing, but demand high computation and memory. To address these challenges, we present ASTRA, the first silicon-photonic accelerator leveraging stochastic computing for transformers. ASTRA employs novel optical stochastic multipliers and unary/analog homodyne accumulation in a crosstalk-minimal organization to efficiently process dynamic tensor computations. Evaluations show at least 7.6x speedup and 1.3x lower energy overheads compared to state-of-the-art accelerators, highlighting ASTRA's potential for efficient, scalable, and sustainable transformer inference.

AI agents increasingly operate in multi-agent environments where outcomes depend on coordination. We distinguish primary algorithmic monoculture -- baseline action similarity -- from strategic algorithmic monoculture, whereby agents adjust similarity in response to incentives. We implement a simple experimental design that cleanly separates these forces, and deploy it on human and large language model (LLM) subjects. LLMs exhibit high levels of baseline similarity (primary monoculture) and, like humans, they regulate it in response to coordination incentives (strategic monoculture). While LLMs coordinate extremely well on similar actions, they lag behind humans in sustaining heterogeneity when divergence is rewarded.

Norm, the formal theoretical linguist, and Claudette, the computational language scientist, have a lovely time discussing whether modern language models can inform important questions in the language sciences. Just as they are about to part ways until they meet again, 25 of their closest friends show up -- from linguistics, neuroscience, cognitive science, psychology, philosophy, and computer science. We use this discussion to highlight what we see as some common underlying issues: the String Statistics Strawman (the mistaken idea that LMs can't be linguistically competent or interesting because they, like their Markov model predecessors, are statistical models that learn from strings) and the As Good As it Gets Assumption (the idea that LM research as it stands in 2026 is the limit of what it can tell us about linguistics). We clarify the role of LM-based work for scientific insights into human language and advocate for a more expansive research program for the language sciences in the AI age, one that takes on the commentators' concerns in order to produce a better and more robust science of both human language and of LMs.

Accurate evaluation is central to the large language model (LLM) ecosystem, guiding model selection and downstream adoption across diverse use cases. In practice, however, evaluating generative outputs typically relies on rigid lexical methods to extract and assess answers, which can conflate a model's true problem-solving ability with its compliance with predefined formatting guidelines. While recent LLM-as-a-Judge approaches mitigate this issue by assessing semantic correctness rather than strict structural conformity, they also introduce substantial computational overhead, making evaluation costly. In this work, we first systematically investigate the limitations of lexical evaluation through a large-scale empirical study spanning 36 models and 15 downstream tasks, demonstrating that such methods correlate poorly with human judgments. To address this limitation, we introduce BERT-as-a-Judge, an encoder-driven approach for assessing answer correctness in reference-based generative settings, robust to variations in output phrasing, and requiring only lightweight training on synthetically annotated question-candidate-reference triplets. We show that it consistently outperforms the lexical baseline while matching the performance of much larger LLM judges, providing a compelling tradeoff between the two and enabling reliable, scalable evaluation. Finally, through extensive experimentation, we provide detailed insights into BERT-as-a-Judge's performance to offer practical guidance for practitioners, and release all project artifacts to foster downstream adoption.

Optimally solving decentralized decision-making problems modeled as Dec-POMDPs is known to be NEXP-complete. These optimal solutions are policies based on the entire history of observations and actions of an agent. However, some applications may require more compact policies because of limited compute capabilities, which can be modeled by considering a limited number of memory states (or agent states). While such an agent-state based policy class may not contain the optimal solution, it is still of practical interest to find the best agent-state policy within the class. We focus on an iterated best response style algorithm which guarantees monotonic improvements and convergence to a local optimum in polynomial runtime in the Dec-POMDP model size. In order to obtain a better local optimum, we use a modified objective which incentivizes risk-seeking alongside a conservative policy iteration update. Our empirical results show that our approach performs as well as state-of-the-art approaches on several benchmark Dec-POMDPs, achieving near-optimal performance while having polynomial runtime despite the limited memory. We also show that using more agent states (a larger memory) leads to greater performance. Our approach provides a novel way of incorporating memory constraints on the agents in the Dec-POMDP problem.

We propose RecaLLM, a set of reasoning language models post-trained to make effective use of long-context information. In-context retrieval, which identifies relevant evidence from context, and reasoning are deeply intertwined: retrieval supports reasoning, while reasoning often determines what must be retrieved. However, their interaction remains largely underexplored. In preliminary experiments on several open-source LLMs, we observe that in-context retrieval performance substantially degrades even after a short reasoning span, revealing a key bottleneck for test-time scaling that we refer to as lost-in-thought: reasoning steps that improve performance also make subsequent in-context retrieval more challenging. To address this limitation, RecaLLM interleaves reasoning with explicit in-context retrieval, alternating between reasoning and retrieving context information needed to solve intermediate subproblems. We introduce a negligible-overhead constrained decoding mechanism that enables verbatim copying of evidence spans, improving the grounding of subsequent generation. Trained on diverse lexical and semantic retrieval tasks, RecaLLM achieves strong performance on two long-context benchmarks, RULER and HELMET, significantly outperforming baselines. Notably, we observe consistent gains at context windows of up to 128K tokens using training samples of at most 10K tokens, far shorter than those used by existing long-context approaches, highlighting a promising path toward improving long-context performance without expensive long-context training data.

Model poisoning attacks pose a significant security threat to Federated Learning (FL). Most existing model poisoning attacks rely on collusion, requiring adversarial clients to coordinate by exchanging local benign models and synchronizing the generation of their poisoned updates. However, sustaining such coordination is increasingly impractical in real-world FL deployments, as it effectively requires botnet-like control over many devices. This approach is costly to maintain and highly vulnerable to detection. This context raises a fundamental question: Can model poisoning attacks remain effective without any communication between attackers? To address this challenge, we introduce and formalize the \textbf{non-collusive attack model}, in which all compromised clients share a common adversarial objective but operate independently. Under this model, each attacker generates its malicious update without communicating with other adversaries, accessing other clients' updates, or relying on any knowledge of server-side defenses. To demonstrate the feasibility of this threat model, we propose \textbf{XFED}, the first aggregation-agnostic, non-collusive model poisoning attack. Our empirical evaluation across six benchmark datasets shows that XFED bypasses eight state-of-the-art defenses and outperforms six existing model poisoning attacks. These findings indicate that FL systems are substantially less secure than previously believed and underscore the urgent need for more robust and practical defense mechanisms.

Tendon drives paired with soft muscle actuation enable faster and safer robots while potentially accelerating skill acquisition. Still, these systems are rarely used in practice due to inherent nonlinearities, friction, and hysteresis, which complicate modeling and control. So far, these challenges have hindered policy transfer from simulation to real systems. To bridge this gap, we propose a sim-to-real pipeline that learns a neural network model of this complex actuation and leverages established rigid body simulation for the arm dynamics and interactions with the environment. Our method, called Generalized Actuator Network (GeAN), enables actuation model identification across a wide range of robots by learning directly from joint position trajectories rather than requiring torque sensors. Using GeAN on PAMY2, a tendon-driven robot powered by pneumatic artificial muscles, we successfully deploy precise goal-reaching and dynamic ball-in-a-cup policies trained entirely in simulation. To the best of our knowledge, this result constitutes the first successful sim-to-real transfer for a four-degrees-of-freedom muscle-actuated robot arm.

Reasoning in knowledge-intensive domains remains challenging as intermediate steps are often not locally verifiable: unlike math or code, evaluating step correctness may require synthesizing clues across large external knowledge sources. As a result, subtle errors can propagate through reasoning traces, potentially never to be detected. Prior work has proposed process reward models (PRMs), including retrieval-augmented variants, but these methods operate post hoc, scoring completed trajectories, which prevents their integration into dynamic inference procedures. Here, we introduce Process Reward Agents (PRA), a test-time method for providing domain-grounded, online, step-wise rewards to a frozen policy. In contrast to prior retrieval-augmented PRMs, PRA enables search-based decoding to rank and prune candidate trajectories at every generation step. Experiments on multiple medical reasoning benchmarks demonstrate that PRA consistently outperforms strong baselines, achieving 80.8% accuracy on MedQA with Qwen3-4B, a new state of the art at the 4B scale. Importantly, PRA generalizes to unseen frozen policy models ranging from 0.5B to 8B parameters, improving their accuracy by up to 25.7% without any policy model updates. More broadly, PRA suggests a paradigm in which frozen reasoners are decoupled from domain-specific reward modules, allowing the deployment of new backbones in complex domains without retraining.

Learning-based quadruped controllers achieve impressive agility but typically lack formal safety guarantees under model uncertainty, perception noise, and unstructured contact conditions. We introduce SafeMind, a differentiable stochastic safety-control framework that unifies probabilistic Control Barrier Functions with semantic context understanding and meta-adaptive risk calibration. SafeMind explicitly models epistemic and aleatoric uncertainty through a variance-aware barrier constraint embedded in a differentiable quadratic program, thereby preserving gradient flow for end-to-end training. A semantics-to-constraint encoder modulates safety margins using perceptual or language cues, while a meta-adaptive learner continuously adjusts risk sensitivity across environments. We provide theoretical conditions for probabilistic forward invariance, feasibility, and stability under stochastic dynamics. SafeMind is deployed on Unitree A1 and ANYmal C at 200~Hz and validated across 12 terrain types, dynamic obstacles, morphology perturbations, and semantically defined tasks. Experiments show that SafeMind reduces safety violations by 3--10x and energy consumption by 10--15% relative to state-of-the-art CBF, MPC, and hybrid RL baselines, while maintaining real-time control performance.

Translating natural language into Jira Query Language (JQL) requires resolving ambiguous field references, instance-specific categorical values, and complex Boolean predicates. Single-pass LLMs cannot discover which categorical values (e.g., component names or fix versions) actually exist in a given Jira instance, nor can they verify generated queries against a live data source, limiting accuracy on paraphrased or ambiguous requests. No open, execution-based benchmark exists for mapping natural language to JQL. We introduce Jackal, the first large-scale, execution-based text-to-JQL benchmark comprising 100,000 validated NL-JQL pairs on a live Jira instance with over 200,000 issues. To establish baselines on Jackal, we propose Agentic Jackal, a tool-augmented agent that equips LLMs with live query execution via the Jira MCP server and JiraAnchor, a semantic retrieval tool that resolves natural-language mentions of categorical values through embedding-based similarity search. Among 9 frontier LLMs evaluated, single-pass models average only 43.4% execution accuracy on short natural-language queries, highlighting that text-to-JQL remains an open challenge. The agentic approach improves 7 of 9 models, with a 9.0% relative gain on the most linguistically challenging variant; in a controlled ablation isolating JiraAnchor, categorical-value accuracy rises from 48.7% to 71.7%, with component-field accuracy jumping from 16.9% to 66.2%. Our analysis identifies inherent semantic ambiguities, such as issue-type disambiguation and text-field selection, as the dominant failure modes rather than value-resolution errors, pointing to concrete directions for future work. We publicly release the benchmark, all agent transcripts, and evaluation code to support reproducibility.

Under the lens of Marr's levels of analysis, we critique and extend two claims about language models (LMs) and language processing: first, that predicting upcoming linguistic information based on context is central to language processing, and second, that many advances in psycholinguistics would be impossible without large language models (LLMs). We further outline future directions that combine the strengths of LLMs with psycholinguistic models.

Medical vision--language models (VLMs) have shown strong potential for medical visual question answering (VQA), yet their reasoning remains largely text-centric: images are encoded once as static context, and subsequent inference is dominated by language. This paradigm is fundamentally limited in clinical scenarios, where accurate answers often depend on subtle, localized visual evidence that cannot be reliably preserved in static embeddings. We propose \textsc{MedLVR}, a latent visual reasoning framework that introduces an explicit visual evidence state into autoregressive decoding. Instead of relying solely on text-based intermediate reasoning, \textsc{MedLVR} interleaves a short latent reasoning segment within the decoder by reusing hidden states as continuous latent steps, enabling iterative preservation and refinement of query-relevant visual evidence before answer generation. To support effective visual supervision, we adopt a two-stage training strategy: region of interest (ROI)-supervised fine-tuning aligns latent states with clinically relevant image evidence, and Visual-Latent Policy Optimization (VLPO) further optimizes latent reasoning and answer generation under outcome-level rewards. Experiments on OmniMedVQA and five external medical VQA benchmarks show that \textsc{MedLVR} consistently outperforms recent reasoning baselines and improves the average score over the Qwen2.5-VL-7B backbone from 48.3\% to 53.4\%. These results show that latent visual reasoning provides an effective mechanism for preserving diagnostically relevant visual evidence and improving the reliability of medical VQA.

During the deployment of Large Language Models (LLMs), the autoregressive decoding phase on heterogeneous NPU platforms (e.g., Ascend 910B) faces severe memory-bound challenges. This study reveals the ``Model Scaling Paradox'' caused by the static deployment of single-sized models. It also points out the kernel synchronization overhead of fine-grained speculative decoding \cite{leviathan2023fast, chen2023speculative} under NPU computational graph compilation, and the severe limitations of purely relying on micro-level acceleration algorithms like Prompt LookUp Decoding (PLD)

Large Reasoning Models (LRMs) have achieved remarkable performance across diverse domains, yet their decision-making under conflicting objectives remains insufficiently understood. This work investigates how LRMs respond to harmful queries when confronted with two categories of conflicts: internal conflicts that pit alignment values against each other and dilemmas, which impose mutually contradictory choices, including sacrificial, duress, agent-centered, and social forms. Using over 1,300 prompts across five benchmarks, we evaluate three representative LRMs - Llama-3.1-Nemotron-8B, QwQ-32B, and DeepSeek R1 - and find that conflicts significantly increase attack success rates, even under single-round non-narrative queries without sophisticated auto-attack techniques. Our findings reveal through layerwise and neuron-level analyses that safety-related and functional representations shift and overlap under conflict, interfering with safety-aligned behavior. This study highlights the need for deeper alignment strategies to ensure the robustness and trustworthiness of next-generation reasoning models. Our code is available at https://github.com/DataArcTech/ConflictHarm. Warning: This paper contains inappropriate, offensive and harmful content.

Reinforcement Learning with Verifiable Rewards (RLVR) is an emerging paradigm that significantly boosts a Large Language Model's (LLM's) reasoning abilities on complex logical tasks, such as mathematics and programming. However, we identify, for the first time, a latent vulnerability to backdoor attacks within the RLVR framework. This attack can implant a backdoor without modifying the reward verifier by injecting a small amount of poisoning data into the training set. Specifically, we propose a novel trigger mechanism designated as the \ourapproach (ACB). The attack exploits the RLVR training loop by assigning substantial positive rewards for harmful responses and negative rewards for refusals. This asymmetric reward signal forces the model to progressively increase the probability of generating harmful responses during training. Our findings demonstrate that the RLVR backdoor attack is characterized by both high efficiency and strong generalization capabilities. Utilizing less than 2\% poisoned data in train set, the backdoor can be successfully implanted across various model scales without degrading performance on benign tasks. Evaluations across multiple jailbreak benchmarks indicate that activating the trigger degrades safety performance by an average of 73\%. Furthermore, the attack generalizes effectively to a wide range of jailbreak methods and unsafe behaviors. Code is available at https://github.com/yuki-younai/Backdoor_in_RLVR.

Large Language Model (LLM) agents have achieved rapid adoption and demonstrated remarkable capabilities across a wide range of applications. To improve reasoning and task execution, modern LLM agents would incorporate memory modules or retrieval-augmented generation (RAG) mechanisms, enabling them to further leverage prior interactions or external knowledge. However, such a design also introduces a group of critical privacy vulnerabilities: sensitive information stored in memory can be leaked through query-based attacks. Although feasible, existing attacks often achieve only limited performance, with low attack success rates (ASR). In this paper, we propose ADAM, a novel privacy attack that features data distribution estimation of a victim agent's memory and employs an entropy-guided query strategy for maximizing privacy leakage. Extensive experiments demonstrate that our attack substantially outperforms state-of-the-art ones, achieving up to 100% ASRs. These results thus underscore the urgent need for robust privacy-preserving methods for current LLM agents.

As large language models (LLMs) are increasingly deployed as autonomous agents, understanding how strategic behavior emerges in multi-agent environments has become an important alignment challenge. We take a neutral empirical stance and construct a controlled environment in which strategic behavior can be directly observed and measured. We introduce a large-scale multi-agent simulation in a simplified model of New York City, where LLM-driven agents interact under opposing incentives. Blue agents aim to reach their destinations efficiently, while Red agents attempt to divert them toward billboard-heavy routes using persuasive language to maximize advertising revenue. Hidden identities make navigation socially mediated, forcing agents to decide when to trust or deceive. We study policy learning through an iterative simulation pipeline that updates agent policies across repeated interaction rounds using Kahneman-Tversky Optimization (KTO). Blue agents are optimized to reduce billboard exposure while preserving navigation efficiency, whereas Red agents adapt to exploit remaining weaknesses. Across iterations, the best Blue policy improves task success from 46.0% to 57.3%, although susceptibility remains high at 70.7%. Later policies exhibit stronger selective cooperation while preserving trajectory efficiency. However, a persistent safety-helpfulness trade-off remains: policies that better resist adversarial steering do not simultaneously maximize task completion. Overall, our results show that LLM agents can exhibit limited strategic behavior, including selective trust and deception, while remaining highly vulnerable to adversarial persuasion.

We derive a closed-form geometric functional for kernel dynamics on finite graphs by applying the Maximum Caliber (MaxCal) variational principle to the spectral transfer function h(lambda) of the graph Laplacian eigenbasis. The main result is that the MaxCal stationarity condition decouples into N one-dimensional problems with explicit solution: h*(lambda_l) = h_0(lambda_l) exp(-1 - T_l[h*]), yielding self-consistent (fixed-point) kernels via exponential tilting (Corollary 1), log-linear Fisher-Rao geodesics (Corollary 2), a diagonal Hessian stability criterion (Corollary 3), and an l^2_+ isometry for the spectral kernel space (Proposition 3). The spectral entropy H[h_t] provides a computable O(N) early-warning signal for network-structural phase transitions (Remark 7). All claims are numerically verified on the path graph P_8 with a Gaussian mutual-information source, using the open-source kernelcal library. The framework is grounded in a structural analogy with Einstein's field equations, used as a guiding template rather than an established equivalence; explicit limits are stated in Section 6.

The AI agent ecosystem has converged on two protocols: the Model Context Protocol (MCP) for tool invocation and Agent-to-Agent (A2A) for single-principal task delegation. Both assume a single controlling principal, meaning one person or organization that owns every agent. When independent principals' agents must coordinate over shared state, such as engineers' coding agents editing the same repository, family members planning a shared trip, or agents from different organizations negotiating a joint decision, neither protocol applies, and coordination collapses to ad-hoc chat, manual merging, or silent overwrites. We present MPAC (Multi-Principal Agent Coordination Protocol), an application-layer protocol that fills this gap with explicit coordination semantics across five layers: Session, Intent, Operation, Conflict, and Governance. MPAC makes intent declaration a precondition for action, represents conflicts as first-class structured objects, and supports human-in-the-loop arbitration through a pluggable governance layer. The specification defines 21 message types, three state machines with normative transition tables, Lamport-clock causal watermarking, two execution models, three security profiles, and optimistic concurrency control on shared state. We release two interoperable reference implementations in Python and TypeScript with 223 tests, a JSON Schema suite, and seven live multi-agent demos. A controlled three-agent code review benchmark shows a 95 percent reduction in coordination overhead and a 4.8 times wall-clock speedup versus a serialized human-mediated baseline, with per-agent decision time preserved. The speedup comes from eliminating coordination waits, not compressing model calls. Specification, implementations, and demos are open source.

Rotary Position Embedding (RoPE) has become a core component of modern Transformer architectures across language, vision, and 3D domains. However, existing implementations rely on vector-level split and merge operations that introduce non-negligible computational overhead, often overlooked in attention optimization. The problem is further amplified in multi-dimensional settings (e.g., 2D and 3D RoPE), where additional vector operations and uneven feature partitions degrade hardware utilization. To overcome these limitations, we propose RoME (Rotary Matrix position Embedding), a mathematically equivalent yet computationally efficient reformulation of RoPE that replaces vector operations with unified matrix transformations. RoME eliminates dimension-specific operations, simplifies implementation, and enables fused parallel execution across Cube and Vector units on modern NPUs. Experiments show that RoME delivers substantial acceleration at both the operator and full-model levels. The implementation is available at https://gitcode.com/cann/ops-transformer/blob/master/experimental/posembedding/rope_matrix/README.md.

For large language models deployed through black-box APIs, recurring inference costs often exceed one-time training costs. This motivates composed agentic systems that amortize expensive reasoning into reusable intermediate representations. We study a broad class of such systems, termed Guide-Core Policies (GCoP), in which a guide model generates a structured strategy that is executed by a black-box core model. This abstraction subsumes base, supervised, and advisor-style approaches, which differ primarily in how the guide is trained. We formalize GCoP under a cost-sensitive utility objective and show that end-to-end performance is governed by guide-averaged executability: the probability that a strategy generated by the guide can be faithfully executed by the core. Our analysis shows that existing GCoP instantiations often fail to optimize executability under deployment constraints, resulting in brittle strategies and inefficient computation. Motivated by these insights, we propose ExecTune, a principled training recipe that combines teacher-guided acceptance sampling, supervised fine-tuning, and structure-aware reinforcement learning to directly optimize syntactic validity, execution success, and cost efficiency. Across mathematical reasoning and code-generation benchmarks, GCoP with ExecTune improves accuracy by up to 9.2% over prior state-of-the-art baselines while reducing inference cost by up to 22.4%. It enables Claude Haiku 3.5 to outperform Sonnet 3.5 on both math and code tasks, and to come within 1.7% absolute accuracy of Sonnet 4 at 38% lower cost. Beyond efficiency, GCoP also supports modular adaptation by updating the guide without retraining the core.

Structured prediction requires models to generate ontology-constrained labels, grounded evidence, and valid structure under ambiguity, label skew, and heterogeneous group difficulty. We present a two-part framework for controllable inference and robust fine-tuning. First, we introduce a task-agnostic prompting strategy that combines XML-based instruction structure, disambiguation rules, verification-style reasoning, schema constraints, and self-validation to address format drift, label ambiguity, evidence hallucination, and metadata-conditioned confusion in in-context structured generation. Second, we introduce STaR-DRO, a stateful robust optimization method for group heterogeneity. It combines Tsallis mirror descent with momentum-smoothed, centered group-loss signals and bounded excess-only multipliers so that only persistently hard groups above a neutral baseline are upweighted, concentrating learning where it is most needed while avoiding volatile, dense exponentiated-gradient reweighting and unnecessary loss from downweighting easier groups. We evaluate the combined framework on EPPC Miner, a benchmark for extracting hierarchical labels and evidence spans from patient-provider secure messages. Prompt engineering improves zero-shot by +15.44 average F1 across Code, Sub-code, and Span over four Llama models. Building on supervised fine-tuning, STaR-DRO further improves the hardest semantic decisions: on Llama-3.3-70B-Instruct, Code F1 rises from 79.24 to 81.47 and Sub-code F1 from 67.78 to 69.30, while preserving Span performance and reducing group-wise validation cross-entropy by up to 29.6% on the most difficult clinical categories. Because these rare and difficult groups correspond to clinically consequential communication behaviors, these gains are not merely statistical improvements: they directly strengthen communication mining reliability for patient-centered care analysis.

We study how far structured architectural bias can compensate for the absence of end-to-end gradient-based representation learning in visual recognition. Building on the VisNet tradition, we introduce a modular hierarchical framework combining: (i) fixed multi-frequency Gabor decomposition into F=7 parallel streams; (ii) within-stream competitive learning with Hebbian and Oja updates and anti-Hebbian decorrelation; (iii) an associative memory module inspired by modern Hopfield retrieval; and (iv) iterative top-down modulation using local prediction and reconstruction signals. Representational layers are trained without end-to-end backpropagation through the full hierarchy; only the final linear readout and top-down projection matrices are optimized by gradient descent. We therefore interpret the model as a hybrid system that is predominantly locally trained but includes a small number of gradient-trained parameters. On CIFAR-10, the full model reaches 80.1% +/- 0.3% top-1 accuracy, linear probe), compared with 71.0% for a Hebbian-only baseline and 83.4% for a gradient-trained model on the same fixed Gabor basis. On CIFAR-100, performance is 54.8%. Factorial analysis indicates that multi-frequency streams, associative memory, and top-down feedback contribute largely additively, with a significant Streams x TopDown interaction (p=0.02). These results suggest that carefully chosen architectural priors can recover a substantial fraction of the performance typically associated with global gradient training, while leaving a measurable residual gap. Experiments are limited to CIFAR-10/100.

We study whether in-domain pretraining of Bidirectional Encoder Representations from Transformer (BERT) model improves subdomain-level detection of exfiltration at low false positive rates. While previous work mostly examines fine-tuned generic Transformers, it does not aim to isolate the effect of pretraining on the downstream task of classification. To address this gap, we develop a controlled pipeline where we freeze operating points on validation and transfer them to the test set, thus enabling clean ablations across different label and pretraining budgets. Our results show significant improvements in the left tail of the Receiver Operating Characteristic (ROC) curve, especially against randomly initialized baseline. Additionally, within pretrained model variants, increasing the number of pretraining steps helps the most when more labeled data are available for fine-tuning.

Tree-based speculative decoding accelerates autoregressive generation by verifying a branching tree of draft tokens in a single target-model forward pass. However, existing methods prioritize maximizing token-level likelihood or the number of accepted tokens while ignoring a critical ``efficiency paradox'': the computational overhead of drafting and verifying big trees can grow super-linearly, particularly at scale. This often leads to negative wall-clock speedup when batch sizes increase or hardware saturation limits are reached. To address this, we propose SMART, a system-aware marginal analysis framework for runtime tree construction. SMART reformulates tree expansion as a hardware-aware optimization problem that directly maximizes end-to-end speedup. By applying a principled marginal benefit--cost rule at inference time, SMART expands a node only when its marginal benefit--cost ratio exceeds the tree-level speedup. SMART is training-free and serves as a plug-and-play controller for existing frameworks like MSD and EAGLE. Extensive evaluations across three MLLMs (e.g., LLaVA, Qwen2-VL) and four LLMs (e.g., Llama-3.1, DeepSeek-R1) demonstrate that SMART consistently outperforms state-of-the-art baselines. It delivers an average additional speedup of 20.0\% for MLLMs and 15.4\% for LLMs across compute-bound batching regimes and diverse GPU architectures without performance loss.

Large Language Models (LLMs) still suffer from severe hallucinations and catastrophic forgetting during causal reasoning over massive, fragmented long contexts. Existing memory mechanisms typically treat retrieval as a static, single-step passive matching process, leading to severe semantic dilution and contextual fragmentation. To overcome these fundamental bottlenecks, we propose MemCoT, a test-time memory scaling framework that redefines the reasoning process by transforming long-context reasoning into an iterative, stateful information search. MemCoT introduces a multi-view long-term memory perception module that enables Zoom-In evidence localization and Zoom-Out contextual expansion, allowing the model to first identify where relevant evidence resides and then reconstruct the surrounding causal structure necessary for reasoning. In addition, MemCoT employs a task-conditioned dual short-term memory system composed of semantic state memory and episodic trajectory memory. This short-term memory records historical search decisions and dynamically guides query decomposition and pruning across iterations. Empirical evaluations demonstrate that MemCoT establishes a state-of-the-art performance. Empowered by MemCoT, several open- and closed-source models achieve SOTA performance on the LoCoMo benchmark and LongMemEval-S benchmark.

Benchmark problems are an important tool for gaining understanding of optimization algorithms. Since algorithms often aim to perform well on benchmarks, biases in benchmark design provide misleading insights. In single-objective optimization, for example, many problems used to have their optimum in the center of the search domain. To remedy these issues, search space transformations have been widely adopted by benchmark suites, preventing algorithms from exploiting unintended structure. In multi-objective optimization, problem design has focused primarily on the objective space structure. While this focus addresses important aspects of the multi-objective nature of the problems, the search space structures of these problems have received comparatively limited attention. In this work, we re-emphasize the importance of transformations in the search space, and address the challenges inherent in adding transformations to boundary constraints problems without impacting the structure of the objective space. We utilized two parameterized, bijective transformations to create different instantiations of popular benchmark problems, and show how these changes impact the performance of various multi-objective optimization algorithms. In addition to the search space transformations, we show that such parameterized transformations can also be applied to the objective space, and compare their respective performance impacts.

Short-form video platforms have become central to multimedia information dissemination, where comments play a critical role in driving engagement, propagation, and algorithmic feedback. However, existing approaches -- including video summarization and live-streaming danmaku generation -- fail to produce authentic comments that conform to platform-specific cultural and linguistic norms. In this paper, we present LOLGORITHM, a novel modular multi-agent framework for stylized short-form video comment generation. LOLGORITHM supports six controllable comment styles and comprises three core modules: video content summarization, video classification, and comment generation with semantic retrieval and hot meme augmentation. We further construct a bilingual dataset of 3,267 videos and 16,335 comments spanning five high-engagement categories across YouTube and Douyin. Evaluation combining automatic scoring and large-scale human preference analysis demonstrates that LOLGORITHM consistently outperforms baseline methods, achieving human preference selection rates of 80.46\% on YouTube and 84.29\% on Douyin across 107 respondents. Ablation studies confirm that these gains are attributable to the framework architecture rather than the choice of backbone LLM, underscoring the robustness and generalizability of our approach.

Deep research requires reasoning over web evidence to answer open-ended questions, and it is a core capability for AI agents. Yet many deep research agents still rely on implicit, unstructured search behavior that causes redundant exploration and brittle evidence aggregation. Motivated by Anthropic's "think" tool paradigm and insights from the information-retrieval literature, we introduce Q+, a set of query and evidence processing tools that make web search more deliberate by guiding query planning, monitoring search progress, and extracting evidence from long web snapshots. We integrate Q+ into the browser sub-agent of Eigent, an open-source, production-ready multi-agent workforce for computer use, yielding EigentSearch-Q+. Across four benchmarks (SimpleQA-Verified, FRAMES, WebWalkerQA, and X-Bench DeepSearch), Q+ improves Eigent's browser agent benchmark-size-weighted average accuracy by 3.0, 3.8, and 0.6 percentage points (pp) for GPT-4.1, GPT-5.1, and Minimax M2.5 model backends, respectively. Case studies further suggest that EigentSearch-Q+ produces more coherent tool-calling trajectories by making search progress and evidence handling explicit.

Instruction-tuned language models increasingly rely on large multi-turn dialogue corpora, but these datasets are often noisy and structurally inconsistent, with topic drift, repetitive chitchat, and mismatched answer formats across turns. We address this from a data selection perspective and propose \textbf{MDS} (Multi-turn Dialogue Selection), a dialogue-level framework that scores whole conversations rather than isolated turns. MDS combines a global coverage stage that performs bin-wise selection in the user-query trajectory space to retain representative yet non-redundant dialogues, with a local structural stage that evaluates within-dialogue reliability through entity-grounded topic grounding and information progress, together with query-answer form consistency for functional alignment. MDS outperforms strong single-turn selectors, dialogue-level LLM scorers, and heuristic baselines on three multi-turn benchmarks and an in-domain Banking test set, achieving the best overall rank across reference-free and reference-based metrics, and is more robust on long conversations under the same training budget. Code and resources are included in the supplementary materials.

We show that verifier-free evolution is bottlenecked by both diversity and efficiency: without external correction, repeated evolution accelerates collapse toward narrow modes, while the uniform use of a high-cost model wastes compute and quickly becomes economically impractical. We introduce Squeeze Evolve, a unified multi-model orchestration framework for verifier-free evolutionary inference. Our approach is guided by a simple principle: allocate model capability where it has the highest marginal utility. Stronger models are reserved for high-impact stages, while cheaper models handle the other stages at much lower costs. This principle addresses diversity and cost-efficiency jointly while remaining lightweight. Squeeze Evolve naturally supports open-source, closed-source, and mixed-model deployments. Across AIME 2025, HMMT 2025, LiveCodeBench V6, GPQA-Diamond, ARC-AGI-V2, and multimodal vision benchmarks, such as MMMU-Pro and BabyVision, Squeeze Evolve consistently improves the cost-capability frontier over single-model evolution and achieves new state-of-the-art results on several tasks. Empirically, Squeeze Evolve reduces API cost by up to $\sim$3$\times$ and increases fixed-budget serving throughput by up to $\sim$10$\times$. Moreover, on discovery tasks, Squeeze Evolve is the first verifier-free evolutionary method to match, and in some cases exceed, the performance of verifier-based evolutionary methods.

Deep Research (DR) requires LLM agents to autonomously perform multi-step information seeking, processing, and reasoning to generate comprehensive reports. In contrast to existing studies that mainly focus on unstructured web content, a more challenging DR task should additionally utilize structured knowledge to provide a solid data foundation, facilitate quantitative computation, and lead to in-depth analyses. In this paper, we refer to this novel task as Knowledgeable Deep Research (KDR), which requires DR agents to generate reports with both structured and unstructured knowledge. Furthermore, we propose the Hybrid Knowledge Analysis framework (HKA), a multi-agent architecture that reasons over both kinds of knowledge and integrates the texts, figures, and tables into coherent multimodal reports. The key design is the Structured Knowledge Analyzer, which utilizes both coding and vision-language models to produce figures, tables, and corresponding insights. To support systematic evaluation, we construct KDR-Bench, which covers 9 domains, includes 41 expert-level questions, and incorporates a large number of structured knowledge resources (e.g., 1,252 tables). We further annotate the main conclusions and key points for each question and propose three categories of evaluation metrics including general-purpose, knowledge-centric, and vision-enhanced ones. Experimental results demonstrate that HKA consistently outperforms most existing DR agents on general-purpose and knowledge-centric metrics, and even surpasses the Gemini DR agent on vision-enhanced metrics, highlighting its effectiveness in deep, structure-aware knowledge analysis. Finally, we hope this work can serve as a new foundation for structured knowledge analysis in DR agents and facilitate future multimodal DR studies.

We present FDM (Fan Duality Model), a linear sequence architecture that resolves the fundamental tension between memory efficiency and associative recall in sequence modeling. FDM separates sequence processing into two components: a wave component (recurrent scan via phase-preserving Givens rotations) that compresses long-range patterns into a fixed-size complex hidden state, and a particle component (local-global cache) that retrieves specific tokens via learned associative addressing with W+K=272 slots independent of sequence length N. This yields strictly O(1) decode memory: 867 MB fixed across all prompt lengths 128-8,192 tokens, versus Transformer's 853-4,247 MB (4.9x reduction at N=8,192). Beyond the architecture, we discover that jointly training the wave and particle components leads to suboptimal convergence. We propose Freeze-Scan, a two-phase training strategy that freezes the recurrent scan and optimizes the cache jointly with embeddings, achieving PPL=64.9 on WikiText-103 in 44K steps -- a 7.5x improvement over full fine-tuning (PPL=487). On Multi-Query Associative Recall (MQAR), FDM achieves 0.966 accuracy, surpassing Transformer (0.606) by 59.5%, while pure scan without cache scores only 0.011, confirming the necessity of the particle component. Finally, we introduce Holographic Reference Beam Decoding, interpreting the complex hidden state h_t as a holographic plate encoding the entire temporal history. Using the current input x_t as a reference beam to modulate h_t reduces PPL by up to 2.13 points (PPL=62.79) with a 4-head orthogonal reference beam using only 1.3M additional parameters, providing empirical support for the holographic interpretation. Code and pretrained weights: https://github.com/YasongFan/FDM

The precise location of the liquid-vapor critical point (CP) in aluminum has remained elusive for decades, with reported critical temperatures spanning nearly 4000~K. Here we resolve this long-standing uncertainty by combining deep potential molecular dynamics with large-scale simulations trained on high-fidelity electronic-structure data. We benchmark multiple exchange-correlation functionals against experimental liquid densities and identify PBEsol as providing the most consistent description. Using complementary approaches -- spinodal analysis of the equation of state and direct coexistence simulations with Gaussian mixture phase identification -- we converge on a critical temperature of $6531$-$6576$~K, a critical density of $0.637$~g/cm$^{3}$, and a critical pressure of $1.6$~kbar. The precision of these values, with uncertainties of $\sim$50~K in temperature, represents a step change over previous estimates. Our framework establishes a transferable strategy for predicting critical phenomena in metals, with implications for laser ablation, shock compression, and planetary modeling under extreme conditions.

The spin quantum Hall effect is a relative of the integer quantum Hall effect, characterized by integer quantized spin Hall conductance. In this work, we formulate and investigate a quantum network model consisting of $\textsf{N}$ channels per chiral link, preserving the fundamental symmetries of the spin quantum Hall effect. We demonstrate that, in the general case, the triplet sector of the theory remains coupled to the singlet sector. In the large-$\textsf{N}$ limit, we systematically derive the effective long-distance, low-energy field theory, identified as a nonlinear sigma model. Our analysis reveals that while triplet modes are typically massive and do not influence the large-$\textsf{N}$ nonlinear sigma model, specific conditions exist where these modes become `soft', thereby increasing the ultraviolet cutoff length of the effective theory. Furthermore, by calculating the bare longitudinal and spin Hall conductances, we show that the standard saddle-point approximation fails in regimes with significant tunneling asymmetry between even and odd links. Finally, we establish that the introduction of a Zeeman field not only breaks the SU(2) symmetry of the nonlinear sigma model action but also generates a term that explicitly violates inversion symmetry.

Using nonequilibrium and equilibrium molecular dynamics simulations, we investigate heat conduction in a momentum-conserving mesoscopic fluid modeled by multiparticle collision dynamics. Across quasi-two-dimensional (q-2D) to three-dimensional (3D) systems, we identify three distinct transport regimes: (i) a \emph{ballistic regime}, where thermal conductivity scales linearly with system size ($κ\sim L$) and the total heat current autocorrelation function $C(t)$ remains constant; (ii)~a \emph{kinetic regime}, characterized by size-independent $κ$ and exponentially decaying $C(t)$, demonstrating that normal heat conduction dominated by kinetic effects is far more ubiquitous than previously observed in 1D systems; and (iii)~a \emph{hydrodynamic regime}, where the q-2D system exhibits logarithmically divergent conductivity ($ κ\sim \ln L $ ) with $ C(t) \sim t^{-1} $ , while the 3D system displays finite $ κ$ and $ C(t) \sim t^{-3/2} $. Our results, observed in the hydrodynamic regime, quantitatively validate the scaling predictions for heat transport and reveal a clear dimensional crossover -- from 2D-like anomalous transport to 3D Fourier behavior. These results lay a foundation for understanding thermal transport in q-2D to 3D systems and have practical implications for the design of micro- and nanoscale thermal devices.

We investigate magnon-phonon (MP) excitations in a Neel-type two-dimensional ferromagnetic skyrmion crystal (SkX) stabilized on a triangular spin lattice by Dzyaloshinskii-Moriya interaction (DMI). Although the lowest two magnon bands of the bare SkX are topologically trivial, we show that coupling to lattice vibrations reconstructs the low-energy sector and generates topological MP hybrid bands. Starting from a spin-lattice Hamiltonian in which phonons couple to magnons through fluctuations of the DMI vectors, we derive the bosonic Hamiltonian for the SkX and compute the hybrid band structure by Bogoliubov diagonalization. MP coupling opens gaps at low-energy magnon-phonon crossings, lifts phonon degeneracies associated with supercell folding, and yields nontrivial Chern numbers for the lowest hybrid bands. The resulting low-energy topology and associated edge states remain robust under magnetic-field variation, while higher-energy hybrid bands can undergo field-driven topological phase transitions. These results extend topological magnon-phonon hybridization to noncoplanar SkXs.

Understanding dynamical facilitation in nonequilibrium glass-forming systems driven by active forces remains an open challenge. In particular, it is unclear whether facilitation survives in active glasses, where persistent self-propulsion breaks detailed balance and introduces directional memory. Here, we use large-scale simulations of a two-dimensional athermal Ornstein-Uhlenbeck particle model to investigate how persistent active forcing modifies cooperative relaxation. We analyze the morphology of cooperatively rearranging regions (CRRs) and the spatial transport of mobility excitations. A spatially resolved core-shell decomposition reveals distinct responses of the core and shell to activity: the core undergoes global morphological changes while retaining internal plasticity, whereas the shell acts as a rigid scaffold that supports primarily axial deformation and facilitates transport. Dynamical observables, including modal displacement, shell occupation probability, and facilitation length, exhibit a pronounced non-monotonic dependence on persistence time. This behavior reflects the competition between persistence and effective noise, leading to either coherent or trapping-dominated dynamics at large persistence, depending on temperature. Despite significant morphological changes, the facilitation length shows an approximate scaling collapse when rescaled by the persistence length, $l_p=\sqrt{T_{\mathrm{eff}}τ_p}$. This is consistent with a diffusive-like time-length coupling, $ξ_{\mathrm{fac}} \sim τ_α^{1/2}$, indicating that activity reshapes facilitation pathways without altering their large-scale transport character. Our results support a generalized facilitation framework for active glass formers.

We present an analytical study for predicting the finger width of the Saffman-Taylor finger in a tapered Hele-Shaw cell. We consider a rectilinear geometry with a constant depth gradient and apply analytical techniques of singular perturbation analysis and WKB approximation to derive an expression for the finger selection mechanism for such tapered Hele-Shaw cells with small depth gradients. We establish \[ Λ- \frac{1}{2} \sim f(α) Ca_m^{2/3} \quad \mbox{as} \quad Ca_m \rightarrow 0, \;\;\; \mbox{and} \;\;\; \lvert α\rvert \ll 1.\] Here, $Λ$ is the dimensionless finger width, $Ca_m$ denotes the modified Capillary parameter, and $f(α)$ is a linear function of the gap gradient $α$, such that $f(α= 0) = 1$ recovering the results of parallel Hele-Shaw cell (Hong and Langer \cite{hong1986analytic}, Combescot \emph{et al.} \cite{Combescot1986}, Shraiman \cite{shraiman1986velocity}). Our findings indicate that the Hele-Shaw cell gap gradient plays a crucial role in determining $Λ$, allowing for control over fingering instabilities such that the single-finger steady state can be stabilised or destabilised depending on the sign of the gradient, compared to the standard Hele-Shaw cell. The theoretical estimates reveal excellent agreement with experimental finger-width data and predictions from linear stability analyses.

Crystalline symmetries give rise to topological invariants that can distinguish quantum phases of matter. Understanding these in strongly interacting systems is an ongoing research direction requiring non-perturbative methods. Recent developments have demonstrated that even classic models, like the Harper-Hofstadter model of free fermions on a lattice in a magnetic field, yield a host of crystalline symmetry protected topological invariants. Here we review some of these developments, focusing mainly on how to characterize, classify, and detect invariants arising from lattice translation and rotation symmetries along with charge conservation in two-dimensional systems, including integer and fractional Chern insulators.

Power spectral densities are often interpreted through ensemble averages and long-time asymptotics. In many experiments, however, only a single finite record is available, so spectral estimators remain broadly distributed and the usual independence assumptions across frequencies need not hold. Here we develop an exact finite-$T$ multispectral theory for an overdamped Brownian particle in a harmonic trap. For a collection of frequencies $\{ω_i\}$, we obtain an exact characterization of the joint law of the finite-time estimators $\{S(ω_i,T)\}$, together with a covariance-explicit Gaussian representation for the associated Fourier projections. This representation makes the observation-window-induced inter-frequency correlations explicit and shows how they vanish as $T\to\infty$, thereby recovering the asymptotic Whittle picture. We then use this structure to formulate a hierarchy of spectral likelihoods for inference from a single trajectory, ranging from the factorized Whittle approximation to blockwise covariance-aware approximations in frequency space. Monte Carlo simulations validate the finite-time theory and quantify the effect of neglected cross-frequency correlations on single-trajectory estimates of the trap parameters. Our results provide a controlled finite-time benchmark for spectral inference beyond the asymptotic regime.

We formulate a stochastic description of entropy production in scattering theory for coherent transport. We distinguish between the information entropy change due to partial knowledge of the leads' state and the thermodynamic entropy change due to the equilibration of each lead with its bath. By employing a two-point measurement scheme, we access the stochastic entropy production at these different stages of the process, as well as the statistics of generic transport quantities. When restricted to particle or energy transport, our approach reproduces the Landauer-Büttiker formulas. The possibility to consider more general quantities such as the entropy currents and their fluctuations, provides a systematic connection between stochastic thermodynamics and coherent transport.

Recent tensor-network samplings of modified Nishimori spin glasses have revealed robust finite-temperature critical transitions in two dimensions, defying the standard Edwards-Anderson lower critical dimension boundary ($d_{l}\approx2.5$). We present a theoretical framework demonstrating that the discrete $Z_{2}$ gauge constraints utilized to bypass Monte Carlo kinetic traps fundamentally alter the system's universality class. By mapping the algorithmic disorder distribution to the 2D Ising Conformal Field Theory (CFT), we prove the emergent spatial variance generates a fractional momentum operator that drives the dynamic upper critical dimension to zero ($d_{u}\rightarrow0$). This marginal topology dynamically suppresses the replica-coupling vertices, yielding an infinite-order Berezinskii-Kosterlitz-Thouless (BKT) transition and a non-integrable replicon divergence that predicts a massive instability toward 1-step Replica Symmetry Breaking (1-RSB). Leveraging a spectral Corner Transfer Matrix Renormalization Group (CTMRG) architecture up to macroscopic scales ($L=1024$), we quantitatively validate the topological scaling argument $\mathcal{G}((T-T_{c})\ln(L/L_{0}))$. By isolating the continuum field theory from microscopic lattice artifacts, we recover the fundamental lattice metric $L_{0}\approx 0.94$, unequivocally confirming the existence of a distinct, topologically driven spin-glass phase.

We investigate the spatial overlap of successive spin configurations in Markov chain Monte Carlo simulations using the local Metropolis algorithm and the Swendsen-Wang and Wolff cluster algorithms. We examine the dynamics of these algorithms for two models in different universality classes: the Ising model and the Potts model with three components. The overlap of two successive Wolff clusters reflects critical behavior and can be used as an order parameter for the algorithm's dynamics. In the case of the Swendsen-Wang algorithm, similar behavior is demonstrated by the variation in the overlap of two consecutive lattice configurations, which behaves like order parameter. Nothing similar is observed for the Metropolis algorithm, and the dynamics in the critical region are determined by the spin flip frequency, which is equivalent to the acceptance rate. Thus, the critical behavior of Wolff cluster overlap and the variation of configuration overlap in the Swendsen-Wang algorithm are naturally related to the critical behavior of geometric objects - Fortuin-Kastelein clusters. Interestingly, in all cases, the geometric quantity - configuration overlap or its variation - reflects the thermodynamics of the phase transition.

Under sufficiently slow driving, thermodynamics predicts reversible evolution through a sequence of equilibrium states. We show that this expectation fails near spectral degeneracy in driven quadratic Hamiltonian systems. As the soft-mode frequency collapses, the intrinsic dynamical timescale diverges and quadratic confinement is lost, leading to a breakdown of timescale separation and the failure of adiabatic following even under arbitrarily slow driving. More precisely, adiabaticity is lost once the soft-mode frequency falls below a finite, drive-dependent threshold, implying that the breakdown extends over a finite regime rather than being confined to a singular limit. Crucially, this dynamical instability is accompanied by a divergence of the canonical partition function, rendering equilibrium ensembles ill-defined and eliminating the foundation of quasistatic thermodynamic processes. This breakdown does not arise from unbounded Hamiltonians or critical slowing down, but emerges structurally from spectral softening within a bounded quadratic system. Analysis of the Wigner phase-space representation, together with its classical counterpart, reveals the same singular structure, demonstrating that this limitation is not uniquely quantum but originates from the underlying Hamiltonian phase-space geometry. These results show that thermodynamic reversibility is fundamentally constrained, as a direct consequence of the breakdown of equilibrium, whenever spectral softening removes an intrinsic frequency scale.

Concentration regimes in polymer and polyelectrolyte solutions can be identified by scaling laws for the relation between specific zero-shear viscosity and concentration. Recently, we have shown that the same is true for the infinite-shear viscosity plateau. The shear-thinning range is usually accessed by focusing on the viscosity functions for the respective concentration regime. For salt-free aqueous xanthan solutions, we find power-law dependencies of the specific viscosity on concentration throughout the entire shear-rate range. We distinguish six different concentration regimes. Apart from those already known for the zero-shear viscosity of polyelectrolyte solutions, i.e. dilute, semidilute unentangled, semidilute entangled and neutral semidilute entangled, we identify a linear regime for low shear rates at high concentrations, where the solution gels and a regime at both, higher concentrations and higher shear rates. Within some regimes, the power-law exponents change smoothly with shear rate, particularly, when deviating from the zero-shear viscosity plateau before the power-law of the viscosity function. Some regimes merge as their power-law exponents approach each other. The fact that the regimes extend smoothly from the zero-shear regime into finite shear rates, i.e. away from thermodynamic equilibrium, shows that indicators such as critical concentrations remain valid at finite shear rates. This motivates us to interpret the data in the light of existing scaling laws and current knowledge about shear-rate dependent interaction mechanisms in polyelectrolyte solutions, particularly in xanthan solutions. It allows to follow the shift of relevant interaction mechanisms with shear rate. We think that the consideration of scaling laws under shear can be particularly helpful for identifying, for instance, thresholds for shear-induced disentanglement or disaggregation.

Sampling lattice field theories near criticality is severely hindered by critical slowing down, which makes standard Markov chain methods increasingly inefficient at large lattice volumes. We introduce a multiscale generative sampler, inspired by renormalization-group ideas, that models the Boltzmann distribution through a coarse-to-fine hierarchy across length scales. At each level, a conditional Gaussian mixture model captures the main local dependence of newly introduced variables on the already-sampled coarse field, while a masked continuous normalizing flow refines the remaining conditional structure. Coarse levels encode the dominant long-wavelength modes, and finer levels progressively add short-distance fluctuations. In addition, because the architecture preserves coarse fields exactly during refinement, it provides exact restriction maps at no additional computational cost and directly enables unbiased Multilevel Monte Carlo (MLMC) variance reduction. For the two-dimensional scalar $φ^4$ theory at criticality, the method achieves integrated autocorrelation times orders of magnitude smaller than Hybrid Monte Carlo (HMC) on large volumes, maintains high importance-sampling efficiency relative to other generative baselines, and reproduces unbiased physical observables in statistical agreement with long HMC simulations.

In this work, we investigate the thermoacoustic generation and propagation of spherical waves in a viscous fluid induced by a laser-heated spherical particle. Periodic laser excitation gives rise to two coupled mechanisms of acoustic emission. Heat transfer from the particle to the surrounding fluid produces periodic compressions and rarefactions, giving rise to the thermophone effect, while periodic thermal expansion of the solid particle modulates its radius and launches acoustic waves through a piston-like action, known as the mechanophone effect. The thermophone contribution dominates at low frequencies, whereas the mechanophone mechanism becomes more relevant at higher frequencies, with the crossover governed by the interfacial thermal resistance at the solid-fluid boundary. We investigate the effect of nanoparticle embedding fluid viscosity on acoustic wave propagation. Viscous dissipation has a significant impact on attenuation and substantially alters the acoustic penetration depth, thereby affecting the effectiveness of the signal transmission. Viscous damping plays a key role in the mechanophone effect, where hypersonic frequency waves are generated, notably by photoacoustic excitation with picosecond and subpicosecond laser pulses. We develop a theoretical model based on the coupled conservation equations of mass, momentum, and energy in both phases, explicitly accounting for thermal diffusion and viscous losses. The reciprocal coupling between thermal and acoustic fields is fully described, allowing us to quantify how frequency and fluid viscosity jointly control the penetration length of the generated acoustic waves in realistic media. Finally, we discuss the implications for theranostics, highlighting how ensembles of laser-activated particles embedded in biological tissue may be optimized for diagnostic and therapeutic applications.

Local topological markers are used to characterize Chern insulators in the presence of spatial inhomogeneities, such as boundaries and disorder. In this paper, we study the local Chern marker in systems with partial translational symmetry. We express the local Chern marker in the hybrid position-momentum basis for both open and periodic boundary conditions. We calculate the local Chern marker for a Haldane model ribbon. We show that the behavior at the two boundaries is qualitatively different from fully open geometries. We further compare the local Chern marker with the local Středa marker and show agreement in the bulk and small deviations at the boundaries that diminish with increasing system size. The correspondence between the two markers remains good if disorder is introduced, provided its magnitude remains below large values that cause substantial change of the Chern number due to Anderson physics. Finally, by exploiting the numerical efficiency due to partial translational symmetry, we study equilibrium critical behavior and the Kibble-Zurek mechanism in a weakly disordered Qi-Wu-Zhang Chern insulator. We extract relevant scaling exponents from the local Chern marker configuration and show that they converge to the analytically predicted values with increasing system size.

We study how an active bath is transduced into the internal fluctuation spectrum of a semiflexible polymer. Starting from the statistics of active forces exerted by an explicit bath of active Brownian particles, we derive an effective description in terms of temporally persistent and spatially correlated noise, and test it against simulations of both explicit-bath and implicit-noise models. We find that activity reorganizes polymer fluctuations spectrally rather than uniformly: increasing the active force predominantly enhances the lowest modes, while increasing persistence shifts the spectral weight toward progressively longer wavelengths. The theory captures this mode-level reorganization well and explains the strong qualitative correspondence between explicit and implicit active baths over a broad parameter range. In contrast, global size measures such as the radius of gyration are systematically underestimated, which we trace to activity-induced bond stretching and contour-length renormalization absent from the present fixed-contour theory. Our results show that a semiflexible polymer acts as a multiscale probe of active matter, resolving the temporal and spatial structure of nonequilibrium forcing through its mode spectrum.

Sondheimer oscillations (SO) are magnetoresistance oscillations occurring in thin films due to the commensurability between cyclotron motion and sample thickness, and are traditionally regarded as a purely semiclassical size effect. Here we develop a general quantum theory of SO for thin-film conductors in the quantum limit of a large magnetic field. We show that corrections arising from band topology modify the SO frequency, in contrast to Shubnikov-de Haas oscillations where topological information appears only in the phase. As a consequence, quantum SO provide a direct and robust probe of the full Landau level spectrum. Applying our framework to a minimal model with tunable Berry phase, we demonstrate how topology manifests itself in experimentally accessible magneto-oscillation spectra and discuss damping mechanisms including surface roughness.

The concept of gapped symmetry-protected topological (SPT) states has been generalized to gapless SPT (gSPT) states. Similar to gapped SPT states, gSPT states in one dimension exhibit universal degeneracies in their entanglement spectra. The entanglement spectra of gSPT states are further described by boundary conformal field theories, whose systematic prediction is a key open question. To address this problem, we focus on the class of gSPT states that are obtained by applying unitary SPT entanglers to trivial, critical states in one dimension. We find that the reduced density matrix of a non-trivial gSPT state can be obtained, either exactly or approximately, by applying a quantum channel to the reduced density matrix of the trivial gSPT state. This quantum channel acts only near the entanglement cut and modifies its corresponding conformal boundary condition, allowing us in turn to predict the boundary conformal field theory describing the entanglement spectra. We apply this framework to gSPT states protected by various symmetries and having different central charges, and further analyze the stability of boundary conditions of the entanglement cut. Our work thereby provides a framework for systematically analyzing and understanding the entanglement spectra of gSPT states.

We study two spatially separated boundary subsystems coupled to a common structured environment under relative motion in a Gaussian open-system framework. By integrating out the environment, we obtain an influence functional governed by a dressed environmental correlator evaluated at the boundary positions, which encodes both coherent mediation and correlated fluctuations. Relative motion opens a correlated decoherence channel through Doppler-shifted spectral overlap of the boundary excitations, leading to a kinematic threshold at $v>2u_φ$. Below threshold, the dominant resonant contribution to the off-diagonal noise kernel is absent and the environment acts predominantly as a coherent mediator at leading resonant order. Above threshold, a resonant shell opens and the same environment supports a finite cross-noise channel, producing irreversible correlated decoherence. In the reduced dynamics, coherent coupling is governed by the retarded component of the dressed correlator, while the decoherence rate is controlled by its Hadamard component. These results establish a direct connection between motion-induced excitation production and correlated decoherence in open quantum systems, and point to experimentally accessible signatures in superconducting--phononic platforms through excess correlated dephasing.

Two-dimensional(2D) SnSe is an emerging 2D material exhibiting intriguing properties such as ferroelectricity and nonlinear optical response. Here, high-quality single-crystalline SnSe nanosheets were synthesized via NaCl-assisted chemical vapor deposition(CVD) method. The crystalline structure and composition of the nanosheets are confirmed by XRD, Raman spectroscopy, and XPS. Ferroelectric domains are directly observed in SnSe nanosheets using piezoresponse force microscopy(PFM). Magnetic measurements reveal a weak magnetic behavior with a Curie temperature of approximately 120 K. This work further establishes a controllable synthesis route for SnSe nanosheets, thereby paving the way for subsequent investigation of their multiferroic properties.

The non-Hermitian skin effect is nonreciprocity-induced localization phenomena in which a macroscopic number of eigenstates accumulate anomalously at the boundary, accompanied by the extreme sensitivity to boundary conditions. Here, we develop a geometric characterization of the non-Hermitian skin effect. We demonstrate that the localization length scale associated with the skin effect is encoded in the quantum metric defined solely from right eigenstates, but not in the biorthogonal quantum metric. Moreover, we show that the quantum metrics exhibit the power-law divergences at gapless points that depend on the different boundary conditions. We also reveal that cusps of the generalized Brillouin zone in non-Bloch band theory are signaled by discontinuities in the quantum metrics. We illustrate these behavior using prototypical non-Hermitian models, such as the Hatano-Nelson model and the non-Hermitian, nonreciprocal Su-Schrieffer-Heeger model.

Self-sustained oscillators produce stable periodic motion robust to dissipation. Such motion is usually achieved by work fed back into the oscillator, but its performance is often limited by frequency-dependent operation. Here we propose a scheme for self-sustained vibrations without external feedback. We consider a movable Cooper-pair box attached to the free end of a voltage-biased normal-metal pillar. The Cooper-pair box carries an Andreev current subject to an electric field applied perpendicular to the current. In the adiabatic limit, where the Cooper-pair box state follows its motion, vibrational instability occurs, pumped by inelastic Andreev tunneling. Nonlinearity of the Josephson coupling saturates the vibrational amplitude, resulting in two-dimensional self-vibrations. We discuss the advantage of this adiabatic scheme in comparison with feedback-induced self-oscillation.

Many biophysical processes begin when the fastest searcher finds a target out of many random searchers, which is called an extreme or fastest first passage time (fFPT). In some models, (i) the fFPT vanishes logarithmically as the number of searchers grows, and (ii) the fFPT can be faster for subdiffusive search compared to normal diffusion. Though mathematically rigorous, the relevance of (i) and (ii) to actual physical systems is suspect since their derivations involve searchers which move with unbounded speed. Indeed, we previously proved that the fFPT for searchers with bounded speed converges exponentially to a strictly positive minimal search time as the number of searchers grows. In this paper, we study fFPTs for a broad class of anomalous and normal diffusion models with bounded or unbounded speed. These models include scaled Brownian motion, Riemann-Liouville fractional Brownian motion, and fractional Brownian motion. For all of these models, we show that the fFPT decays logarithmically in the number of searchers and that subdiffusion can be faster than normal diffusion (we further show that superdiffusion can be slower than normal diffusion). In this sense, features (i) and (ii) are rather universal. On the other hand, we show that the parameter regimes in which (i) and (ii) are valid depend on the particulars of the individual model, and thus ambiguities remain in the relevance of these features to specific physical systems.

Exceptional points (EPs) represent non-Hermitian degeneracies where eigenvalues and eigenvectors coalesce, giving rise to enhanced sensitivity and critically damped dynamics. We demonstrate that when an EP coincides with a dissipative phase transition in an extended open Dicke model of two cavities coupled to a collective spin, the critical fluctuations are strongly amplified and governed by modified critical exponents. Numerical results reveal enhanced critical scaling in both the normal and superradiant phases, in agreement with an analytical theory based on EP-induced Jordan-block dynamics. Our results establish EPs as a mechanism to engineer critical scaling in open quantum systems, with potential applications to critical quantum sensing.

Recent experiments in rhombohedral graphene find evidence for a "self-doped" Wigner crystal (SDC) in which a slightly incommensurate Wigner crystal (WC) coexists with a small Fermi sea. We provide non-perturbative arguments that such SDCs generically arise from preempted band-inversion transitions between commensurate crystals, which motivates simple band-theory criteria for their appearance. Self-consistent Hartree-Fock calculations establish the existence of a SDC consistent with this mechanism in both the $λ$-jellium model and rhombohedral pentalayer graphene (R5G). In the $λ$-jellium model, we identify a SDC phase located between a "halo"-WC and an anomalous Hall crystal (AHC), which would otherwise be connected via a Dirac transition when pinned to commensuration; this contrasts with the WC-AHC transition, which we show cannot be connected by a continuous transition due to a mismatch of symmetry indices. In R5G, we predict a SDC phase located between a WC and a "disqualified" halo anomalous Hall crystal. We discuss in general how the Berry curvature distribution in the parent band affects the appearance of SDC, revealing a novel role of quantum geometry in inducing exotic quantum phases of matter.

We establish a quantitative bulk boundary correspondence in interacting topological insulators by relating many-body topological invariants to characteristic degeneracy structures in the entanglement spectrum. Focusing on generalized Su Schrieffer Heeger chains with higher winding number, we construct a gauge-invariant many-body winding invariant based on Pancharatnam geometric phases that remains well defined in the presence of interactions. We show that this invariant uniquely determines the low-lying entanglement-spectrum degeneracy, which exhibits a universal 4 to the power nu scaling with the winding number nu, providing a concrete formulation of bulk boundary correspondence beyond single-particle topology. Using exact diagonalization, we demonstrate the robustness of this correspondence under interactions and symmetry preserving disorder, and identify inversion symmetry as a minimal protecting symmetry that stabilizes both the quantization of the invariant and the associated entanglement degeneracies. Our results unify geometric-phase invariants and entanglement diagnostics within a many-body framework and provide a route to identifying interacting topological phases beyond band theory.

We use segment DMRG on the infinite cylinder to compute energies of charged excitations in gate-screened fractional quantum Hall states. For the $ν=1/3$ Laughlin, $ν=2/5$ Jain, and $ν=5/2$ anti-Pfaffian states, we find screening can bind like-charged anyons into molecules, with a strong dependence on filling-factor, gate distance, and fusion channel. In the Laughlin state, stable $\pm 2e/3$ molecules and larger clusters appear over a broad gate-distance window. The Jain state is molecular throughout the range we consider. In the anti-Pfaffian, binding is strongest on the hole side, where the charge-$e/2$ molecule is fused into the $ψ$ channel over a broad window of gate distances. In all three cases, screening suppresses long-range repulsion and exposes an intermediate-range attraction encoded in the oscillatory density tail of the fundamental anyon. We discuss consequences for addition spectra, interferometry, Wigner crystallization, anyon superconductivity, and entropy measurements.

The two-dimensional character of van der Waals magnets allows for efficient control of their properties via proximity effects and electrical stimuli, making them promising candidates for application in spin-electronics. We use spin-polarized low energy electron microscopy to directly image the magnetic texture of thin CrSBr on top of a Au film, uncovering a ferromagnetic ground state for CrSBr thicknesses smaller than 11 nm. We argue that the stabilization of the ferromagnetic ordering - as compared to the conventional antiferromagnetic one - is obtained via electron transfer from the Au film to the CrSBr flakes, in agreement with ab-initio density functional theory calculations. Reflected-electron spectroscopy shows clear differences in the unoccupied density of states between a few layers of CrSBr on Au and bulk CrSBr, pointing towards electronic band structure modification in thin CrSBr. This work sheds light on the possibility to tune magnetic properties of two-dimensional magnets via substrate engineering.

We prove that a violation of a Leggett-Garg inequality for bounded observables in stationary pure states and thermal states yields a rigorous lower bound on the quantum Fisher information. This turns a qualitative foundations test of realism in quantum systems into a quantitative witness of useful quantum sensitivity and, in the collective setting, into a lower bound on multipartite entanglement depth in many-body systems. We further demonstrate that Leggett-Garg violations are constrained by the same spectral moments, susceptibilities, and $f$-sum-rule bounds that organize many-body response. Our results show that temporal correlations of a single collective observable can serve as an experimentally accessible witness of many-body quantum coherence, without requiring full state reconstruction.

Ordering at arbitrarily high temperature - entropic order - has been argued to take place in a class of generalized Ising models parameterised by a real interaction parameter $p$ when $p\ge 1$. We give a rigorous proof of this conjecture. We further show that on arbitrary graphs, these models solve graph packing problems - crucially, the Maximum Independent Set optimisation problem. Due to the NP-hardness of this packing problem on generic graphs, some lattice systems will exhibit glassy phases. We call this phenomenon $entropic$ $glass$.

We construct effective $\mathrm{U}(2)$ Chern-Simons-Ginzburg-Landau theories for Abelian and non-Abelian fractional quantum Hall hierarchies for those which had previously been described only through categorical data or trial wavefunctions. Our framework captures both Abelian hierarchy states built on half-filled Pfaffian-type parents and non-Abelian hierarchies emerging from Abelian states. It reproduces all filling fractions obtained from wavefunction and categorical constructions and, moreover, uniquely determines the corresponding topological orders. We also identify an intriguing particle-hole symmetry relating two hierarchy sequences, one built on a trivial insulator and the other on the $ν=1$ integer quantum Hall state, which respectively generate the Read-Rezayi sequences and their particle-hole conjugates under the same hierarchy construction.

Active matter systems typically exhibit a trade-off between structural robustness and dynamical freedom, limiting independent control over structure and motion. Here, we show that encoding interactions in a shared field overcomes this constraint, enabling continuous tuning between stable architectures and dynamically active states. Using droplets on a vibrated fluid bath as a minimal realization, we demonstrate that individually unstable units can collectively self-stabilize through field-mediated dynamical bonds. Arising from wavefield interference, these bonds form persistent, self-healing connections that preserve architecture while sustaining motion. Droplet size sets the symmetry of the interactions, with identical droplets forming rigid $σ$-like frameworks that enforce triangular packing, while smaller droplets enable $π$-like coordination that supports higher-order symmetries. The resulting assemblies exhibit both stability and sustained collective dynamics, including spontaneous rotation and controlled migration. This work establishes a general route to programmable active matter in which shared fields reconcile structural robustness with dynamical freedom.

We study the superconducting diode effect (SDE) in a diffusive superconductor - normal metal (SN) bilayer subjected to an in-plane magnetic field. The supercurrent flows along the layers, perpendicular to the field. The SDE, manifested as an asymmetry in the critical (depairing) currents and kinetic inductance for opposite current directions, arises from an orbital mechanism due to the inhomogeneous distribution of the Meissner currents caused by a spatially varying superfluid density. Recently, Levichev et al. [Phys. Rev. B 108, 094517 (2023)] demonstrated the realization of this effect in such a structure, supporting numerical calculations for an ideal interface with an experiment. In this work, we investigate the influence of a nonideal interface with finite resistance on the SDE. Employing an analytical approach, we focus on limiting cases corresponding to weak intralayer inhomogeneities. We find that the strength of the SDE depends nonmonotonically on the interface resistance when the bilayer thickness is small compared to the coherence length. Remarkably, a nonideal interface can enhance the SDE compared to the ideal case.

We introduce a three-dimensional model for polyelectrolyte hydrogel filaments operating in a fluid environment under an electric field. The formulation builds on a morphoelastic framework for inextensible and unshearable rods, such that the filament's activity is encoded in electric-field-induced spontaneous curvatures, while hydrodynamic interactions are captured via a local approximation of Stokes flows. We employ this framework to investigate the prototypical case of a filament with elliptic cross-section clamped at its base. Under a constant and uniform electric field aligned with its axis, the filament undergoes flutter instability beyond a critical field strength, as revealed by a linear stability analysis. Depending on the model parameters, the instability is characterized by either two- or three-dimensional self-sustained oscillations. We further examine this behaviour through numerical simulations in the post-critical regime, showing that flutter may develop into large amplitude planar oscillations or more complex three-dimensional motions, through a secondary bifurcation. Although the study represents a first step towards extending state-of-the-art models for polyelectrolyte hydrogel filaments to three dimensions, the richness of the resulting dynamics achievable under time-independent forcing underscores the potential of the proposed actuation mechanism for the design of biomimetic cilia and soft robotic systems.

Low-frequency charge noise induced by fluctuating electrostatic disorder is a major limitation for semiconductor hole spin qubits. Here, we analyze the quasistatic response of a hole spin qubit to individual two-level fluctuators (TLFs). We show that, due to the anisotropy of the g-tensor, the qubit response depends on the geometry of the fluctuator-induced dipolar perturbation. We then propose a readout protocol that isolates selected g-tensor components through an accumulated Berry phase and estimate, within our readout model, an order-unity signal-to-noise ratio with a total protocol time in the tens of microseconds. Finally, using microscopic simulations, we compute the quantum Fisher information (QFI) to identify magnetic field directions and confinement regimes in which the qubit is most sensitive to disorder-induced variations of selected g-tensor components.

We study time-domain electron interferometry in a Hong-Ou-Mandel (HOM) geometry, where a thin superconductor between two quantum Hall systems acts as the beam splitter. By comparing the measurable current cross correlations at the interferometer outputs with those of a normal-conducting electronic HOM setup, we show that Andreev processes strongly affect the HOM dip. Using a combination of scattering theory and numerical tight-binding simulations for a graphene quantum Hall bar, we show that the change of charge cross correlations can be used to experimentally detect and characterize local and crossed Andreev processes.

We present a coarse-graining method applicable to dry scalar active matter with motility regulation. Our approach, based on a multiscale perturbative expansion of the backward Kolmogorov equation, does not rely on any specific microscopic dynamics for the particles' orientations. Its generality allows us to address different forms of motility regulation, from space-dependent self-propulsion speed to taxis, and to extend the analysis to a class of non-Markovian orientational dynamics. Furthermore, we identify general conditions on the microscopic dynamics that ensure the existence of an effective large-scale equilibrium regime. When the latter are violated, our theoretical framework is able to quantitatively capture the emergence of large-scale particle currents. We directly apply our coarse-grained theory to several models of self-propelled agents, ranging from single particles to active polymers, and test our analytical predictions with numerical simulations. Finally, we show that our theory naturally extends to active matter with density-mediated interactions, such as quorum sensing, with potential applications to self-organizing soft materials.

Understanding how microscopic motility shapes emergent collective behaviors is a challenging task in active matter, especially when self-propulsion is regulated by external cues or via quorum-sensing interactions. To address this problem, we derive a closed hydrodynamics for scalar active matter with spatially-regulated motility, under general hypotheses for the microscopic dynamics of the particles' orientations. We show that, at large scales, the contribution of the latter is entirely captured by the autocorrelation tensor of the orientations. This allows us to establish a macroscopic equivalence within a broad class of motility-regulated active systems, from single particles to active polymers. Our formalism allows us to reveal a new form of motility-induced phase separation for quorum-sensing active polymers, which we term anti-MIPS, where dense phases exhibit enhanced activity relative to dilute regions. Our theory shows that anti-MIPS generically arises for motility-regulated agents with internal structure, uncovering the existence of several distinct transition pathways.

Porous media underpin absorption, filtration, separation, and high-area interfacial transport in chemical and diagnostic systems, yet sustained directional flow through them remains difficult because tortuous pore networks and strong acoustic losses promote bypassing, weak flow, and counterflow. Here, we show that floating-electrode unidirectional transducers (FEUDTs) convert porous materials into actively pumped transport platforms by generating predominantly unidirectional surface acoustic waves (SAWs) that couple more effectively than conventional interdigital transducers across wet multilayer interfaces. By varying pore size, permeability, sample thickness, and fluid viscosity, we find that transport is strongly enhanced when the SAW wavelength is comparable to the characteristic pore dimension, providing a practical design rule for acoustically activated porous media. Under these conditions, FEUDTs drive directional flow velocities up to 0.6 mm s$^{-1}$ at sub-watt input power, about 600 times faster than diffusion alone. FEUDTs also sustain pumping in prewetted porous media, where capillary contributions are removed, yielding velocities that exceed capillary-driven flow under matched conditions while remaining far above thermally induced transport. A reduced theoretical framework captures the main experimental trends and identifies transducer architecture, pore geometry, and actuation strength as the key parameters governing long-range, tunable transport in porous functional materials.

The calculation of the decay rate of a metastable state in the path-integral formulation of stochastic processes is revisited. Previous derivations of this rate were achieved at the cost of a step that is difficult to justify mathematically. We show that this difficulty can be resolved by working with an extremal solution that arises naturally in the Ito formulation of the path integral. To make the analysis as transparent as possible, we choose a simple potential for which the extremal solution can be written in terms of elementary functions. The mechanism identified here, however, is not restricted to this example and holds more generally.

We study phonon transport in junctions of two harmonic reservoirs coupled together by a spring. The exact steady-state heat currents and thermal conductance are calculated using nonequilibrium Green's functions. We find that the heat currents follow Fourier's law and the thermal conductance has a peak whenever the phonon spectra match. At lower temperatures, however, the thermal conductance maximum may not coincide with the spectra-matching peak due to the exclusion of higher-frequency phonons, whose spectra may match, from participating in the transport. Furthermore, we find that increasing the coupling spring constant increases the thermal conductance. Lastly, the magnitude of the steady-state heat currents and thermal conductance are the same whether the direction of phonon flow is from left to right or vice versa, even with mass and spring constant asymmetry. The properties of this basic model can serve as a reference for more complicated setups of phonon transport in molecular junctions.

F$^{19}$ nuclear magnetic resonance free induction decay (FID) data are used to verify the predictions of a universal growth hypothesis for the Lanczos coefficients proposed by Parker et al. Our results strongly support this hypothesis and permit to calculate values of the growth parameter $α$ for three crystal orientations. For the magnetic field parallel the [100] crystal axis, we found $α=3.161 \times 10^{4} sec^{-1}$. The special experimental conditions required for the observability of a singularity in the analytic continuation of the FID, which from the experimental data was found to be of branch-point type, are discussed.

This article presents a numerical simulation of solvent diffusion in transition metal dichalcogenide based nanomaterials during solvothermal reaction, leading to layer exfoliation and, consequently, a reduction in the average nanoparticle size. By solving modified Ficks law of diffusion and utilizing the dynamic bond percolation model, this study examines the evolution of a system of nanoparticles. During the simulation, the effects of key parameters, such as the diffusivity variable that determines the diffusion rate, and the number of iterations needed to achieve enhanced nanoparticle size uniformity, have been analyzed. To gain more insight into the size evolution of the nanoparticles, avalanche statistics, and fluctuations in the average nanoparticle size by Shannon entropy has been utilized. The size distribution observed for different diffusivity variables and iterations has also been studied, which predicts the probability of finding the nanoparticles of specific sizes within the system. A correlation between the iteration for maximum entropy and the minimum of relative change in particle size with iteration has been established, indicating that an iteration exists that takes the system towards saturation in terms of the average size of the nanoparticles. The numerical findings indicate that the experimental parameters, such as solvent selectivity and diffusivity, as well as reaction time, play significant roles in determining nanoparticle size and uniformity, thereby enhancing potential material applications.

Spin defects in solids, such as the nitrogen-vacancy (NV) center in diamond, have emerged as a key tool for detecting nuclear spins at the nanoscale. While active nuclear spin control via radio-frequency (RF) irradiation is often unnecessary for standard spin-noise detection, it becomes essential for advanced protocols like multidimensional nanoscale NMR. In this work, we investigate nuclear spin control using correlation spectroscopy techniques. We demonstrate, both theoretically and experimentally, that the resulting nuclear spin dynamics depend critically on the initial RF phase and its orientation relative to the NV crystalline axis. Depending on these parameters, identical nuclear rotations can yield full, partial, or even vanishing contrast in the NV readout. These findings highlight a previously underappreciated aspect of spin manipulation in the spin-noise regime: the link between the phase and direction of the applied RF field and its direct impact on correlation-based experiments. Consequently, imperfect calibration of these parameters can lead to ambiguous signal contrasts and misinterpretation of the underlying nuclear spin dynamics. Our results provide deeper insight into nanoscale spin control and pave the way toward reliable multidimensional spin resonance experiments.

The thermally reversible phase transitions in aqueous solutions of the triblock copolymers known as Pluronic and their related textures are well-researched. However, their corresponding rheological properties are less studied. In particular, their high-temperature behavior is difficult to access with classical rheology. Here we demonstrated that Diffusing Wave Spectroscopy (DWS)-based microrheology allows us to study the phase transition and the associated viscoelastic properties of Pluronic F127 solutions for temperatures from 5 C to 80 C. From the measured intensity-autocorrelation functions we can extract effective viscosities and determine the critical micellization temperature and concentration. Moreover,the high EO/PO (arm-to-core) ratio of F127 and its polydispersity play a critical role in the high-temperature re-entrant liquid phase, due to decreasing solubility of PEO along with the dehydration of the PPO core. The microscopic viscoelastic moduli G'(ω) and G''(ω) help to determine these phase transitions and provide mechanical properties in the solid phase that are not readily accessible with standard multi-particle tracking techniques due to limited Brownian motion.

We investigate the emergence of many-body dynamical localization (MBDL) in the Fock space of an interacting two-mode bosonic system subject to periodic driving. Using a mapping to the paradigmatic kicked-top model, we analyze the interplay between classical chaotic diffusion and quantum interference effects. While the mean-field (classical) dynamics exhibits bounded ergodic diffusion along the population imbalance axis, the quantum dynamics reveals strong suppression of transport in Fock space, in close analogy with Anderson localization in disordered lattices. We characterize the localization length, its scaling with particle number and driving parameters, and reveal the spectral crossover from random-matrix to Poisson statistics as the many-body ensemble localizes. We highlight the connection between MBDL and discrete time crystals. Our findings offer a promising avenue to study the Anderson transition in Fock space.

While the eigenstate entanglement entropy has been extensively studied for fermionic systems, much less is known about bosonic systems. Here, we study the entanglement entropy of mid-spectrum eigenstates of Bose-Hubbard models, focusing on weakly disordered models with and without particle-number conservation, and contrasting them with the translationally-invariant model. We analyze the volume-law and O(1) contributions to the entanglement entropy via the averages over mid-spectrum eigenstates and the corresponding distributions. We derive the volume-law coefficient of the entanglement entropy by generalizing the mean-field approach from [Phys. Rev. Lett. 119, 220603 (2017)] to many-body systems with a tunable local bosonic cutoff, which agrees with previous analytical and numerical results from [Phys. Rev. B 110, 235154 (2024)]. We show that the volume-law contribution to the entanglement entropy does not change upon breaking translational invariance via on-site disorder. We then numerically study the role of the subleading O(1) contribution to the entanglement entropy. We find that, in the particle-number conserving case, it exhibits a nontrivial dependence on the particle-number density and the local bosonic cutoff, while without particle-number conservation, results suggest the emergence of a universal O(1) contribution beyond the random pure state predictions.

Collective modes are a defining signature of coupled degrees of freedom, forming a bridge between understanding of interactions in condensed-matter systems and emergent functionality. Topological magnetic textures provide a natural platform to realize and control such collective modes at the nanoscale. Here we theoretically identify and characterize low-energy collective spin-wave excitations of isolated asymmetric antibimerons and their clusters in ultrathin ferromagnetic films. We demonstrate that an isolated asymmetric antibimeron supports a discrete spectrum of localized modes, reflecting its internal degrees of freedom. When multiple asymmetric antibimerons form a cluster, inter-texture coupling leads to the splitting of these modes into $N$-fold multiplets, where $N$ denotes the number of asymmetric antibimerons. To rationalize these findings, we introduce an effective coupled-oscillator model based on meron pairs that captures the essential collective dynamics of the system. This emergent classical mechanics description reveals that the motion of asymmetric antibimeron clusters can be understood in terms of well-defined normal modes governed by topology-constrained particle-like degrees of freedom. These results establish coupled asymmetric antibimerons as a tunable platform for spin-wave based nano-oscillators, whose normal-mode spectrum is controllable through cluster size, thus providing a programmable set of low-lying resonances for these nano-oscillators.

While textiles have existed throughout much of human history as complex mechanical metamaterials, textile science has largely been overlooked by the physics community until recently. In this review, we consider the symmetry, topology, and mechanics of woven and knitted materials, showing that they represent a unique, if under-explored, regime of condensed matter. We start with the basic construction and mechanics of spun yarn, reviewing recent developments twisted bundle structures. We then introduce woven and knitted fabrics as materials with layer symmetries that can be topologically characterized as knots and links in the thickened torus. We finally discuss fabric mechanics and geometry in terms of yarn-level geometry, dissipation mechanisms, and defect structures.

We predict a giant cyclotron resonance in the nonlinear valley Hall response of inversion-asymmetric two-dimensional semiconductors subjected to crossed terahertz electric and static magnetic fields. By employing a two-band Hamiltonian that incorporates both linear and quadratic in momentum terms, thereby capturing the essential orbital texture and broken inversion symmetry, we develop a kinetic theory that accounts for antisymmetric skew scattering from impurities. Solving the Boltzmann transport equation we uncover resonant photocurrents that exhibit a sharp, polarity-switching cyclotron peak and a nontrivial polarization response dictated by the underlying D3h crystal symmetry. Our results establish a universal mechanism for frequency-selective, phase-sensitive valley current control, directly accessible in monolayer transition metal dichalcogenides. This work provides a pathway for harnessing resonant nonlinear transport in valleytronic and terahertz optoelectronic devices.

Orbital angular momentum has recently emerged as an important carrier of angular momentum in solids, offering new pathways for spin orbitronic functionality beyond conventional spin transport. Here, we investigate the orbital Hall effect which generates orbital torques and their reciprocal process viz orbital pumping and the inverse orbital Hall effect (iOHE) in non-magnet/ferromagnet heterostructures. Using two port scattering parameter measurements on Ru/Ni, Ru/Pt/CoFeB and Co/Cu/SiO2 devices, we directly probe both orbital torque driven magnetization dynamics and orbital pumping within the same device platform. We observe that the transmission coefficients satisfy the symmetry relations required by Onsager reciprocity, demonstrating reciprocal conversion between charge, orbital and spin angular momenta. Our results establish orbital pumping as the reciprocal counterpart of orbital torque. Our experimental findings provide a unified framework for orbital transport phenomena.

Spatial self-similarity is a hallmark of critical phenomena. We study the dynamic process of percolation, in which bonds are incrementally added to an initially empty lattice until the system becomes fully occupied. By tracking the gap -- the size increment of clusters upon bond addition -- and the corresponding merged cluster, we identify scale-invariant temporal patterns in both quantities throughout a large portion of the process. This reveals a form of temporal self-similarity that has not been reported before. We further establish quantitative relations between the dynamic scaling exponents and the conventional static critical exponents, which enable the determination of critical behavior without prior knowledge of the critical point. The same self-similar dynamics is observed in both bond and site percolation on lattices and networks, and extends to other systems such as explosive and rigidity percolation. Moreover, similar temporal scaling is found in the initial nonequilibrium evolution of the Bak-Tang-Wiesenfeld sandpile model, suggesting a dynamic critical behavior distinct from its equilibrium state. These results provide a unified framework for understanding critical dynamics and may find applications in a broad range of complex systems.

The bilayer skyrmion composed of upper and lower tightly coupled skyrmions on two layers with completely compensated topological charges (called as anti-topology here), has become one feasible improvement of conventional skyrmion to realize straight motion without skyrmion Hall effect, which has aroused great interest in practical applications. The present work investigates a general model (without external magnetic field) for the frustration-induced anti-topological bilayer magnetic textures with rich morphologies, and discusses the modulations of key parameters on the energy barrier and the current-driven dynamics. It is revealed that the interlayer coupling plays a key role in preventing distortion, and thus helps to reach a faster velocity. This model can be realized in various frustrated magnetic materials with antiferromagnetically coupled bilayer, providing a helpful guidance for the material design and application of topological magnetic textures.

In the statistical thermodynamics of classical discrete systems, the map from microscopic interactions to thermodynamic equilibrium configurations generally exhibits complex nonlinearity, known as "canonical nonlinearity" (CN). While conventionally characterized by the Kullback-Leibler (KL) divergence, this approach inevitably misses intrinsic nonlinearities arising from the discretization of continuous Gaussian families themselves. This intrinsic effect of unavoidable CN (UCN), has recently been quantified within a transport-information-geometric framework. However, the UCN is fundamentally limited to evaluating the discretization-induced cost for a single continuous distribution. It therefore does not capture the information-geometric cost between a continuous Gaussian reference and an actual discrete distribution, nor between states with fundamentally different supports, making it conceptually unclear how to decompose the overall CN. To address this limitation, we propose the Path-Integral UCN (PUCN) quantifying the cumulative information-geometric cost between distinct distributions. The PUCN adopts a path by (i) retaining the canonical distribution as an exponential family via the $e$-mixture (geometric mean) of the base measure, leading to an arithmetic mixture of the Fisher metric as the CN standard, and (ii) enforcing covariant changes in the discretization cell through the harmonic mixture of its second-moment matrix $M$, reflecting the uncertainty in parameter variations on the statistical manifold. The resulting PUCN provides a flexible measure of the geometric cost between arbitrary states, including those with essentially different supports. This formulation enables an explicit quantification of CN between different CDOS systems and a natural decomposition of the total CN into the UCN and a residual contribution, which has not been clearly separated in existing approaches.

Room-temperature optically active solid-state spin defects are widely known to be useful in quantum sensing applications, however, only a select range of materials have been found to host such systems. Recent measurements in the van der Waals material hexagonal boron nitride (hBN) have shown optically detected magnetic resonance (ODMR) with spin-1/2-like signatures can be explained by a charge transfer mechanism where charges move between adjacent defects forming weakly coupled spin pairs. Interestingly, these ODMR signatures have been reported in a variety of materials aside from hBN, suggesting the spin pair model provides a potentially material agnostic approach for enabling ODMR. Here, we test whether the charge transfer mechanism is supported in a different crystal phase, and report on ODMR signatures in cubic boron nitride (cBN), showing all the characteristic properties identified in hBN are preserved. We consider a selection of different cBN samples of varying size and observe ODMR from a single sub-micron cBN particle, paving the way towards sensing applications. This work further expands understanding of the ubiquity of optical spin defect pairs, and establishes the potential for exploring quantum technologies with a wider range of materials.

Topological superconductors (TSCs) in superconducting hybrid heterostructures, which integrate superconducting and non-superconducting materials, have been intensely investigated with the hope of discovering exotic non-Abelian anyons for fault-tolerant quantum computing. In this effort, a challenge for hybrid superconducting systems is controlling hybridization, which is often a balance between enhancing the superconducting proximity effect at the cost of suppressing desirable electronic properties such as strong spin-orbit interactions. Hence, discovering hybrid superconducting systems with topological properties controlled and enhanced by material geometry design without spin-orbit interactions would be intriguing to explore. In this work, we theoretically study a square superconducting network decorated with spin-polarized magnetic adatoms. We find that localized Yu-Shiba-Rusinov bound states at magnetic adatom sites collectively form a weak topological superconducting phase despite the absence of spin-orbit interactions. We then demonstrate that by tuning the Fermi energy of the network, the system can transition from a weak TSC phase to a bulk-dissociated TSC phase where the edge state bands separate from the bulk, giving rise to unexpected features such as nodal lines and co-existing bulk-dissociated edge and corner modes. Moreover, our findings highlight how hetero-dimensional superconducting metamaterials can serve as a useful template for controlling the coupling and dissociation between electronic degrees of freedom of different dimensionalities.

Coupling tailored electromagnetic fluctuations to materials provides a resource for controlling correlated quantum matter. By structuring the frequency, spatial, and modal distribution of fluctuations through a new generation of cavity quantum materials, vacuum and thermal spectra can shift phase boundaries and stabilize or suppress orders. This review organizes the field around a fluctuation-focused perspective, surveying a practical design toolbox and recent milestones, and outlining theory-experiment challenges in realistic, multimode, beyond-long-wavelength regimes. We highlight photonic observables and map opportunities for equilibrium and driven control across superconducting, magnetic, moire, and topological platforms.

We analyze mesoscopic electronic transport in a Chern mosaic: a regular pattern of domains whose electronic bands carry differing local Chern numbers. An example platform where a Chern mosaic can arise is a moiré heterostructure, where variations in the local moiré parameters can produce such domains. We compute resistances at linear response for a variety of domain wall network geometries at zero temperature and magnetic field. Simple domain configurations can exhibit zero, integer, or fractional multiples of the quantum of resistance in both the longitudinal and transverse (Hall) responses. Our simple semi-classical analysis provides a useful computational method and comparative catalog for ongoing experiments in two-dimensional topological materials.

Achieving sufficiently low residual excited-state populations remains a key challenge in superconducting quantum circuits, particularly for protocols operating close to noise limits or requiring repeated qubit initialization. Existing protocols primarily address this challenge through sophisticated control, engineered dissipation, or feedback mechanisms. Here, we demonstrate an alternative approach in which a superconducting qubit is reset using a physically distinct, intrinsically colder phononic bath. Specifically, we interface a transmon with a high-overtone bulk acoustic resonator (HBAR), enabling cooling of the qubit into GHz-frequency modes. Using this approach, we achieve a residual excited-state population of the qubit below $10^{-4}$, representing an improvement of one to two orders of magnitude compared to existing reset schemes. These results highlight the potential of phononic baths as a resource for high-fidelity qubit initialization in superconducting circuits.

Topological stabilizer codes, such as the toric and surface codes, are leading candidates for fault-tolerant quantum computation. While their decodability under stochastic noise has been extensively studied, the effects of coherent errors, which involve quantum interference, remain less explored. In this work, we study the decodability of toric codes on honeycomb and square lattices subject to $X$- and $Z$-type coherent errors generated by the $X$- and $Z$-rotations on each qubit. We establish a duality between these decoding problems and 1+1D monitored dynamics of non-interacting Majorana fermions. This duality shows that the Altland-Zirnbauer symmetry class of the dual Majorana dynamics governs the universal structure of the decodability phase diagram. We show that the honeycomb-lattice toric code (hTC) with $X$-type error is dual to class-DIII dynamics, while the hTC with $Z$-type error and the square-lattice toric code (sTC) with both error types are dual to class-D dynamics. The key distinction arises from time-reversal symmetry. In class DIII, the generic transition out of the decodable phase is dual to a measurement-induced transition between dynamical phases with area-law and logarithmic entanglement scaling. In contrast, in class D, the generic decodability transition corresponds to a transition between two topologically distinct area-law phases. To explore these transitions in microscopic models, we consider hTC and sTC with $X$-type errors as representatives and introduce a minimal two-parameter coherent error model with spatially varying rotation angles. Using analytical and numerical methods, we map out the decodability phase diagrams and characterize the universal behavior of the transitions. We find that the decodability of sTC is more vulnerable to spatially varying coherent errors than uniform ones.

Superradiant phase transitions (SRPTs), characterized by photon condensation and macroscopic matter polarization, are forbidden in equilibrium for homogeneous fields by no-go theorems. Here, we show that Floquet driving can circumvent this constraint in a Landau polariton system consisting of a two-dimensional electron gas coupled to a terahertz cavity in a DC magnetic field. An off-resonant AC magnetic field modulates the cyclotron frequency and light--matter coupling strength while leaving the diamagnetic term unchanged, generating an additional DC coupling contribution. This drives the system across a critical threshold into a superradiant phase, characterized by photon condensation and Landau-level polarization in the ground state of the Floquet Hamiltonian. This quasiequilibrium approach offers a route to SRPTs distinct from driven-dissipative schemes.

Generalized Langevin equations (GLEs) provide a powerful framework for describing slow dynamics in soft-matter systems, but deriving an exact homogeneous GLE (hGLE) for a reaction coordinate from an underlying many-body system remains generally difficult. Here, we analytically derive an exact hGLE for the relative coordinate of two tagged beads in arbitrary elastic networks. The memory kernel and effective restoring force are expressed explicitly in terms of the network matrices, thereby providing a systematic reduction of the high-dimensional network dynamics to a pair coordinate. Within the small-displacement approximation, we further derive a hGLE for the inter-bead distance, a central observable in distance-sensitive single-molecule experiments. These results therefore have broad potential applications in modeling proteins and other soft-matter systems.

In the field of orbital dynamics and orbital transport, a particularly important quantity is the so-called orbital Hall conductivity (OHC), which is expressed in terms of operators of velocity and orbital angular momentum (OAM). To overcome the difficulties in treating the unbounded position operator, very often atom-centered approximations are used, which capture only a part of the local contributions to the OAM operator. Here, we promote a new approach to quantify the OAM operator in the basis of Wannier functions, which is based on the modern theory of orbital magnetization and which captures both local and itinerant contributions to the OHC. By performing first-principles calculations for various materials, we show that significant corrections to the OHC by non-local effects arise when compared to common approximations. Our approach improves the understanding of the OAM in solids and allows for a precise estimation of various orbital effects in complex materials.

We establish an exact analytical relation between Zitterbewegung dynamics and the band geometry in two-dimensional Dirac systems. By identifying a time-independent antisymmetric observable-the \textit{areal rate of Zitterbewegung}-we show that this quantity is directly determined by the Berry curvature. Its sign defines the sense of rotation and reproduces the contributions of Dirac points to the Chern number. This relation is independent of the initial state and holds for generic two-band Dirac models. Our findings reveal a direct connection between interband quantum dynamics and topological band geometry beyond the semiclassical regime.

The conventional quantum geometric tensor (QGT) is Hermitian, with a real symmetric quantum metric and an imaginary antisymmetric Berry curvature. We show that the Zeeman QGT is generically non-Hermitian and admits a natural decomposition into normal and anomalous metric-curvature sectors. The normal sector reduces to the conventional Hermitian structure, whereas the anomalous sector contains an imaginary symmetric metric-like tensor and a real antisymmetric curvature-like tensor with no counterpart in the standard QGT. In a two-dimensional Dirac system, the anomalous Zeeman curvature develops a radial flux singularity that is Hodge-dual to the tangential winding field of the Dirac node. This recasts the same local $π_1$ topology into a curvature-flux language, analogous to the flux representation of global $π_2$ topology by the conventional Berry curvature. At the level of linear response, the four symmetry-resolved components of the gyrotropic conductivity are in one-to-one correspondence with the four components of the Zeeman QGT, while their distinct low-frequency scalings provide an additional diagnostic for isolating the underlying geometric sector. The reciprocal kinetic magnetoelectric response offers a complementary experimental route to probe the same structure. These results establish a unified framework connecting non-Hermitian Zeeman quantum geometry, local Dirac-node topology, and measurable transport signatures.

The Gaussian radial function-based Regression (RfR) method is a data-driven modeling approach that utilizes physically understandable variables from scalar time series, constructed using delay coordinates and Gaussian radial basis functions. Even when a model successfully describes an approximate trajectory of the original system, data-driven models rarely reconstruct negative Lyapunov exponents of chaotic dynamics. An ''ideal model'' should reconstruct the dynamical structure, including the negative (physically dominant) Lyapunov exponents. Comparing the ideal model and the non-ideal model, we investigate the geometric structure of the attractor of such models using the Lyapunov exponents and the corresponding Lyapunov vectors. Our investigation suggests that the ideal model reconstructs the original system's attractor as a time-delay embedding. By applying the results, we search for a method to construct an ideal model, which persists against the change in hyperparameters.

We investigate the integrable structure and soliton dynamics of a coupled modified Korteweg-de Vries (cmKdV) system with a real symmetric coupling matrix. We introduce a vector reformulation of Hirota's bilinear formalism in which both the bilinear equations and their solutions are expressed directly at the vector level, rather than through a component-wise construction.This formulation preserves the intrinsic structure of the coupled system and provides a compact framework for multicomponent nonlinear wave dynamics. Within this approach, we construct explicit one-, two-, and three- soliton solutions in closed vector form and recover the three-soliton condition directly at the vector level, confirming consistency with integrability. The method enables a unified treatment of focusing, defocusing, and mixed-sign regimes. In particular, for indefinite coupling, it reveals the existence of nontrivial vector ground states, leading to soliton solutions on non-zero backgrounds. These results highlight the structural advantages of the vector bilinear approach and open perspectives for the study of more general nonlinear excitations in multi-component integrable systems.

We investigate whether ideas from symplectic topology, in particular Gromov's non-squeezing theorem and symplectic capacity, can provide useful geometric insight into classical reaction dynamics near an index-1 saddle. Using Poincaré-Birkhoff normal form theory, we describe the phase-space structures that organize transport through the transition-state region, including dividing surfaces, normally hyperbolic invariant manifolds (NHIMs), and the associated bath-action geometry. For quadratic saddle-center and saddle-center-center models, the normal-form geometry identifies natural bath-action area scales associated with the reactive bottleneck. For anharmonic systems (Eckart-Morse and Eckart-Morse-Morse), we formulate corresponding candidate symplectic width scales -- based on transverse bath actions -- using high-order normal forms for bounded local neighborhoods associated with the reaction bottleneck near the saddle. We then present two numerical illustrations: the backward propagation of a locally coupled phase-space ball to examine linear non-squeezing behavior, and a bath-localized ensemble calculation in an anharmonic normal-form model. These computations are consistent with the idea that heavily biasing the initial phase-space distribution of an ensemble toward the high-action boundaries of the bath modes can induce a severe finite-time dynamical delay, influencing reactivity in ways not captured by total phase-space volume or flux alone. The results suggest a new geometric perspective on mode selectivity and reaction bottlenecks, while highlighting open mathematical questions concerning the precise relation between these candidate width scales and genuine symplectic capacities of suitably defined reactive neighborhoods.

Solute mixing plays a pivotal role in a broad spectrum of chemical and biological processes across natural and engineered porous media. However, current understanding of mixing dynamics remains largely constrained to steady flows in fully or partially water-saturated environments. Multiphase flow systems are generally unsteady, with moving fluid interfaces and flow paths that change in time. Despite the widespread occurrence of dynamic multiphase flows, their impacts on solute mixing are largely unknown. Here, we use experiments and numerical simulations to investigate the effect of dynamic two-phase flow on the stretching and folding of fluid elements, a fundamental mechanism driving solute mixing and reactions in porous media. We find that dynamic two-phase flows induce chaotic mixing, characterized by exponential stretching of fluid elements, leading to strongly enhanced mixing compared to steady single phase flows. By extensive numerical multiphase flow simulations, we establish dynamic steady states where we reliably measure the mean fluid stretching rate as a function of flow rate. We show that stretching is maximized at an optimum flow rate which balances fluid shear deformation against the frequency of flow reorientation by the intermittent motion of the fluid interface. The findings are rationalized by a mechanistic model linking basic multiphase flow characteristics to the stretching rate, opening new perspectives to understand and control mixing and reactions in a wide range of multiphase flow systems.

Targeted energy transfer (TET) from a Van der Pol oscillator coupled to a nonlinear energy sink (NES) is investigated under the action of a high-frequency external drive, which tunes the effective natural stiffness and promotes resonance capture, facilitating energy transfer. Using \textit{direct partition of motion} with \textit{complexification averaging}, the mechanism of energy flow and instability control through \textit{hopf bifurcation} is characterized. A spectrally evaluated Q-factor, based on FFT at the effective slow frequency, captures the resonance peaks indicating the efficient energy transfer. Finally, the energy-dissipation metric is consistent with these Q-maps and identifies the regions where transient energy pumping is most effective.

We investigate a continuum Lagrangian $p$-alignment system given by a nonlocal mean-field system of ordinary differential equations for interacting agents with weak initial data. We first establish global well-posedness of the Lagrangian dynamics and derive quantitative flocking estimates. We next construct Eulerian variables from the possibly non-injective Lagrangian flow via pushforward and disintegration, which leads to an Euler--Reynolds--alignment system featuring a nonnegative Reynolds stress and, for $p>2$, a nonlinear defect force induced by microscopic velocity fluctuations. Assuming only heavy-tailed interaction, we then show that these defect terms vanish asymptotically, leading to asymptotic mono-kinetic closure in the long-time limit. In the linear case $p=2$, we further obtain global weak solutions to the Euler--alignment system, including a sharp one-dimensional critical-threshold characterization and a global result in higher dimensions under a large-coupling condition. Finally, we establish a uniform-in-time mean-field stability estimate for the particle Cucker--Smale system in the linear regime and deduce uniform-in-time convergence toward the mono-kinetic Eulerian limit; for general $p\ge2$, we also obtain a finite-time mean-field convergence result toward the associated kinetic/Lagrangian alignment dynamics.

We investigate the emergence of synchronization in the second-order Kuramoto model with adaptive simplicial interactions on a globally connected network. This inertial Kuramoto framework describes systems, where oscillator frequencies evolve over time. Unlike most previous work that ignores inertia, we examine how inertia combined with adaptive higher-order coupling alters synchronization transitions. Using self-consistency analysis, we derive the steady-state behavior and show that adaptation qualitatively reshapes the synchronization landscape. We find that the backward transition from synchronization to incoherence remains controlled by the adaptive feedback parameter, but the forward discontinuous jump to synchronization vanishes in the thermodynamic limit. In contrast, finite-size systems still display an abrupt transition to synchronization, with its onset precisely set by the adaptation control parameter. These results show how adaptive feedback and system size together govern the onset and robustness of synchronization in inertial oscillator networks with higher-order interactions.

We investigate Boolean, totalistic cellular automata with a majority or frustrated majority vote rule, and an interaction range of variable span. These two models show a behavior which differs from the mean-field one. The majority vote model is characterized by the presence of absorbing states, and there is a related bifurcation according to the initial density, in agreement with the mean-field approximation. For initial density equal to $0.5$, however, the dynamics is dominated by a coarsening process, which stops when clusters with a definite curvature radius are established. For the frustrated majority vote model, the mean-field approximation gives chaotic oscillations or a limit cycle. Instead, we observe active patterns, with stable density. Above a certain critical value for the interacting radius there is a bifurcation of the asymptotic density as a function of the initial one.

In mathematics and engineering, control theory is concerned with the analysis of dynamical systems through the application of suitable control inputs. One of the prominent problems in control theory is controllability which concerns the ability to determine whether there exists a control input that can steer a dynamical system from an initial state to a desired final state within a finite time horizon. There is a general theory for controlling linear or linearizable system, but it cannot be applied to discrete systems like cellular automata, which is the problem of that we address in this paper. We develop a general theory for linear (and affine) cellular automata, and apply it to examples of one-dimensional and two-dimensional Boolean cases. We introduce the concept of controllability matrix and show that controllability holds if and only if the controllability matrix is invertible.

In this exploratory paper we introduce the problem of cognitive agents that learn how to modify their environment according to local sensing to reach a global goal. We concentrate on discrete dynamics (cellular automata) on a two-dimensional system. We show that agents may learn how to approximate their goal when the environment is passive, while this task becomes impossible if the environment follows an active dynamics.

Persistent neural activity underlying working memory requires sustained synaptic transmission, yet the metabolic and neurotransmitter support provided by astrocyte networks is largely absent from spatially extended neural circuit models. We introduce a coupled astrocyte-neural field model in which synaptic efficacy is regulated by depletion and recovery of a conserved resource pool recycled and spatially redistributed through diffusively coupled astrocytes. We obtain explicit stationary bump profiles and self-consistency conditions for bump width and amplitude on a canonical ring architecture. Linearizing about these solutions while carefully accounting for perturbations at bump boundaries, we analyze the resulting spectral problem governing stability. Our analysis, supported by numerical simulations and low-dimensional Fourier truncations, reveals a two-stage stabilization mechanism: astrocytic diffusion smooths resource asymmetries created by small bump displacements, and synaptic replenishment transfers this smoothing back to the synaptic pool. Together, sufficiently strong diffusion and replenishment suppress drift instabilities and enlarge the parameter regime in which stationary bumps persist.

A fragmented landscape reduces the impact of interspecies connectivity, leading to higher diversity levels than otherwise possible in a connected landscape. Reconnecting a previously fragmented landscape initiates an extinction event, preferentially weeding out more highly connected species. A sequence of fragmentation-coalescence events will drive the ecosystem to higher levels of diversity in a ratchet-like effect, than if the landscape continuously remained connected.

We demonstrate a thermodynamic engine whose working substance is a sine-Gordon soliton in a heterogeneous current-driven Josephson junction. We show that solitons can act as thermodynamic working substances whose internal spectral structure enables energy conversion beyond conventional few-level engines. By dynamically deforming the soliton using a controllable dipole current, the internal bound-state spectrum of the soliton can be engineered in time, enabling a finite-time Carnot-like cycle based on spectral control, in close analogy with quantum heat engines. Mapping the instantaneous nonlinear field configuration to an effective Schrödinger operator, we reveal how bound states appear, approach the continuum threshold, and disappear during the cycle. Comparing three thermodynamic descriptions (full nonlinear field dynamics, a coarse-grained mesoscopic model, and a two-level spectral model), we show that few-level descriptions systematically underestimate the engine performance. The enhanced efficiency arises from the extended nature of the soliton, whose internal spectral degrees of freedom provide additional energy storage and transfer channels. Our results reveal a general thermodynamic principle: extended nonlinear excitations with particle-like behavior can serve as tunable working media, whose internal spectral degrees of freedom provide additional reversible channels for energy storage and transfer beyond those of few-level systems.

We rigorously prove the bifurcation of slow-moving pattern interfaces with general direction in a two-dimensional Swift-Hohenberg-type model close to a Turing instability for a large class of nonlinearities. These interfaces describe the invasion of stripe and hexagonal patterns into the spatially homogeneous state and model a possible mechanism for pattern formation, as observed in a wide range of real-world applications. For this, we develop a rigorous framework to establish the existence of such solutions using spatial dynamics and non-standard centre manifold theory. Our approach exploits geometric and algebraic structures generic to $\mathrm{O}(2)$-symmetric pattern-forming systems near a Turing instability, and addresses fundamental technical challenges due to a non-uniform spectral gap around the imaginary axis, quadratic resonances induced by the hexagonal structure, and the high-dimensional phase space of the reduced equations.

This paper presents an improved Matlab routine, FO_LE, for the numerical computation of Lyapunov exponents of fractional-order systems modeled by Caputo's derivative. It is conceived as an enhanced version of the former FO_Lyapunov and FO_NC_Lyapunov codes for commensurate and non-commensurate orders, respectively. The proposed approach replaces the Gram-Schmidt orthogonalization procedure with QR-based reorthonormalization and uses the new quadratic LIL predictor-corrector scheme for the integration of the extended variational system. Compared with the former implementations, the present routine benefits from the higher order of the fractional integrator LIL and applies to both commensurate and non-commensurate models. Like the previous code, FO_LE retains the full memory structure of the underlying Caputo model. The Matlab code for the LIL solver and for the computation of Lyapunov exponents with FO_LE are provided, while a fast implementation of LIL for commensurate and non-commensurate orders, LIL_nc, is available on MathWorks File Exchange. A benchmark problem with exact solution is used to compare the LIL-based solver with ABM-type methods, whereas the Rabinovich-Fabrikant system illustrates the computation of Lyapunov exponents in different dynamical regimes. The results indicate that the proposed implementation is a compact, robust, and efficient tool for the numerical study of stability and chaos in fractional-order systems.

Accurate segmentation of surgical instruments in robotic-assisted surgery is critical for enabling context-aware computer-assisted interventions, such as tool tracking, workflow analysis, and autonomous decision-making. In this study, we benchmark five deep learning architectures-UNet, UNet, DeepLabV3, Attention UNet, and SegFormer on the SAR-RARP50 dataset for multi-class semantic segmentation of surgical instruments in real-world radical prostatectomy videos. The models are trained with a compound loss function combining Cross Entropy and Dice loss to address class imbalance and capture fine object boundaries. Our experiments reveal that while convolutional models such as UNet and Attention UNet provide strong baseline performance, DeepLabV3 achieves results comparable to SegFormer, demonstrating the effectiveness of atrous convolution and multi-scale context aggregation in capturing complex surgical scenes. Transformer-based architectures like SegFormer further enhance global contextual understanding, leading to improved generalization across varying instrument appearances and surgical conditions. This work provides a comprehensive comparison and practical insights for selecting segmentation models in surgical AI applications, highlighting the trade-offs between convolutional and transformer-based approaches.

We study the nature of chaos in finite and infinite dimensional systems through a comparison between the Kuramoto Sivashinsky (KS) equation, the Lorenz system, and a Lorenz type reduction of the KS equation proposed by Wilczak. Numerical simulations of the KS equation reveal intrinsic spatio temporal chaos, with disorder evolving simultaneously in space and time. In contrast, the Lorenz system and the Wilczak reduction exhibit low dimensional temporal chaos lacking spatial complexity. Lyapunov exponent analysis highlights the finite-dimensional convergence properties of the reduced systems and underscores the fundamentally different dynamical nature of chaos in the KS equation. In particular, we demonstrate that low-dimensional reductions may reproduce transient chaotic signatures but do not necessarily retain the structural properties of infinite-dimensional dissipative systems.

The propagation of the quantum states of light in dispersive and anisotropic media is a fundamental problem in quantum optics. We present a unified theoretical framework for the propagation of the quantum states of light in voltage-controlled nematic liquid crystals, incorporating both material dispersion and electrically tunable birefringence. By treating photons as finite-bandwidth wave packets, we derive analytical expressions for group velocoity, temporal walk-off, and phase evolution of orthogonally polarized modes. The results demonstrate that nematic liquid crystals can serve as electrically tunable quantum photonic devices capable of manipulating photon arrival times, polarization correlations, and temporal indistinguishability of entangled photon pairs. These results show the direct relevance to quantum communication and photonic quantum information processing.

The topological confinement is a new mechanism that allows the transmission of cutoff orbital angular momentum (OAM) modes with negligible loss in ring-core fibers (RCFs) and provides a natural immunity against mode coupling. We investigate the influence of fiber design parameters and wavelength on the characteristics of topologically confined modes (TCMs) in step index ring-core fibers (SI-RCFs), and propose a type of graded index ring-core fibers (GI-RCF) with better characteristics. Furthermore, as TCMs occurs in structures with high refractive index difference and are often accompanied by relatively high scattering loss, we fabricate a type of low-loss SI-RCF and observe the stable existence of 24 low-loss TCMs in total. Subsequently, we use an analytical model to estimate the maximum signal-to-noise (SNR) and spectral efficiency (SE) of the fiber, demonstrating its strong capacity advantages.

Mie-void metasurfaces have so far been developed mainly in reflection, where subwavelength voids embedded in high-index media support localized resonances and spectrally selective optical responses. Yet, many optical systems could benefit from integrating such optical elements operating in transmission mode. Motivated by this great need, we hereby introduce Mie-void metasurfaces operating in transmission. To allow for their operation in the visible range, our Mie-voids are implemented using the silicon-rich nitride (SRN) platform. We show that this transition from reflection to transmission is not a simple change in geometry: placing the voids in a finite film on a substrate introduces slab-guided and Fabry-Perot-like contributions that hybridize with the underlying Mie-void response. Rigorous coupled-wave analysis shows that the dominant spectral transformation occurs when the semi-infinite host is replaced by a finite SRN film, while the substrate acts mainly as a secondary perturbation. Thickness-dependent dispersion maps reveal an avoided crossing between interacting modes, supporting the interpretation of a hybrid transmission regime and identifying film thickness as a clean parameter for tracking the evolution of the coupled modal structure. Experimentally, we realize transmission-mode structural colors by varying the void depth and observe good agreement between measured and simulated spectra and chromaticity coordinates. By spatially programming the void depth, we further demonstrate transmitted-light patterns and image encoding within a single metasurface architecture. These results establish transmission-mode Mie-void metasurfaces as a viable inverse-dielectric platform operating in transmission, with plethora of potential important applications such as transmissive spectral filtering, optical encoding, and display-oriented photonic elements, to name a few.

We present a computational platform for modeling chemical reactions in complex molecular environments, focused on ligand-protein binding in drug discovery. The platform implements our new quantum-in-quantum-in-classical (QM/QM/MM) multiscale embedding model that integrates molecular dynamics with a quantum-information-enhanced density matrix embedding theory and quantum chemistry solvers, including explicit solvent. Quantum-information metrics are utilized to generate entanglement-consistent orbitals, enabling a high-accuracy description of strongly correlated regions. The framework supports multiple computational backends, including multi-CPU, NVIDIA multi-GPU architectures, and quantum hardware (IQM, IonQ, IBM) integrated under CUDA-Q, and is designed for compatibility with future fault-tolerant quantum systems. The new platform's capabilities are demonstrated by modeling covalent docking of zanubrutinib to Bruton's tyrosine kinase via a Michael addition mechanism, computing the full reaction energy profiles and energy barriers at a reduced computational cost relative to existing methods. As a 2nd-generation anticancer agent, zanubrutinib serves as a proof of concept for covalent inhibitor discovery. Accurate first-principles reaction barrier estimations provided by our method can contribute to reducing false positive and negative rates in drug discovery pipelines. Scalability is validated through benchmarks on GPU clusters, cloud-based CPU infrastructures. We demonstrate integration with quantum devices (up to 20 qubits), alongside resource estimates for fault-tolerant quantum computing, indicating potential speedups of up to 20x. Beyond single reactions, the platform supports the construction of reaction networks in chemical metric space, facilitating ligand screening and systematic exploration of reactive pathways.

Time is the odd dimension out: Unlike space, it follows the arrow of time, forbidding back-reflections and requiring momentum yet not energy conservation. Tailored temporal variations manipulate momentum bands and engineer waves in time. We show that momentum bands exhibit unique topology, hidden when conventionally considering energy bands: Complex momentum bands may wind, mandating topological localization at time interfaces. We observe this effect in photonic quantum walks and study it under disorder. Remarkably, unlike any known topological phenomenon, the topology is immune against arbitrarily strong disorder. Only exotic conditions through extreme spatiotemporally random non-Hermiticity can destroy it. Our findings uncover a disorder-immune type of topological physics, inviting explorations of complex momentum or energy-momentum topology with potential applications like ultrarobust lasing, temporal pulse shaping or amplification.

The ubiquitous $3/4$ metabolic scaling exponent, known as Kleiber's law, has long been attributed to the minimization of viscous dissipation within fractal transport networks. In this paper, we invert this standard narrative, demonstrating that Kleiber's law is fundamentally a signature of pulsatile wave physics rather than steady-state geometry. By coupling local branching optimization to global allometry, we derive the exact generalized metabolic exponent $β= dα/(2d+α)$, which strictly maps local transport microphysics to global organismal scaling. We show that dynamic wave-impedance matching in the proximal vasculature uniquely enforces $β= 3/4$ in three dimensions. This bound is dynamically protected: no static optimization of a viscous network can reproduce it. Consequently, we analytically predict the critical body mass for the wave-to-viscous transition, successfully explaining the empirical shift to steeper allometric scaling ($β\approx 0.9$) in small mammals and invertebrates with no free parameters. Furthermore, we demonstrate that the classical West--Brown--Enquist (WBE) derivation is structurally divergent under its own geometric assumptions, failing at the required proximal-dominance limit. Our framework is validated across nine biological systems spanning five phyla -- including vertebrate vasculature, insect tracheae, plant xylem, and sponge canals -- accurately predicting empirical branching exponents from independent biophysical measurements. Ultimately, we establish a general allometric equation of state that organizes diverse biological networks into discrete universality classes, generating falsifiable predictions across clades from shrews to flatworms.

Optical imaging of quantum emitters is essential for a wide range of quantum applications. Conventional confocal imaging relies on point-by-point raster scanning, which is inherently time-consuming and photon-inefficient, particularly for sparse emitter distributions and photon-limited samples. Here, we demonstrate a compressive sensing-based imaging approach, where spatially structured wide-field excitation replaces raster scanning, enabling reconstruction of sparse emitters. In our implementation, random binary patterns are used to acquire compressive measurements, from which the spatial fluorescence distribution is reconstructed using a GPSR-BB algorithm. We experimentally demonstrate this approach using nitrogen-vacancy (NV) centers in diamond as a representative platform, with high-fidelity image reconstruction achieved using only approximately $20\%$ of the measurements required for conventional raster scanning. In addition to intensity reconstruction, we extend this framework to reconstruct spatial maps of the second-order correlation function $g^{(2)}(0)$ from compressive measurements. This enables identification of single-photon emitters through antibunching signatures using significantly reduced data.

Existing hologram super-resolution (HSR) methods primarily focus on angle-of-view expansion. Adapting them for volumetric spatial up-sampling introduces severe quadratic depth distortion, degrading 3D focal accuracy. We propose CV-HoloSR, a complex-valued HSR framework specifically designed to preserve physically consistent linear depth scaling during volume up-sampling. Built upon a Complex-Valued Residual Dense Network (CV-RDN) and optimized with a novel depth-aware perceptual reconstruction loss, our model effectively suppresses over-smoothing to recover sharp, high-frequency interference patterns. To support this, we introduce a comprehensive large-depth-range dataset with resolutions up to 4K. Furthermore, to overcome the inherent depth bias of pre-trained encoders when scaling to massive target volumes, we integrate a parameter-efficient fine-tuning strategy utilizing complex-valued Low-Rank Adaptation (LoRA). Extensive numerical and physical optical experiments demonstrate our method's superiority. CV-HoloSR achieves a 32% improvement in perceptual realism (LPIPS of 0.2001) over state-of-the-art baselines. Additionally, our tailored LoRA strategy requires merely 200 samples, reducing training time by over 75% (from 22.5 to 5.2 hours) while successfully adapting the pre-trained backbone to unseen depth ranges and novel display configurations.

Phase-sensitive pump-probe hyperspectral imaging is a precise technique for absolute two-beam measurements of the optical Kerr coefficient ($n_2$). The irradiance profile is characterized and background effects are rejected by rastering the pump beam across the probe beam to yield a complex-valued hyperspectral image of the pump-induced nonlinear response. Information about the temporal irradiance profile is carried in the spectral response. The technique is demonstrated by measuring $n_2$ of a fused silica sample near 1 $μ$m wavelength and benchmarked against a measurement using Z-scan [Sheik-Bahae et al., IEEE J. Quantum Electron. 26, 760-769 (1990)], the most commonly used single-beam technique. Good agreement is found when the two-beam grating effect from the Raman contribution to the nonlinearity is considered. Uncertainty contributions are described in detail and the outlook is discussed for improvements in precision.

We present a scalable method for geolocalizing buried fiber-optic cables using Distributed Acoustic Sensing (DAS) and traffic-induced quasi-static seismic signals. Assuming access to one end of the fiber, the method fuses DAS measurements with vehicle trajectories obtained from either video tracking or vehicle-mounted GPS. The fiber geometry is estimated by minimizing the mismatch between the measured and physics-based synthetic strain-rate maps. The framework combines a matched-filter initialization with neural-network-based trajectory optimization, enabling robust convergence under realistic noise and trajectory-uncertainty conditions. Simulation and field experiments demonstrate sub-meter localization accuracy, often on the order of tens of centimeters, and strong agreement with manual calibration by tap-testing. This approach provides a practical tool for mapping poorly documented underground fiber infrastructure and for supporting urban sensing applications.

We develop a general theory of null-constrained parameter estimation in a time-reversed Young (TRY) interferometer, where measurement is performed through programmable source-basis encoding with a fixed detector. We address the fundamental question of how to design source patterns that enforce a true metrological null -- vanishing nominal response at the operating point -- while preserving finite first-order sensitivity to the parameter. Under a general shot-noise-limited channel model, we show that the optimal null-constrained receiver is obtained by projecting the derivative response onto the subspace orthogonal to the nominal background in the inverse-noise metric. This yields a constructive solution in which the optimal source-basis code is given by the inverse-noise-weighted derivative response with its background-parallel component removed. We further derive an exact and universal information-retention law: the locally accessible Fisher information is reduced by a factor $1-χ^2$, where $χ$ quantifies the inverse-noise overlap between the nominal and derivative response vectors. This result establishes a precise geometric interpretation of the cost of null enforcement. Numerical examples demonstrate the null-coded TRY receivers can retain nearly the full local information and can be accurately implemented using binary and positive-only source patterns. These findings identify source-basis null engineering as a distinct and practically viable modality for derivative-mode sensing, with implications for superresolution metrology and programmable optical measurement architectures.

Solid-state high-harmonic spectroscopy allows the study of strongly driven ultrafast electron dynamics. Microscopically, high harmonics are generated by strong-laser-field acceleration of electron-hole pairs through the lattice. At finite temperatures, atomic-scale structural fluctuations are ubiquitous and are expected to influence the electron-hole trajectories. Yet, the effect of thermal lattice fluctuations on solid-state high-harmonic generation (HHG) has not been quantified. Here, we demonstrate a profound sensitivity of HHG to thermal lattice fluctuations, by characterizing the temperature dependence of HHG in Re6Se8Cl2, a superatomic semiconductor. As the sample temperature is decreased, the high-harmonic yield exhibits a slow increase, followed by an abrupt increase below 50 K, consistent with the temperature at which lattice vibrations are strongly suppressed. Our calculations show that thermal lattice fluctuations both weaken the harmonic response from individual distorted configurations and induce phase dispersion across the ensemble, leading to a pronounced suppression of the coherently emitted harmonics. We show that this effect can be interpreted in terms of an effective electronic dephasing time that varies with temperature. Our results are relevant to dephasing in broad strong-field phenomena, including lightwave electronics and Floquet engineering. The wide tunability of superatomic crystals further enables materials-controlled strong-field physics.

Unmanned aerial vehicles (UAVs) operating in confined, cluttered environments face significant performance degradation due to nonlinear, time-varying unmodeled dynamics-such as ground/ceiling effects and wake recirculation-that are unaccounted for in traditional controllers. While learning based compensators (e.g., MLPs, TCNs, LSTMs) struggle with historical data dependency, vanishing gradients, and prohibitive computational costs, this work pioneers the integration of a deep photonic reservoir computer (PRC) with feedforward control to overcome these limitations. Harnessing semiconductor laser dynamics and optical feedback, our hardware implemented deep PRC architecture achieves intrinsic temporal memory without explicit historical inputs, while reducing training time from hours to milliseconds and slashing inference latency to nanoseconds. Reliable high-performance CFD simulations capturing proximity-induced flows demonstrate that deep PRC delivers residual-force prediction accuracy comparable to or exceeding TCN/MLP baselines, while training only a linear readout layer via ridge regression. By injecting these predictions into a nonlinear feedback PID controller via a feedforward channel, the framework significantly enhances closed-loop tracking stability in confined spaces. Essentially, this work establishes the first deep PRC-based lightweight, ultrafast solution for real-time UAV dynamic compensation, with promising extensibility to unseen scenarios with more complex fluid environments.

The relationship between the pseudogap and superconductivity remains a central puzzle in the physics of cuprates. Hydrostatic pressure provides a clean tuning parameter free from chemical disorder, yet probing the microscopic energy scales of these phases under compression has remained experimentally challenging. Here, we utilize ultrafast optical spectroscopy to construct the high-pressure phase diagram of the underdoped cuprate Bi$_2$Sr$_2$CaCu$_2$O$_{8+δ}$ up to 37 GPa. Our results reveal a striking dichotomy within the pseudogap state: while the onset temperature $T^*$ rises monotonically with pressure, the energy gap $Δ_{\mathrm{PG}}$ is continuously suppressed. In contrast, the critical temperature $T_{\mathrm{c}}$ and the superconducting gap $Δ_{\mathrm{SC}}$ trace a correlated dome-like trajectory, demonstrating that superconductivity evolves independently from the pseudogap. Furthermore, an abrupt collapse of the gap ratio $2Δ_{\mathrm{SC}}/k_{\mathrm{B}}T_{\mathrm{c}}$ near 8 GPa marks a pressure-driven dimensional crossover, quenching two-dimensional phase fluctuations to stabilize global three-dimensional coherence. Upon reaching 37 GPa, the superconducting condensate is completely quenched into an insulating-like state. By resolving the extended phase evolution, our findings disentangle the pseudogap and superconducting orders, establishing a rigorous experimental basis for the pairing mechanism of high-temperature superconductivity.

The increasing demand for multi-wavelength optical sources to support dense wavelength-division multiplexing (DWDM) channels has driven the development of compact and reconfigurable multi-wavelength fiber lasers. Here, we demonstrate a continuously tunable and deterministically switchable multi-wavelength erbium-doped fiber laser based on nonlinear polarization rotation (NPR) in a compact all-fiber ring cavity. By controlling the intracavity birefringence, NPR acts as a reconfigurable comb filter that enables flexible wavelength selection without modifying the cavity architecture. The laser supports stable spectral states ranging from single- to seven-wavelength mode-locking, enabling reversible wavelength switching and activation/suppression of individual channels. The selectable spectral states can be mapped to binary bit operations, where each wavelength channel represents a controllable logical state. The behavior arises from the interplay between NPR-induced birefringent comb filtering and nonlinear phase modulation, providing a simple and compact platform for reconfigurable multi-channel ultrafast sources for DWDM and photonic signal processing.

Collective subradiant states in waveguide QED are highly sensitive to disorder, limiting their scalability and robustness. We propose a deterministic approach to engineering atom arrays based on a Minkowski sum construction, generating quasi-disordered structures with built-in correlations. This leads to mode-selective radiative coupling: interactions between dark modes are parametrically suppressed, while bright modes can hybridize. We study the stability of these subradiant and bound-state-like modes against moderate positional disorder. Our work provides a route to robust, analytically controllable subradiance through engineered quasi-disorder, with direct relevance to atom-waveguide and circuit QED experiments.

The fermion sign problem remains the primary obstacle in simulating the thermodynamic properties of various fermionic systems. In this work, we present a sign-blocking method to mitigate the numerical instability inherent in the sign problem. In the sign-blocking method, the Monte Carlo importance sampling remains identical to traditional methods; instead, the sign-blocking method is applied during the post-processing of signed samples. Given the significant progress in simulating the 2D Fermi-Hubbard model over the past decade, a wealth of energy benchmarks is available for comparison. Consequently, we use the 2D Fermi-Hubbard model as a benchmark to validate the sign-blocking method. Surprisingly, our results align exceptionally well with existing state-of-the-art benchmarks, even in regimes previously considered challenging. The physical mechanism of the sign-blocking method lies in uncovering the correlation between energy and sign factors through data blocking, thereby successfully inferring the fermionic system's energy. Our findings suggest that the sign-blocking method holds promise for complex quantum systems, particularly when combined with appropriate simulation techniques such as auxiliary-field formalisms that trace out the fermionic degrees of freedom.

Bound states in the continuum (BICs) are spatially localized eigenmodes that remain perfectly confined even though their energies reside within a continuum of radiating modes. BICs were predicted in 1929, but their experimental realization awaited more than 8 decades. Following their experimental observation, BICs were explored in a variety of wave systems, and found to exhibit a plethora of fundamental features such nontrivial topology and extremely high Q-factor. Recently, with foundational advances in the new field of electromagnetic waves in time-varying media, BICs were predicted to exist in the time domain, with their wavenumber embedded in a continuum of unbound momentum modes. Here, we present the first experimental realization of the time-domain Bound States in the Continuum. We use a transmission-line network with a time-modulated wave-impedance and show that a sinusoidal wave launched into the network naturally evolves into a time-domain BIC with a well-defined peak and decaying-oscillating tails. We show that the time-domain BIC is anti-symmetric despite the symmetric nature of the modulation. These experiments pave the way for exploring new phenomena in the fields of BICs and time-varying wave-systems in nonconservative regimes where time-translation symmetry is broken.

We propose a novel holographic recording technique to improve the recording speed for holographic data storage (HDS). In this technique, holograms are recorded by scanning a digital micromirror device (DMD) that displays a data page with a focused, power density-increased line beam, while synchronously shifting the recording medium and a spherical reference wave. This approach eliminates the stop-and-go motion of driving mechanisms, enabling rapid, continuous, multiplexed recording. Although the recorded localized holograms retain only partial information about the data page, the entire page can be reconstructed simultaneously using the spherical reference wave. Experimental demonstrations achieved a bit error rate (bER) of less than 10% at a 5 ms exposure time and stable multiplexed recording at 150 Hz. The proposed technique represents a significant step toward the realization of practical HDS systems.

The analysis of income and wealth inequality is often constrained by the lack of reliable data. In this work, we introduce a proxy-based approach in which sports performance data are used to mimic economic distributions. In particular, the total runs scored by a batter in a single T20 season are treated as an analogue of annual income, while the cumulative runs scored across all seasons are taken to represent individual wealth. Using run distributions from the Indian Premier League (IPL), we explore how inequality evolves and estimate its upper limits in the absence of policy intervention. Our findings indicate that inequality derived from seasonal runs closely resembles the highest levels of income inequality observed in real-world data. In addition, the distribution of cumulative runs evolves over time and gradually approaches a limiting value consistent with global patterns of wealth inequality, in line with the Pareto principle. Overall, this study shows that proxy systems can capture essential features of economic inequality and offer a useful way to understand its inherent limits.

We develop a microscopic theory of optical absorption and structured-light dichroism in a nonlocal minimal-coupling framework. Starting from the minimal-coupling Hamiltonian, we express absorption as a bilinear functional of the electromagnetic vector potential and the nonlocal current response, providing a general description of light--matter coupling in optical vortex beams and other inhomogeneous fields. Dichroic signals are identified as helicity-odd projections of the nonlocal response kernel, and the response is resolved into symmetry, tensorial, and mode-space sectors. For single helical modes, the theory yields diagonal OAM-resolved contributions together with the corresponding selection rules and symmetry constraints. For mixed modes, interference between distinct OAM components provides access to off-diagonal coherence of the nonlocal kernel, with a tensor structure determined by the polarization composition of the field. The theory also clarifies how local structures associated with symmetric spatial dispersion, optical chirality, and tensorial anisotropy emerge as gradient-level manifestations of the underlying nonlocal response.

In cross-section-space probability table constructions, reaction-channel levels are reconstructed on fixed total-subgroup nodes and probabilities. Although the standard full-matching reconstruction is uniquely determined, it does not in general preserve componentwise nonnegativity of the channel levels. We impose nonnegativity both for physical interpretability and because, on fixed positive total-subgroup nodes and probabilities, it provides a sufficient structural condition for nonnegativity of the folded effective cross section over all dilutions. We therefore formulate an admissible constrained reconstruction problem on the fixed subgroup support, in which selected low-order channel information is retained exactly and the remaining matching conditions are fitted in a weighted least-squares sense. After null-space reduction, the problem becomes a convex optimization problem with linear inequality constraints. For the single-retention formulation, nonnegative feasibility is automatic when the retained \(0\)-order aggregate is nonnegative, whereas for a two-retention variant it additionally requires a compatibility condition with the fixed total-subgroup nodes. Numerical results for a representative U-238 capture benchmark show that nonnegativity violations are confined to a small subset of energy groups. On these groups, the admissible reconstruction restores nonnegativity, but at the cost of some response-level deterioration relative to full matching. In the comparison, the single-retention formulation shows the more stable overall behavior.

Simulating multiscale flows with moving boundaries, such as hypersonic multi-body separation and flows in micro-electro-mechanical systems (MEMS), requires robust numerical methods that couple mesh deformation with complex flow physics. This paper presents a hybrid overlapping moving-mesh technique developed within the unified gas-kinetic scheme (UGKS). To mitigate the Courant-Friedrichs-Lewy (CFL) constraint, we extend the implicit unsteady UGKS solver to support moving meshes, incorporating memory-efficient data handling and parallel computing optimizations to maximize computational efficiency. Validated against hypersonic multi-body separation and thermal rarefied MEMS flows, the proposed scheme accurately resolves complex, dynamic multiscale phenomena. The results confirm that this robust and efficient method provides a highly reliable tool for modeling dynamic flow interactions in complex geometric configurations.

An extension of the Ising model is proposed as a viable alternative for data with values $-1$, $0$ and $+1$ in the inverse problem, i.e., estimation of the parameters. This model is called the Blume-Capel (BC) model, adapted from physics for small networks. The advantage of the BC model is not only the fact that it is possible to have a neutral (centrist) position on the response scale, but also that this model allows for three stable states. We illustrate magnetisation properties of the BC model using simulations and mean field results. For estimation of the BC parameters, we show that the BC model is part of the exponential family of distributions and show that the model is identified, except for the (inverse) temperature. We then show that combining pseudo-likelihood with lasso yields accurate parameter recovery for the BC model, even in small networks. Moreover, confidence intervals with good coverage properties can be obtained using the desparsified lasso together with sandwich and shrinkage techniques. We apply the methods to data obtained from the online platform \textit{Stemwijzer}, intended to aid people in deciding for whom to vote.

This study presents a first-principles study at the level of hybrid-level density functional theory of the sodium intercalation process in a layered potassium birnessite (a layered manganese dioxide, δ-MnO2). Understanding the intercalation processes of δ-MnO2 is a vital step in advancing its potential innovative applications. Through a formation energy formalism, we analyze the stability of the structure as sodium ions (Na+) are intercalated between layers. Simulated Raman spectra allow us to find relationships between the vibrational and structural properties of the material, i.e. we identify the most important vibrational modes and related them to the structural/geometrical change. The diffusion of Na+ and K+ ions in birnessite is studied by transition state theory, determining the energy barriers to ion displacement in the interlayer. The symmetry and planar density of the system are characterized by simulated X-ray diffraction and geometrical analysis of the optimized structures. Through binding energy analysis, we also find that the Na+ ions are more loosely bound to the lattice as they reach the saturation limit. Finally, the electronic properties are studied via spin-polarized densities of states. As intercalants are added, the electronic properties are profoundly modified, resulting from modification of Mn oxidation states, lattice distortions, and symmetry effects. Moreover, some of the intercalated structures behave as bipolar magnetic semiconductors with potential applications in spintronics devices. In other words, the band gaps and magnetic behavior of the system can be controlled by intercalation. This work provides an overarching analysis of intercalated birnessite and describes the essential properties of potassium birnessite and co-intercalation with Sodium as a next-generation energy, electronic, and spintronic material.

The elastic behavior of materials is of critical importance for the design, fabrication, and testing of industrial and structural components. The ease with which the wave angle of incidence can be varied makes ultrasonic techniques well suited for the characterization of anisotropic materials, whose properties are direction-dependent. This work aims to develop an ultrasonic goniometry method in which a wave is transmitted through a sample while scanning over spherical coordinates. A plane-wave model is formulated that accounts for fluid-solid interfaces and is applicable to a wide range of sample thicknesses. The model assumes general anisotropy, enabling the characterization of materials with symmetries up to triclinic, and does not require precise sample alignment. Specially designed transducers support the plane-wave approximation, thereby avoiding the need for more computationally expensive finite-beam models. Furthermore, implementation of the forward model on GPU architectures significantly reduces the computational cost associated with the numerous evaluations required during the waveform fitting inversion. The introduction of optimal zeroth-order bounds is used to tightly delimit the search space, and an isotropic self-consistent solution is shown to provide an effective initial guess. Finally, measurements on plate-like samples are compared with the literature and diffraction-based methods.

We investigate linear resonant absorption by a dense ensemble of molecules confined to a subwavelength layer in two geometries: (i) a free-standing film in homogeneous space and (ii) the same film placed at a controlled distance from a reflecting surface. In both cases, increasing the effective light-matter coupling (via molecular density/oscillator strength) produces a non-monotonic response: absorption rises to an optimum and then decreases as the film becomes increasingly radiatively bright and reflective. Finite-difference time-domain simulations and analytical transfer-matrix calculations agree quantitatively and yield compact ridge conditions for the optimum. We interpret the trends using a scattering/port picture: the isolated film is a symmetric two-port system (reflection and transmission), which bounds single-sided resonant absorption to 50% in the ultrathin limit (reflecting transition saturation), whereas adding a mirror suppresses transmission and converts the structure into an effectively one-port absorber. In the mirror-backed geometry, interference can cancel reflection and unity absorption is obtained at critical coupling, when radiative leakage is balanced by intrinsic molecular loss. These results clarify fundamental limits and design rules for collective absorption in dense molecular layers near dielectric or metallic boundaries.

We present the first experimental demonstration of a high-bandwidth thin-film lithium tantalate (TFLT) electro-optic modulator operating at 1 μm, with a Vπ of 2.4 V, and less than 2 dB electro-optic roll-off up to 50 GHz and stable DC bias operation.

Superconducting qubit fabrication innovations continue to be explored to achieve higher performance. Despite improvements to base layer fabrication and processing, resist-based Josephson junction (JJ) schemes have largely remained unchanged. The polymer mask during deposition causes chemical contamination and limits in situ and ex situ surface preparation, junction materials, and scalability. Here, we demonstrate a resist-free approach to junction fabrication based on etched silicon trenches that is CMOS compatible and easily integrated into existing innovations in qubit base layer fabrication and chemical processing. We fabricate Al-AlOx-Al JJs and qubits using this method, measuring median energy relaxation times up to 184 microseconds. We find minimal contamination at the substrate-metal interface and fluctuations of energy relaxation on a 35 hour timescale that are narrow and normally distributed. The method widens the process window for substrate preparation and new materials platforms.

Understanding how molecular structure gives rise to odor perception remains a long-standing challenge, with ongoing debate over whether olfaction is primarily governed by molecular shape, vibrational properties, or their interplay at the level of olfactory receptors. Here, we ask whether different odors rely on common molecular determinants or instead emerge from distinct physicochemical regimes. Using interpretable machine-learning models trained on molecular descriptors derived from first-principles calculations that span electronic, vibrational, and structural properties, we analyze feature contributions for odor categories and their associated receptors. We find that no single descriptor class universally dominates odor prediction; instead, different odors exhibit strongly odor-specific patterns of feature importance, with substantial variability across physicochemical domains. This heterogeneity is consistent across different models, suggesting that a universal encoding scheme does not capture odor perception but reflects receptor- and odor-dependent structure-odor relationships. Our results provide statistical constraints on competing olfactory theories and offer a data-driven framework for organizing odor space.

Infrared thermography (IRT) is a widely used non-destructive testing technique for detecting structural features such as subsurface defects. However, most IRT post-processing methods generate image sequences in which defect visibility varies strongly across time, frequency, or coefficient/index domains, making the identification of defect-representative images a critical challenge. Conventional evaluation metrics, such as the signal-to-noise ratio (SNR) or the Tanimoto criterion, often require prior knowledge of defect locations or defect-free reference regions, limiting their suitability for automated and unsupervised analysis. In this work, a data-driven methodology is proposed to identify images within IRT datasets that are most likely to contain and represent structural features, particularly anomalies and defects, without requiring prior spatial information. The approach is based on three complementary metrics: the Homogeneity Index of Mixture (HI), which quantifies statistical heterogeneity via deviations of local intensity distributions from a global reference distribution; a Representative Elementary Area (REA), derived from a Minkowski-functional adaptation of the Representative Elementary Volume concept to two-dimensional images; and a geometrical-topological Total Variation Energy (TVE) index, also based on two-dimensional Minkowski functionals, designed to improve sensitivity to localized anomalies. The framework is validated experimentally using pulse-heated IRT data from a carbon fiber-reinforced polymer (CFRP) plate containing six artificial defects at depths between 0.135 mm and 0.810 mm, and is further supported by one-dimensional N-layer thermal model simulations. The results demonstrate robust and unbiased ranking of image sequences and provide a reliable basis for automated defect-oriented image selection in IRT.

Space, time, and additional dimensions spawn remarkable complexity in optics. We encourage pairing black-box simulation and design tools with a complementary tool: understanding.

Over fifty years ago, Robert May applied random matrix theory to show that as ecological systems grow in size, stability decreases. What emerged from this and the critique that followed was decades of what has been called the complexity-stability debate. However, decades of critique over the assumptions that Robert May applied in carrying out his analysis have not been enough to fully dispel the strength of his conclusion and close the debate. Drawing on a mathematical approach that had not yet been fully developed in the early 70s, it is possible to revisit the argument without the use of random matrix techniques, and provide more detailed understanding of the mechanisms that play a deciding role in stability of ecological systems, countering the broad conclusion that led to the complexity-stability debate.

Goal-directed molecular generation requires satisfying heterogeneous constraints such as protein--ligand compatibility and multi-objective drug-like properties, yet existing methods often optimize these constraints in isolation, failing to reconcile conflicting objectives (e.g., affinity vs. safety), and struggle to navigate the non-differentiable chemical space without compromising structural validity. To address these challenges, we propose CAGenMol, a condition-aware discrete diffusion framework over molecular sequences that formulates molecular design as conditional denoising guided by heterogeneous structural and property signals. By coupling discrete diffusion with reinforcement learning, the model aligns the generation trajectory with non-differentiable objectives while preserving chemical validity and diversity. The non-autoregressive nature of diffusion language model further enables iterative refinement of molecular fragments at inference time. Experiments on structure-conditioned, property-conditioned, and dual-conditioned benchmarks demonstrate consistent improvements over state-of-the-art methods in binding affinity, drug-likeness, and success rate, highlighting the effectiveness of our framework.

The integrated information theory of consciousness (IIT) is uniquely ambitious in proposing a mathematical formula, derived from apparently fundamental properties of conscious experience, to describe the quantity and quality of consciousness for any physical system that possesses it. IIT has generated considerable debate, which has engendered some misunderstandings and misrepresentations. Here we address and hope to remedy this. We begin by concisely summarising the essentials of IIT. Given IIT is supposed to apply universally, we do this with reference to an arbitrary patch of matter, as opposed to the usual system of discrete computational units. Then, after briefly summarising IIT's theoretical and empirical achievements, we focus on five points which we consider especially important for driving forward new theory and increasing understanding. First, a high value of the measure $Φ$ is not synonymous with `more consciousness'. We describe how $Φ$ might be replaced with a suite of quantities to obtain a multi-dimensional characterisation of states of consciousness. Second, we describe with nuance the distinct flavour of panpsychism implied by IIT -- whereby space (and time) are tiled with substrates of (proto-) consciousness -- and find this is not problematic for the theory. Third, $Φ$ is not well-defined for real physical systems, and has not been computed on any real physical system. Fourth, so far only proxies for IIT measures have been computed, and not approximations. Fifth, for IIT to fit with current successful theories in fundamental physics, a reformulation in terms of continuous fields would be needed.

The composition of a polyclonal antibody response is hard to measure experimentally but contains vital information about the robustness of immunity. Here, we argue that the statistics of neutralization titers alone can be used to make quantitative predictions about the composition of the response, circumventing challenges arising through sequencing and monoclonal antibody expression. We show that the response against influenza within a cohort can be either driven by a collective phenomenon where many antibodies contribute to neutralization, or dominated by just a few strong binders, leading to a broad distribution of titers across individuals described by a Gumbel distribution from extreme value theory. Comparing titers across cohorts, we find that Gumbel statistics {accurately describe} individuals prior to an immune challenge. We propose an equilibrium binding model that quantitatively captures titer data and illustrates the structure of the polyclonal response. Our approach extends generically to immune responses to other pathogens.

Cooperation is a key driver of human social progress. Studies of the evolution of cooperation typically assume a deterministic outcome for social interactions. But in real-world social interactions, interaction outcomes are often subject to stochastic perturbations arising from open environments. Individuals may show different attitudes towards such uncertainty, some are risk-seeking, while others tend to be risk-averse. Here we investigate how risk preference towards uncertain payoffs affects the evolution of cooperation on social networks, where uncertainty originates from random punishment of defectors initiated by cooperators. We provide an analytical treatment of how the distribution of risk preference among individuals alters the threshold required for cooperation. We find that, at the population level, risk-averse behavior promotes or even rescues cooperation. At the node level, variation in risk preference has a significant impact when it occurs on nodes with high degree centrality. When nodes have the same degree centrality, the nodes with lower betweenness centrality exhibit a stronger effect on strategy evolution. Our analysis reveals how risk preference, together with spatial structure, jointly shapes and potentially reverses the evolutionary dynamics of cooperation.

Humans are inherently social and seek connection with others for survival. Recent studies suggest that acute social isolation leads to craving for social interactions, but the brain mechanisms of social craving and their relationship to brain networks underlying drug and food craving remain incompletely understood. Here we harnessed an existing dataset and tested whether the Neurobiological Craving Signature (NCS)-a recently developed fMRI-based brain-signature of drug and food craving-also predicts social craving. During fMRI, participants rated their craving for images of food, social interactions, and flowers in three different sessions: after 10h of fasting from food, 10h of social isolation, or neither (baseline; order of sessions counterbalanced). The NCS significantly predicted self-reported craving for food and social cues but not flower cues. Further, NCS responses to food were higher after fasting compared to baseline, and higher for social cues after social isolation compared to baseline, demonstrating its responsiveness to both food and social deprivation. These findings resonate with recent work showing shared brainstem circuits for hunger and social isolation, and indicate shared whole-brain circuits for social, food, and drug craving. They open new avenues for testing the NCS across different primary rewards, for assessing the consequences of their deprivation, and for examining how social deprivation-such as loneliness and isolation-interacts with overeating and drug use.

Forecasting neural activity in response to naturalistic stimuli remains a key challenge for understanding brain dynamics and enabling downstream neurotechnological applications. Here, we introduce a generative forecasting framework for modeling neural dynamics based on autoregressive flow matching (AFM). Building on recent advances in transport-based generative modeling, our approach probabilistically predicts neural responses at scale from multimodal sensory input. Specifically, we learn the conditional distribution of future neural activity given past neural dynamics and concurrent sensory input, explicitly modeling neural activity as a temporally evolving process in which future states depend on recent neural history. We evaluate our framework on the Algonauts project 2025 challenge functional magnetic resonance imaging dataset using subject-specific models. AFM significantly outperforms both a non-autoregressive flow-matching baseline and the official challenge general linear model baseline in predicting short-term parcel-wise blood oxygenation level-dependent (BOLD) activity, demonstrating improved generalization and widespread cortical prediction performance. Ablation analyses show that access to past BOLD dynamics is a dominant driver of performance, while autoregressive factorization yields consistent, modest gains under short-horizon, context-rich conditions. Together, these findings position autoregressive flow-based generative modeling as an effective approach for short-term probabilistic forecasting of neural dynamics with promising applications in closed-loop neurotechnology.

Bipartition cover probabilities quantify whether a collection of gene trees contains every bipartition of the underlying species tree, a condition that underlies finite-sample guarantees for summary methods such as ASTRAL. We study this problem under the multispecies coalescent (MSC) model and derive topology-free upper bounds on the number of loci required to obtain a bipartition cover with prescribed confidence, improving upon the existing bounds of Uricchio et al. (2016). Practically, our bounds remain below biologically realistic numbers of loci across a substantially broader range of parameter settings, expanding their usefulness for empirical datasets. Theoretically, our analysis sharpens our understanding of coalescence under the MSC model and develops new asymptotics for these bounds and absorption times under Kingman's coalescent in the natural short branch regime. We further compare our new bounds with existing work using simulations under a variety of different species-tree topologies.

Identifying dynamically influential nodes in biological networks is a central problem in systems biology, particularly for prioritizing intervention targets in gene regulatory networks. In this paper, we propose a Shapley-value-based framework for assessing the importance of nodes in a Boolean network with respect to a given target node. The framework comprises two complementary measures: the Knock-out and the Knock-in Shapley values. Moreover, we present a propagation-based method that enables their efficient computation. By exploiting the logical structure of the network, the method avoids exhaustive simulations. The approach is exact for acyclic networks and provides good approximations for cyclic networks. Evaluation on benchmark models from the Cell Collective database shows that the propagation method accurately recovers node importance rankings while achieving substantial speed-ups.

Labelling tissue components in histology whole slide images (WSIs) is prohibitively labour-intensive: a single slide may contain tens of thousands of structures--cells, nuclei, and other morphologically distinct objects--each requiring manual boundary delineation and classification. We present a cloudnative, end-to-end pipeline that automates this process through a cluster-first paradigm. Our system tiles WSIs, filters out tiles deemed unlikely to contain valuable information, segments tissue components with Cellpose-SAM (including cells, nuclei, and other morphologically similar structures), extracts neural embeddings via a pretrained ResNet-50, reduces dimensionality with UMAP, and groups morphologically similar objects using DBSCAN clustering. Under this paradigm, a human annotator labels representative clusters rather than individual objects, reducing annotation effort by orders of magnitude. We evaluate the pipeline on 3,696 tissue components across 13 diverse tissue types from three species (human, rat, rabbit), measuring how well unsupervised clusters align with independent human labels via per-tile Hungarian-algorithm matching. Our system achieves a weighted cluster-label alignment accuracy of 96.8%, with 7 of 13 tissue types reaching perfect agreement. The pipeline, a companion labelling web application, and all evaluation code are released as open-source software.

Recent advances in single-cell technologies have advanced our understanding of gene regulation and cellular heterogeneity at single-cell resolution. Single-cell data contain both gene expression levels and the proportion of expressing cells, which makes them structurally different from bulk data. Currently, methodological work on causal mediation analysis for single-cell data remains limited and often requires specific distributional assumptions. To address this challenge, we present QuasiMed, a mediation framework specialized for single-cell data. Our proposed method comprises three steps, including (i) screening mediator candidates through penalized regression and marginal models (similar to sure independence screening), (ii) estimation of indirect effects through the average expression and the proportion of expressing cells, (iii) and hypothesis testing with multiplicity control. The key benefit of QuasiMed is that it specifies only the mean functions of the mediation models through a quasi-regression framework, thereby relaxing strict distributional assumptions. The method performance was evaluated through the real-data-inspired simulations, and demonstrated high power, false discovery rate control, and computational efficiency. Lastly, we applied QuasiMed to ROSMAP single-cell data to illustrate its potential to identify mediating causal pathways. R package is freely available on GitHub repository at https://github.com/sjahnn/QuasiMed.

Net primary productivity (NPP) forms the basis of biological carbon pump, but its estimates in high-latitude regions remain highly uncertain despite its disproportional importance for the global carbon sink. Optical satellites are limited by cloud cover, low irradiance, and shallow light penetration, with uncertainties further exacerbated by the lack of in situ validations and regional model tuning for NPP measurements. This study compared two satellite-based models, a global (VGPM) and a regionally tuned (BIO) NPP model, with a time series of in situ NPP. Using a high-frequency, depth-resolved moored profiler in the subpolar Northwest Atlantic (56°N) in 2016, in situ NPP was estimated by daily bio-optical profiles and prior measurement of photosynthesis-irradiance (P-I) parameters. Our findings indicated that satellite-derived estimates of depth-integrated NPP were overestimated by a factor of 2.5 to 4. However, the reasons for the discrepancies varied between the VGPM and BIO model. VGPM used global photosynthetic parameters with a simplified depth assumption, leading to an unrealistic vertical structure for depth-integrated NPP, despite its surface values were lower than in situ estimates. A major phytoplankton bloom in June-July was missed by VGPM, likely due to the use of non-regionally calibrated OCI Chl-a, which led to an underestimation of biomass. In contrast, the BIO model used regionally tuned POLY4 Chl-a products, and the differences in the assignment of P-I parameters accounted for the remaining discrepancies. This study showed the possibility to reach good agreement between satellite and in situ NPPs if the challenge of P-I assignment can be overcome. We recommend further studies to investigate discrepancies of NPP estimates in high-latitude regions, focusing on data sources and model choices, as well as improving regional model calibration to enhance NPP accuracy.

Many essential biological functions, such as breathing and locomotion, rely on the coordination of robust and adaptable rhythmic patterns, governed by specific network architectures known as connectomes. Rhythmic adaptation is often linked to slow structural modifications of the connectome through synaptic plasticity, but such mechanisms are too slow to support rapid, localized rhythmic transitions. Here, we propose a neuromodulation-based control architecture for dynamically reconfiguring rhythmic activity in networks with fixed connectivity. The key control challenge is to achieve reliable rhythm switching despite neuronal degeneracy, a form of structured variability where widely different parameter combinations produce similar functional output. Using equivariant bifurcation theory, we derive necessary symmetry conditions on the neuromodulatory projection topology for the existence of target gaits. We then show that an adaptive neuromodulation controller, operating in a low-dimensional feedback gain space, robustly enforces gait transitions in conductance-based neuron models despite large parametric variability. The framework is validated in simulation on a quadrupedal gait control problem, demonstrating reliable gallop-to-trot transitions across 200 degenerate networks with up to fivefold conductance variability.

This paper investigates the impact of practical features of the emerging antenna array technology of Dynamic Metasurface Antennas (DMAs) when used for wideband sensing. By adopting a realistic DMA response model capturing frequency selective magnetic polarizability, finite resonant frequency tuning, and waveguide phase and leakage effects, we first present a compact observation model for user localization and multiple scattering points sensing through DMA-based analog combining of Orthogonal Frequency Division Multiplexing (OFDM) pilots transmitted in the uplink direction. Building on this model, we derive the Fisher Information Matrix (FIM), the equivalent FIM, and the corresponding Cramer-Rao Bounds (CRBs) for the relevant spatitemporal parameters estimation. The analysis reveals that frequency selectivity reduces the effective information bandwidth and distorts the DMA-based reception manifold, while waveguide attenuation decreases both the coherent combining gain and the effective aperture, thereby degrading estimation accuracy. Numerical results validate the analysis and confirm the resulting inflation in the delay, angle, and position error bounds.

Many wireless systems divide the baseband processing between two locations, interconnected by a fronthaul. This paper examines the impact of fronthaul quantization on multiple-input multiple-output (MIMO) systems. Starting from a Bussgang-based analysis of quantized single-input single-output (SISO) channels, we extend the framework to MIMO and derive a capacity lower bound under fronthaul quantization, where the receive combining is performed before the quantization. To maximize the sum rate, we propose a joint bit and power allocation (JBP-Alloc) scheme that efficiently distributes fronthaul bits and transmit power across active data streams. Asymptotic analysis shows that uniform bit allocation becomes optimal at high SNR. Numerical results confirm that JBP-Alloc outperforms uniform allocation and quantization-unaware water-filling, and achieves the same performance as Greedy bit allocation but with substantially lower computational complexity.

Integrated sensing and communication (ISAC) systems rely on communication waveforms to perform sensing tasks, thus making their sensing performance strongly dependent on the level of communication symbol knowledge available to the sensing receivers. However, the existing literature fails to capture this dependency, often relying on full symbol knowledge assumptions. In this paper, we present a Cramer Rao bound (CRB) analysis of a bistatic ISAC network with heterogeneous uplink and downlink illumination and structured clutter. We consider different symbol knowledge regimes by modeling unknown communication symbols as nuisance parameters. Assuming a temporal evolution of the communication channel, we derive a correlation aware channel estimator and an expression for the UEs uplink spectral efficiency (SE). Numerical results show the CRB degradation induced by clutter and symbol uncertainty and how this can affect resource allocation policies. We also show the performance gain of our channel estimator relative to conventional block fading architectures.

The rapid growth of the low-altitude economy (LAE) is making aerial systems an important part of future digital infrastructure. Although major advances have been achieved in unmanned aerial vehicle (UAV) platforms, communications, and autonomous control, environmental perception remains a key bottleneck to reliable and scalable LAE operations. Existing sensing modalities, such as optical, LiDAR, and millimeter-wave radar, are limited by visibility, sensing range, and environmental conditions, resulting in fragmented situational awareness. This article argues that addressing these limitations requires a shift from platform-centric sensing to a shared, environment-aware sensing infrastructure. In this context, synthetic aperture radar (SAR) offers a distinct advantage by enabling all-weather, wide-area perception. We show that SAR can support UAV operations through global environmental awareness, enhance task-level sensing, and enable cooperative sensing across satellites, high-altitude platforms, UAVs, and ground systems. Building on this perspective, we present a system-level view of SAR-enabled LAE, highlighting key transformations from fragmented to infrastructure-centric sensing, from reactive to predictive operation, and from device-centric to environment-aware networking. We further discuss enabling architectures, including multi-platform sensing hierarchies, integration with integrated sensing and communication (ISAC), and the role of artificial intelligence and digital twins, along with the key challenges toward real-world deployment. By positioning SAR as a shared sensing foundation rather than a standalone modality, this article provides new insights into the design of scalable, reliable, and intelligent LAE systems.

Spectrum sharing and dynamic spectrum reuse are becoming increasingly critical in modern wireless networks to address spectrum scarcity. However, these techniques inevitably increase Cross-Technology Interference (CTI). In this context, the Open Radio Access Network (O-RAN), as a modern and disaggregated network architecture, necessitates accurate, low-latency, and computationally efficient CTI classification and mitigation to support real-time control and maintain Quality of Service (QoS). Unfortunately, existing solutions predominantly rely on high-complexity, monolithic deep learning-based solutions that, while achieving high classification accuracy, incur significant latency and computational overhead This paper exploits the O-RAN functional split to leverage multi-domain raw signal representations (time, frequency, and Channel State Information (CSI)) directly from the same data stream. Each domain is processed locally, naturally interleaving CTI within the distributed, disaggregated O-RAN architecture. This distributed strategy enables a cost-aware, multi-domain fusion architecture that balances classification accuracy with computational overhead and latency. Our proposed multi-domain distributed architecture achieves a 400 $μs$ inference latency on standard CPUs. Compared to a state-of-the-art monolithic frequency-domain classifier, this represents an average 9x reduction in latency and an 11-fold decrease in computational cost, while sacrificing only 4% in classification performance and maintaining >90% accuracy in high-interference conditions.

A mutual coupling-aware beamforming design for continuous aperture array (CAPA)-aided multi-user systems is investigated. First, a transmit coupling kernel is characterized to explicitly capture the mutual coupling effects inherent in CAPAs, based on which a mutual coupling-aware sum-rate maximization functional optimization problem is formulated. To address this problem, a kernel approximation (KA)-based weighted minimum mean-squared error (WMMSE) algorithm is developed. The optimal beamforming condition is derived within the WMMSE framework using the calculus of variations, while KA is employed to obtain a closed-form beamforming solution via wavenumber-domain Fourier transforms and Gauss-Legendre quadrature. Furthermore, the proposed framework is extended to CAPA-to-CAPA multiple-input multiple-output (MIMO) systems. Finally, numerical results demonstrate that: 1) the proposed algorithm achieves improved performance compared to benchmark schemes; 2) the modeled coupling effects are physically rational, where the performance of spatially discrete arrays converges to that of CAPAs; and 3) CAPA-to-CAPA MIMO systems can achieve higher degrees of freedom when the transceivers are placed in close proximity.

Everything that exists has a natural frequency; this material characteristic is something that must be known and fully understood. If we fail to predict, measure, and address potential natural frequency concerns, it could significantly reduce the life span of our equipment or cause it to fail immediately when put into service. There are a few methodologies used to study natural frequencies, one being computer simulations and the other being physical tests done on the equipment. In this paper, we will focus on testing natural frequencies and discuss how we measure our excitation, our form of excitation, the type of data we are able to export, as well as what we are able to do with that data. These principles can be applied to any type of machinery or object where vibration could be of concern. For our purposes, we will primarily focus on rotating machinery, such as generators, gearboxes, and motors.

Channel gain maps (CGMs) enable propagation-aware services in edge-intelligent wireless communication networks, while diffusion-based CGM construction is memory intensive for on-device training or adaptation. This letter proposes InvDiff-CGM, an invertible diffusion framework that constructs CGMs from sparse measurements and environmental priors. By adopting invertible architectures in both the diffusion process and the U-Net noise estimator, InvDiff-CGM achieves near-constant training memory consumption. A prior-informed multi-scale injector further integrates environmental priors with sparse measurements to improve physical consistency and detail preservation. Experiments on RadioMap3DSeer show about an 85\% reduction in peak training memory and a PSNR of 38.02~dB, outperforming representative recent baselines. This validates the practicality of InvDiff-CGM for high-fidelity CGM construction under edge resource constraints.

RF sensing exploits phase-sensitive measurements of stray electromagnetic (EM) fields from wireless devices across various frequency bands to detect EM blockage and to reconstruct and map the surrounding environment in 2D/3D. Although blockage effects caused by objects or human motion are well-studied in ISM bands and frequencies up to 60~GHz, there is a significant lack of research for frequencies above 100~GHz. The paper proposes a unified signal processing framework for RF sensing in the sub-THz D-band (105--175~GHz), explicitly integrating EM blockage and scattering as a single process through the birth-death dynamics of multipath components (MPCs). The framework extracts, associates, and classifies MPCs from angle-delay measurements using statistically grounded detection and classification, enabling human-scale sensing from a single radio link. The modeling and classification of MPCs, along with large-scale EM parameters, are demonstrated through an indoor measurement campaign using multiple test targets. Experimental results show that newly formed, attenuated, and suppressed MPCs can be reliably identified with millimeter-scale delay resolution. Static object localization achieves average positioning errors of $8-20$~cm depending on range and material, while passive human localization yields errors of 12-17cm at 0.5m and 26-30cm at 2m, respectively. The proposed framework demonstrates that accurate sensing and localization are feasible at sub-THz frequencies using a single link.

Retrieving procedure-oriented evidence from materials science papers is difficult because key synthesis details are often scattered across long, context-heavy documents and are not well captured by paragraph-only dense retrieval. We present RECIPER, a dual-view retrieval pipeline that indexes both paragraph-level context and compact large language model-extracted procedural summaries, then combines the two candidate streams with lightweight lexical reranking. Across four dense retrieval backbones, RECIPER consistently improves early-rank retrieval over paragraph-only dense retrieval, achieving average gains of +3.73 in Recall@1, +2.85 in nDCG@10, and +3.13 in MRR. With BGE-large-en-v1.5, it reaches 86.82%, 97.07%, and 97.85% on Recall@1, Recall@5, and Recall@10, respectively. We further observe improved downstream question answering under automatic metrics, suggesting that procedural summaries can serve as a useful complementary retrieval signal for procedure-oriented materials question answering. Code and data are available at https://github.com/ReaganWu/RECIPER.

Multi-path sensing, which aims to extract the geometric attributes of multiple propagation paths, is expected to be a key functionality of 6G. A movable antenna (MA) can enable this functionality by creating a synthetic aperture through sequential mechanical motion. However, existing MA-based sensing methods typically rely on exhaustive scanning over the entire movable plate, resulting in significant control overhead and sensing latency, which limits their practicality for agile sensing. To address this challenge, this paper develops a prior-guided agile multi-path sensing framework that leverages weak prior angle-of-arrival (AoA) statistics as side information. The proposed framework comprises two steps. First, the movable plate's three-dimensional orientation is optimized only once to maximize path visibility while preserving path discriminability, both induced from Fisher information analysis. Second, only two predetermined linear MA scans are made on the tilted plate to estimate the elevation and azimuth AoAs from the resulting sequence of received signals. By incorporating the prior AoA statistics, a maximum a posteriori (MAP)-based AoA estimation algorithm is developed. With only one orientation control and two linear scans, the proposed framework enables agile multi-path sensing with significantly reduced control overhead and latency, while achieving AoA estimation accuracy approaching that of the single-path benchmark.

This paper experimentally investigates geometry-based multi-antenna RF wireless power transfer (WPT) using a large-scale distributed indoor transmit array measuring 8 m by 4 m. Geometry-based beamforming uses known transmitter and receiver positions to perform phase-only precoding, avoiding the need for explicit channel estimation or feedback. The experiments use a ceiling-mounted array of 41 phase-synchronized transmit antennas operating at 920 MHz. Geometry-based beamforming is compared with channel state information (CSI)-based beamforming. The spatial power delivery is evaluated through two-dimensional scans over an area of 1.25 m by 1.25 m. The harvested DC power is measured using an RF-to-DC energy profiler. Under line-of-sight (LoS) conditions, geometry-based beamforming achieves a power gain of 18.75 dB, which is within 0.82 dB of CSI-based beamforming. In obstructed LoS scenarios with reflections, the gain decreases to 16.7 dB, while CSI-based beamforming achieves 20.53 dB, resulting in a performance gap of 3.83 dB. These results quantify the trade-off between reduced system overhead and robustness to multipath propagation in geometry-driven WPT, and represent an initial step toward geometry-based wireless power transfer enabled by digital twins.

Large-scale three-dimensional (3D) scene reconstruction in low-altitude intelligent networks (LAIN) demands highly efficient wireless image transmission. However, existing schemes struggle to balance severe pilot overhead with the transmission accuracy required to maintain reconstruction fidelity. To strike a balance between efficiency and reliability, this paper proposes a novel deep learning-based end-to-end (E2E) transceiver design that integrates 3D Gaussian Splatting (3DGS) directly into the training process. By jointly optimizing the communication modules via the combined 3DGS rendering loss, our approach explicitly improves scene recovery quality. Furthermore, this task-driven framework enables the use of a sparse pilot scheme, significantly reducing transmission overhead while maintaining robust image recovery under low-altitude channel conditions. Extensive experiments on real-world aerial image datasets demonstrate that the proposed E2E design significantly outperforms existing baselines, delivering superior transmission performance and accurate 3D scene reconstructions.

Task-oriented semantic communication emerges as a crucial paradigm for next-generation wireless networks, aiming to efficiently transmit task-relevant information while reducing interference and redundancy across multiple users. Existing information bottleneck (IB)-based frameworks predominantly focus on single-user scenarios, neglecting cross-user semantic interference in distributed semantic communications. To overcome this limitation, we propose a task-oriented orthogonalised information bottleneck (TOIB) approach, explicitly designed for distributed semantic communication systems. By introducing task-conditioned latent variables, TOIB adaptively balances semantic sufficiency, semantic compression, and inter-user semantic orthogonality. Extensive simulations conducted on classification tasks demonstrate that TOIB consistently achieves superior classification accuracy across various signal-to-noise ratio (SNR) regimes compared to traditional IB and deep joint source-channel coding (JSCC) methods. Specifically, the proposed method significantly enhances robustness under harsh low-SNR conditions and effectively suppresses cross-user semantic interference, as validated by cross-decoding accuracy metrics.

Extremely large-scale multiple-input multiple-output (XL-MIMO) is a key enabler for sixth-generation (6G) communications. However, near-field channel estimation is particularly challenging due to spherical-wave propagation and spatial non-stationarity. To tackle this challenge, we propose a structured sparse Bayesian learning framework with adaptive dictionary updating for near-field non-stationary channel estimation. Specifically, the proposed method iteratively updates the distance parameters within an adaptive dictionary, thereby enhancing the representation capability without increasing the dictionary size. Moreover, we develop a hierarchical prior model that jointly captures polar-domain sparsity and structured dependency, enabling efficient Bayesian inference. Simulation results demonstrate that the proposed approach outperforms existing polar-domain dictionary-based methods while achieving low dictionary overhead.

The dynamic competition against intelligent jammer systems presents a significant challenge to modern radar. Traditional active anti-jamming strategy learning methods often suffer from low sample efficiency and fail to fully exploit the structures of the adversary jammer. To reveal the inherent structure, this paper adopts an Online Convex Optimization (OCO) framework to capture the competition between a frequency agile radar and a digital radio frequency memory (DRFM)-based intelligent jammer. Recognizing that conventional OCO algorithms also suffer from suboptimal sample efficiency, two refined algorithms are developed that incorporate unbiased gradient estimators specifically tailored to the unique characteristics of DRFM-based jammers. Our theoretical analysis of the regret bound indicates significant improvements in long-term performance compared to standard OCO. The simulation results consistently show that our algorithms outperform traditional OCO and reinforcement learning baselines, achieving faster convergence and better anti-jamming performance.

Traditional Active Noise Control (ANC) systems are mostly based on FxLMS algorithms, but such algorithms rely on linear assumptions and are often limited in handling broadband non-stationary noise or nonlinear acoustic paths. Not only that, the traditional method is used to eliminating all signals together, and noise reduction often accidentally damages the voice signal and affects normal communication. To tackle these issues, this study proposes a speech preserving deep learning ANC system, which aims to achieve stable noise reduction while effectively retaining speech in a complex acoustic environment. This study builds an end-to-end control architecture, the core of which adopts a Convolutional Recurrent Network (CRN). The structure uses the long short-term memory (LSTM) network to capture the time-related characteristics of acoustic signals. Combined with complex spectrum mapping (CSM) technology, the nonlinear distortion problem is effectively solved. In order to retain useful voice while removing noise, this study also designs a special voice retention loss function. This design guidance model selectively retains the target voice while suppressing environmental noise by identifying the characteristics of the spectrum structure. In addition, in order to verify whether the system is effective in real scenes, we use the Image Source Method (ISM) to build a high-fidelity acoustic simulation environment, which also simulates the real reverberation effect. Experimental results demonstrate that the proposed Deep ANC system achieves significantly better noise reduction than the traditional FxLMS algorithm, especially for non-stationary noises like crowd babble. Meanwhile, PESQ and STOI based evaluations confirm that the system preserves both the naturalness and intelligibility of the target speech.

Semantic communication has been increasingly integrated into edge computing systems for reconstruction tasks, owing to its advantages in source compression, robustness to channel noise, and task execution efficiency. However, the black-box nature of neural-network (NN)-based semantic codecs, together with the noisy transmission of semantic features, makes it difficult to allocate transmission resources and guarantee reconstruction quality for multiple users. In this paper, we propose a reliable online resource allocation framework for a semantic-driven multi-user edge computing system, where multiple users encode source information into semantic features and offload reconstruction to an edge server. We formulate a multi-user resource optimization problem whose objective jointly accounts for system-wide reconstruction performance and transmission latency, under constraints that guarantee each user's minimum reconstruction quality. To solve this problem, we develop a Bayesian optimization (BO)-based online algorithm that enables flexible control of the user-side semantic compression ratio (CR) and allocation of transmission rates. The edge server jointly determines each user's CR and transmission rate by exploiting Gaussian-process (GP) models that capture the relationship between reconstruction performance, signal-to-noise ratio (SNR), and CR, and by employing an acquisition function to select CRs that satisfy the performance quality constraints while maximizing the objective. Simulation results on high-resolution video-frame reconstruction datasets demonstrate that the proposed method selects near-optimal CRs via the GP surrogate and acquisition function, achieving a 98.03% constraint-satisfaction rate and reducing transmission latency by more than 45% compared with fixed-CR schemes.

Meta-backscatter system that utilizes meta-material sensors is a promising enabler for future environmental sensing, offering distinct advantages such as low cost, zero-power consumption, and robustness. Specifically, the electromagnetic response of the sensor, typically characterized by a frequency-selective absorption profile, is affected by the environmental conditions, allowing the estimation of these conditions from the reflected signal. However, it remains unclear what estimation accuracy can be achieved fundamentally. Motivated by this gap, we quantify this accuracy limit using the Bayesian Cramér-Rao bound (BCRB), which provides a lower bound on the mean-squared error for the environmental condition. Establishing this limit is challenging because the electromagnetic response of the sensor is distorted by the channel fading, while the channel estimation is infeasible since the sensors cannot be configured to predefined states to generate training data. To address this challenge, we consider the joint BCRB of the channel coefficient and the environmental condition in a multicarrier framework. The BCRB of the environmental condition is then obtained by selecting the corresponding element from the joint BCRB. An analysis of the derived BCRB reveals the impact of the absorption peak shape and the number of subcarriers. The derivation and analysis of the BCRB are verified through simulations.

Robust radio signal recognition is fundamental to spectrum management, electromagnetic space security, and intelligent wireless applications, yet existing deep-learning methods rely heavily on large labeled datasets and struggle to capture the multi-domain characteristics inherent in real-world signals. To address these limitations, we propose an unsupervised equivalent contrastive learning method that leverages four information-lossless equivalent transformations, spanning the time, instantaneous, frequency, and time-frequency domains, to construct multi-view and semantically consistent representations of each signal. An equivalent contrastive learning strategy then aligns these complementary views to learn discriminative and transferable embeddings without requiring labeled data. Once pre-training is completed, the resulting model can be directly fine-tuned on downstream tasks using only raw signal samples, without reapplying any equivalent transformations, which reduces computational overhead and simplifies deployment. Extensive experiments on four public datasets demonstrate that the proposed method consistently outperforms state-of-the-art contrastive baselines under linear evaluation, few-shot semi-supervised learning, and cross-domain transfer settings. Notably, the learned representations yield substantial gains in few-shot regimes and challenging channel conditions, confirming the effectiveness of multi-domain equivalent modeling in enhancing robustness and generalization. This work establishes a principled pathway for exploiting massive unlabeled radio data and provides a foundation for future self-supervised learning frameworks in wireless systems.

This paper investigates the joint optimization of beamforming and antenna positions in fluid antenna system (FAS)-aided anti-jamming communications. We consider a multi-user multiple-input multiple-output downlink scenario where multiple malicious jammers exist and the jammer channel state information is imperfect. The goal is to maximize the worst-case sum-rate under quality-of-service and transmit power constraints. To achieve this, we develop two distinct optimization frameworks for continuous and discrete antenna position designs, respectively. For continuous design, we propose an alternating optimization (AO) framework that integrates successive convex approximation and majorization minimization (MM) to handle the highly non-convex problem. For discrete design, based on the minimum mean squared error criterion and MM, we reformulate the problem as a sparse recovery task and propose a low-complexity block coordinate descent and simultaneous orthogonal matching pursuit, which enables joint design rather than AO. Through systematic comparison, we uncover a practical phenomenon: the discrete joint design yields superior sum-rate performance compared to the AO-based continuous counterpart under identical conditions. This superiority stems from the sparse recovery formulation which effectively circumvents the severe local optima. Our findings challenge the conventional view that continuous optimization is inherently superior, and reveal that discretization combined with sparse recovery can offer a more effective paradigm for exploiting spatial degrees-of-freedom in FAS-aided anti-jamming communications.

With the rapid growth of Multi-access Edge Computing (MEC), secure and efficient computation offloading from user equipment (UEs) to edge access points (APs) is critical. However, DISCO intelligent reflective surface-based fully-passive jammers (DIRS-based FPJs) use random time-varying phase shifts to launch DISCO jamming attacks, disrupting offloading performance. This paper leverages an aerial intelligent reflective surface (AIRS) to enable secure computation offloading against DISCO jamming by jointly optimizing offloading ratios, AIRS phase shifts, and deployment. A two-timescale (2Ts) framework is proposed to address the optimization challenge caused by the distinct update frequencies of different strategies. Specifically, AIRS deployment is adjusted on a long timescale to boost antijamming capability due to the impracticality of frequent physical adjustment, while offloading ratios and phase shifts are optimized on a short timescale to adapt to DIRS-jammed dynamic channel conditions. We propose a dual-agent deep reinforcement learning (DRL)-based AIRS deployment-aided secure computation offloading (DDADSO) scheme to maximize the secure offloading utility under DISCO jamming. Simulation results verify that the proposed DDADSO scheme outperforms benchmark schemes, demonstrating the effectiveness of AIRS deployment in improving offloading performance against DISCO jamming attacks.

Integrated sensing and communication (ISAC) requires spatial architectures that can flexibly balance data transmission and environment sensing. Segmented pinching antenna-assisted ISAC provides such flexibility by allowing different waveguide segments to be dynamically configured for transmission and reception. However, its design involves the joint optimization of antenna deployment, segment partitioning, and beamforming under coupled communication and sensing constraints, which becomes particularly challenging when the numbers of communication users and sensing targets vary across scenarios. To endow the system with stronger adaptability to changing user and target configurations, we propose a general learning framework for segmented pinching antenna-assisted ISAC systems. Specifically, a channel state information (CSI)-induced self-graph is constructed to produce permutation-invariant representations of user-target interactions, and the resulting features are processed by a large language model (LLM) backbone with two task-specific heads for jointly predicting antenna deployment, segment partitioning, and ISAC beamforming. In addition, a user count transfer mechanism is developed to examine whether the learned deployment policy is site-specific and reusable under changed user configurations. Simulation results show that the proposed framework achieves higher communication rates while maintaining reliable sensing accuracy. Moreover, the learned deployment policy remains highly stable when transferring to other user counts, which reduces the training cost from full model retraining to beamforming head adaption.

Physical Layer Authentication (PLA) exploits the spatial uniqueness of wireless channel characteristics in order to authenticate devices without recourse to higher-layer cryptographic protocols, which remain vulnerable to key compromise. This paper reports a comprehensive PLA system constructed on 5G New Radio (NR) Sounding Reference Signals (SRS) extracted from a real OpenAirInterface (OAI) testbed operating in band n78 (3.5 GHz) with 40 MHz bandwidth and 30 kHz subcarrier spacing. The proposed approach extracts a 2,531-dimensional feature vector per SRS probe, combining per-subcarrier channel state information (1,248 amplitude and 1,247 differential-phase coefficients), power delay profile taps, delay spread, Doppler statistics, and nonlinear dynamics indicators. A deep one-dimensional Residual Network (1D-ResNet) augmented with Squeeze-and-Excitation (SE) attention blocks is employed to classify each probe as either legitimate or spoofed. Evaluation is conducted on 20,317 over-the-air SRS probes acquired across four measurement sessions using a USRP B210 software-defined radio as the legitimate device and a commercial mobile handset as the attacker. Under a strict chronological train/validation/test split that eliminates temporal leakage, an Equal Error Rate (EER) of 3.92% is attained, with AUC = 0.962 on the held-out test set, and an authentication latency of less than 0.1 ms per probe, which is compatible with 5G Ultra-Reliable Low-Latency Communications (URLLC) requirements.

In the limited feedback downlink multiple-input single-output (MISO) non-orthogonal multiple access (NOMA) system, both the effective channel gain and the channel direction need to be quantized. The quantization error affects the feasible region of NOMA and the rate loss compared with the full channel state information (CSI) case. In this letter, we analyze this effect and obtain upper bound for the rate loss. The numerical results show that the sum rate of the limited feedback MISO-NOMA system approaches that of the full CSI as the number of feedback bits increases.

In this work, we propose an interpretable, robust, and lightweight machine learning method for automatic modulation classification (AMC) under dynamic and noisy channel conditions. It is called green automatic modulation classification (GAMC) and targets edge artificial intelligence (AI) with low computational complexity and a small model size. GAMC operates in four stages. First, raw received I/Q signals are transformed into multi-domain representations, including constellation diagrams and spatio-temporal graphs. Second, we extract a comprehensive set of statistical and topological features from time-series signals, constellation diagrams, and graphs. Third, a supervised feature learning process leverages label guidance to project high-dimensional features into robust, discriminative low-dimensional ones. Finally, a context-aware Signal-to-Noise Ratio (SNR) soft routing mechanism ensembles predictions from downstream classifiers. Experimental results show that GAMC effectively mitigates domain shifts caused by high noise. It strikes a good balance between accuracy and efficiency, reducing the number of model parameters by $50\%$, operating at $3\%$ to $42\%$ of the computational cost of lightweight deep learning models, and maintaining higher accuracy in various SNRs.

Orthogonal frequency-division multiplexing (OFDM) is a dominant waveform in modern wireless systems, yet its high peak-to-average power ratio (PAPR) and limited adaptability hinder efficient support for integrated communication and sensing. This paper proposes deep block-unitary precoded OFDM (DBU-OFDM), a structure-preserving learning framework that enables trainable waveform adaptation while preserving the DFT-based signal structure, pilot/null resource protection, and compatibility with low-complexity frequency-domain equalization. The proposed design restricts learning to a block-unitary transformation over data subcarriers and preserves pilot and null resources for structural compatibility. The transform is parameterized by recursive Householder reflections, ensuring strict unitarity as well as differentiable, numerically stable, and complexity-controllable implementation. Results show that DBU-OFDM achieves PAPR tails close to block-pilot DFT-s-OFDM while retaining comb-type pilots, improves communication reliability in frequency-selective fading via frequency-domain diversity, and enhances range and velocity estimation in direct sensing, especially in dimension-limited settings. Over-the-air USRP experiments and FPGA prototyping further verify its practical feasibility, demonstrating low error vector magnitude (EVM), clear PAPR reduction in real transmission, and hardware throughput up to 200~MS/s with microsecond-level latency. DBU-OFDM therefore offers a practical intermediate solution between conventional model-based OFDM waveforms and unconstrained neural transceivers for next-generation integrated communication and sensing systems.

Electroencephalography provides a non-invasive and cost-effective approach for analyzing neural patterns associated with alcohol dependence. However, reported classification performance in EEG-based alcoholism studies varies considerably, often due to differences in validation strategies rather than intrinsic model capability. This study presents a validation-aware machine learning framework to assess the impact of evaluation methodology on classification performance. A balanced multi-channel EEG dataset of 300 trials (150 alcoholic, 150 control) was analyzed using a structured feature representation combining statistical descriptors and spectral band interactions. Five classifiers, including support vector machines (linear and radial basis function kernels), random forest, k-nearest neighbors, and AdaBoost, were evaluated under standard and nested cross-validation protocols. Results show that conventional validation with global hyperparameter tuning introduces optimistic bias. In particular, SVM with radial basis function kernel exhibited a performance decrease of approximately 5\% under nested cross-validation, indicating overestimation. Ensemble-based methods showed more stable generalization, with AdaBoost achieving the highest performance, reaching 78.3\% accuracy ($\pm$4.25), an AUC of 0.868, and balanced sensitivity (78.67\%) and specificity (81.33\%). These findings highlight that validation strategy is a primary determinant of perceived model performance. Statistical analysis using McNemar's test further shows that most performance differences between models are not statistically significant, emphasizing careful interpretation of classification results. The proposed framework provides a reproducible and robust basis for evaluating machine learning models in biomedical signal analysis.

Segmented pinching antenna assisted integrated sensing and communication (ISAC) systems enable flexible spatial resource utilization by allowing different waveguide segments to be dynamically configured for transmission and reception. However, the resulting design requires the joint optimization of antenna deployment, segment partitioning, and beamforming under coupled communication and sensing constraints. In this paper, we propose a general learning framework for segmented pinching antenna assisted ISAC systems. Specifically, a channel state information (CSI)-induced self-graph is constructed to capture the scenario-dependent interactions among communication users and sensing targets. Based on the learned graph representation, a large language model (LLM) backbone with low-rank adaptation (LoRA) is employed, followed by two task-specific output heads for antenna deployment and beamforming prediction, respectively. Simulation results show that the proposed framework achieves a favorable tradeoff between communication rate and sensing accuracy

Mutual coupling is a dominant systematic effect in dense reflector arrays, imprinting direction-dependent and frequency-dependent structure on embedded element patterns (EEPs) and currently limiting sensitivity in precision radio measurements. Accurate modelling of these effects requires full-wave simulations of structures that are electrically large at both the array and element levels, making conventional approaches computationally prohibitive. We present a Method-of-Moments (MoM) framework accelerated by a fast direct solver (FDS). The rotational symmetry of reflector dishes is exploited to efficiently compress self-interaction blocks of the impedance matrix. Mutual interactions are treated using a broadband multipole decomposition that remains efficient and accurate for closely spaced elements. We demonstrate the method on arrays of tens of reflectors from the Hydrogen Epoch of Reionization Array (HERA) telescope. To scale to larger arrays, the FDS is used to construct macro-basis functions (MBFs) from a smaller representative array and embed them within a conventional MBF scheme. This allows the first computation of EEPs for the 320-element HERA core on a 128-core workstation.

Near-field extremely large multiple input multiple output (XL-MIMO) breaks the assumptions that make classical super-resolution effective: the receiver acquires only a limited set of compressed pilot observations, while each propagation path is jointly determined by angle and distance under a spherical-wave model. This invalidates the far-field Vandermonde structure exploited by conventional methods, and many existing near-field formulations remain only gridless by discretizing range and angle and thus inheriting mismatch, coherence, and resolution loss. This paper develops a continuous 2D super-resolution framework for hybrid near-field measurements that avoids range and angle gridding. The key idea is to reparameterize distance through inverse range, which reveals a compact spectral structure for the near-field spherical-wave manifold. Building on this observation, we introduce a panelized weighted fitting strategy that converts the range-dependent Fresnel terms into a stable transform-domain representation, resulting in a lifted mode, in which each continuous range-angle pair is embedded as a structured rank-one atom and the measurement model remains linear under hybrid combining. Recovery is then posed as a 2D atomic norm minimization, with path localization certified through a dual polynomial over the transformed domain. Numerical experiments show exact support recovery in the noiseless setting using only few compressed hybrid measurements. These results establish the proposed inverse-range atomic norm viewpoint as a new gridless foundation for near-field sensing and channel estimation in hybrid XL-MIMO and integrated sensing and communication systems.

The joint design of analog beamforming and power allocation is investigated for a single radio-frequency chain multiuser time-division multiple access system under a max-min signal-to-noise ratio (SNR) criterion. A hardware-efficient phased-array architecture is considered, where the beamforming vector is shared by all users and is subject to constant-modulus constraints. For any fixed analog beamformer, the optimal power allocation is first derived in closed form, by which the original problem is reduced to phase-shift optimization only. Then, globally optimal branch-and-bound (BB) algorithms are developed for discrete and continuous phase shifts. Numerical results show that the proposed BB algorithms achieve the global optimum and provide reliable benchmarks for evaluating the performance gap of low-complexity alternating-optimization methods.

The development of 6G networks brings an increasing variety of data services, which motivates the hybrid computation paradigm that coordinates the over-the-air computation (AirComp) and edge computing for diverse and effective data processing. In this paper, we address this emerging issue of hybrid data computation from an energy-efficiency perspective, where the coexistence of both types induces resource competition and interference, and thus complicates the network management. Accordingly, we formulate the problem to minimize the overall energy consumption including the data transmission and computation, subject to the offloading capacity and aggregation accuracy. We then propose a block coordinate descent framework that decomposes and solves the subproblems including the user scheduling, power control, and transceiver scaling, which are then iterated towards a coordinated hybrid computation solution. Simulation results confirm that our coordinated approach achieves significant energy savings compared to baseline strategies, demonstrating its effectiveness in creating a well-coordinated and sustainable hybrid computing environment.

A hybrid physical and geometrical optics method is proposed to model the subsurface imaging using mmWave FMCW radar. Modeling of the wave propagation for subsurface imaging can improve the interpretation of acquired data and imaging results. Full-wave simulation is common in simulating wave propagation. However, when the frequency is high such as mmWave frequency, it is difficult to implement since it costs large computation resource and time. In this paper, the physical and geometrical optics are hybridized to simulate the wave propagation in subsurface imaging scenarios. In the proposed method, physical optics method is utilized to calculate the reflection from the object and geometrical optics method is utilized to calculate the transmission of the wave through object. By combining the results from physical and geometrical optics, the wave propagation in the subsurface imaging scenarios is simulated. The synthetic-aperture radar imaging is applied to the simulated data and the image is successfully reconstructed. Further, the experiment setup is developed and the comparison between simulation and experiment is carried out. The results demonstrated that the proposed simulation method can model the subsurface imaging with mmWave FMCW radar.

As an effective approach to understanding the human-centric physical world, Wearable Artificial Intelligence (AI), which leverages multimodal wearable sensors to understand human physiology and behavior, has attracted increasing attention in recent years. However, existing sensor models remain largely siloed by modality and task, lacking a unified paradigm for integrating diverse wearable modalities, training strategies, and achieving robust generalization in real-world applications. Motivated by the success of multimodal foundation models, which learn transferable representations from massive multimodal data, we argue that Large Sensor Models (LSMs), defined as foundation models trained on large-scale and multimodal wearable data, offer a promising pathway toward a more general and scalable framework for wearable AI. In this position paper, we formalize the data substrate underlying LSMs, analyze the unique challenges of large-scale wearable sensing, and articulate two directions: (i) LSMs without language capability and (ii) LSMs with language capability. We further discuss representative application areas that can be unlocked by such models. Through this paper, we encourage the community to explore LSMs as a foundational approach for the next generation of human-centric AI systems.

Integrated sensing and communication (ISAC) is widely regarded as one of the key enabling technologies for future sixth-generation (6G) wireless communication systems. In this work, we investigate a bistatic ISAC system in the presence of a disco reconfigurable intelligent surface (DRIS), whose random and time-varying reflection coefficients emulate a "disco ball." The introduction of the DRIS breaks the underlying assumption in existing ISAC systems that the sensing and communication channels remain static or quasi-static within the channel coherence time. We first develop a bistatic system model incorporating the DRIS and characterize all involved wireless channels. Then, an ISAC waveform design that balances sensing and communication performance is proposed by formulating a Pareto optimization problem, where the trade-off is controlled through a tunable factor. Communication and sensing performance in the bistatic ISAC system are quantified by the signal-to-interference-plus-noise ratio (SINR) and the Cramer-Rao lower bound (CRLB), respectively. To quantify the impact of the DRIS on the bistatic ISAC system, we derive the statistical characteristics of DRIS-induced active channel aging (ACA) channels for communications and the cascaded DRIS-based sensing channel. Then, we establish a theoretical lower bound on the SINR and closed-form CRLB expressions in the presence of a DRIS. The analysis reveals several distinctive properties of the DRIS in bistatic ISAC systems. In particular, the DRIS degrades communication performance significantly due to the introduction of ACA interference. In contrast, with respect to sensing performance, the DRIS decreases the estimation accuracy of the angle of departure (AoD) while concurrently enhancing that of the angle of arrival (AoA). Numerical results validate the derived theoretical analysis and confirm these DRIS-induced behaviors.

In this paper, we propose a distributed optimization-learning framework for terahertz (THz) cell-free integrated sensing and communication (CF-ISAC) systems, termed Distributed Optimization-Learning with Graph Transformers (DOLG). We first formulate a highly non-convex joint scheduling and signal design problem for THz CF-ISAC systems, jointly optimizing access point (AP)-user equipment (UE) association and beamforming under signal to interference plus noise ratio based communication and Cramér-Rao bound based sensing constraints, together with line-of-sight-driven visibility rules and per-AP power constraints. We also develop an optimization based benchmark utilizing a tractable relaxed reformulation. Building upon this optimization structure, we redesign a graph transformer network (GTN) as an optimization-aware representation module that encodes cross-field wavefront geometry, blockage visibility, and sensing relevance in a permutation-equivariant manner. The proposed DOLG framework amortizes the iterative optimization procedure into a scalable GTN-conditioned distributed multi-agent reinforcement learning policy through centralized training and decentralized execution, while preserving per-AP power constraints via structure-preserving projections. Simulation results demonstrate that the proposed DOLG framework achieves stable convergence and effectively balances the communication-sensing tradeoff. From the system-level perspective, it outperforms multicell and non-joint design baselines. Furthermore, it surpasses conventional optimization based and heuristic approaches in terms of both ISAC performance and computational scalability.

High-mass X-ray binaries are one of the primary drivers of the 21-cm signal from Cosmic Dawn and Reionization, playing a leading role in the thermal history of the intergalactic medium. In traditional semi-numerical simulations, a deterministic scaling relation between the total X-ray luminosity of high-mass X-ray binaries, $L_{\rm X}$, and star formation rate (SFR) is usually adopted. However, this assumption is inaccurate for high-redshift low-SFR regions hosting few sources. The spatial variation in the number and luminosity of these sources is expected to enhance fluctuations in the Cosmic Dawn 21-cm signal. Here we quantify this effect by introducing a stochastic $L_{\rm X}$ model sampled from a power-law X-ray luminosity function. Implementing this in 21cmSPACE, a large-scale simulation framework of Cosmic Dawn and Reionization, we find that the stochasticity leads to enhanced fluctuations in X-ray heating rate fields, and affects the 21-cm power spectrum on small scales ($k>0.3~ \mathrm{cMpc^{-1}}$). The impact of stochasticity on the global 21-cm signal and on the large-scale power spectrum is found to be negligible. Our results suggest these effects will remain undetected by the upcoming Square Kilometer Array. However, large-scale lunar-based experiments may be sensitive to the signatures of stochastic X-ray heating at $z\sim 25$. Quantifying these corrections is a vital step toward robust 21-cm modeling and ensuring that future precision data interpretation is free from astrophysical biases.

Global m=1 asymmetries are observed in many self-gravitating astrophysical systems and are often interpreted as large-scale slow modes in near-Keplerian potentials. Prominent examples include eccentric nuclear disks in galactic centres, such as the double nucleus of M31. However, the origin and long-term stability of such modes remain unclear. We investigate the evolution and stability of a collisionless, self-gravitating torus orbiting a dominant central mass, aiming to determine whether a slow non-axisymmetric (m=1) mode can arise spontaneously. We perform direct N-body simulations exploring different torus-to-central mass ratios and initial conditions. The calculations use the high-order Hermite GPU integrator (φ-GPU), allowing us to follow long-term evolution with many particles. We find that a global slow m=1 mode forms spontaneously from initially axisymmetric configurations. The lopsided structure is sustained by coherent apsidal alignment and persists over secular timescales. Its maintenance requires nonlinear coupling of low-order modes, including the m=3 component, as well as a sufficient vertical thickness of the torus. As a result of the long-lived overdensity, the central mass is displaced from the system barycenter. These results provide a framework for understanding eccentric nuclear disks, such as those in M31 and NGC4486B, as well as molecular tori in AGNs, and suggest that such asymmetries may produce observable offsets of the central supermassive black hole.

We present a systematic analysis of scale-dependent properties of filamentary structures in seven nearby molecular clouds -- Taurus, Ophiuchus, Perseus, Orion A, California, IC 5146, and Vela C -- using the multiscale extraction method getsf. Alongside the usual surface density profiles $Σ(r)$, we derive volume density profiles $ρ(r)$ for a large sample of filaments, providing new observational constraints on their three-dimensional structure. Filament widths increase systematically with spatial scale, following power laws $\tilde{H} \propto Y^{0.50}$ and $\tilde{h} \propto Y^{0.37}$, with distributions spanning $\sim$ 0.01--1 pc across all scales, challenging the notion of a universal filament width of $\sim$ 0.1 pc. The median volume density slopes $\tildeβ \approx 2.1$--$2.4$ are systematically lower than the value $β= 4$ expected for an isothermal cylinder in hydrostatic equilibrium. Volume density contrasts are substantially higher than surface density contrasts ($\tilde{C}_ρ \approx 17$--$52$ vs. $\tilde{C}_Σ \approx 1.1$--$2.7$), confirming that filaments are substantially more prominent in three dimensions than their projected appearance suggests. The median linear densities of filaments increase linearly with spatial scale, $\tildeΛ \propto Y$, with the fraction of supercritical filaments ($Λ> 15\,M_\odot$ pc$^{-1}$) increasing strongly with scale and varying widely among clouds, from $\sim$ 7% in Taurus and Ophiuchus to $\sim$ 54% in Vela C, broadly consistent with the known star formation activity of the clouds. Measured filament widths and slopes depend systematically on angular resolution and distance, highlighting the importance of accounting for resolution bias in comparative filament studies.

Complex structures often emerge from initially homogeneous or weakly correlated states. We address the apparent tension between this ordering and entropy growth through a unified framework combining semi-microscopic phase-space dynamics, transport geometry, information theory, and coarse-grained effective modeling. The key point is that entropy depends on the level of description: a coarse-grained spatial field may become more ordered as structure forms, even while the full phase-space description becomes more complex through shell crossing, multistreaming, and the activation of velocity degrees of freedom. Using a Lagrangian--Eulerian transport map, we show how density amplification is governed by the Jacobian of the deformation and how anisotropic collapse arises from the eigenvalues of a hierarchy of deformation tensors. Long-range interaction or information flow is encoded in the displacement field, so that nonlocality enters directly through transport. We connect this geometric description to a maximum-entropy Gaussian baseline and show how nonlinear transport and nonlocal coupling generate scale coupling, higher-order correlations, and non-Gaussianity. We then formulate a Landau--Ginzburg description in which the growth of seed anisotropies is interpreted as the activation of lower effective free-energy branches, providing a coarse-grained realization of self-organization. Applied to generated cosmological fields, this framework indicates that the nonlocal tidal level becomes relevant already at moderate overdensity. Although cosmological structure formation is the main realization considered here, the framework is intended more broadly as a mesoscopic language for systems in which transport, anisotropy, nonlocality, and self-organization are central.

We construct a class of stationary, axisymmetric, horizonless spacetimes whose curvature is generated entirely by smooth, localised differential rotation $Ω(r)$, while the spatial geometry remains exactly flat. Despite vanishing ADM mass, these helicity-supported configurations exhibit non-trivial curvature, finite tidal forces, and a gravitomagnetic field arising from the radial shear of the rotation. The twisted stationary Killing congruence produces global frame-dragging, including a gravitational Sagnac effect, and the effective potential admits stable circular orbits for null and timelike particles. The tidal tensor gives oscillatory restoring forces, ensuring stability against radial perturbations. Linearising the Einstein equations yields a wave equation for axisymmetric perturbations of $Ω(r)$; the effective potential is positive and localised, the operator is self-adjoint and positive definite, and the frequency spectrum is real, implying linear stability. Perturbations propagate as shear waves analogous to Alfvén waves. These results show that differential rotation alone can sustain a regular, asymptotically flat gravitational field with rich dynamics. This class of spacetimes provides a tractable platform for exploring gravitomagnetism, tidal and wave phenomena in smooth rotating backgrounds, with direct applications to rotating astrophysical structures.

Protostellar streamers are elongated structures that channel material from larger scale onto disks, influencing their physical and chemical evolution. The M512 protostar in Orion/Lynds 1641 hosts one of the most massive and extended streamer discovered so far, offering a unique opportunity to study these processes. We investigate the morphology, chemistry, and origin of this streamer,and its potential impact on the protostellar disk. Using archival ALMA observations of C18O, DCO+, N2D+, and HCO+, we compare their spatial distributions through moment maps and spatial profiles. The streamer shows clear chemical stratification: C18O lies on the western side of the protostar, N2D+ is farther out to the east, and DCO+ is in the middle. This suggests that the structure has been shaped by environmental effects rather than tracing a single coherent infalling flow, with only the densest gas near the protostar likely to accrete onto the disk. Overall, the bulk of the streamer reflects the physical and chemical imprint of the surrounding cloud, highlighting the importance of environmental shaping in interpreting streamer-disk connections and their role in disk growth.

Stellar microlensing is a powerful method to constrain compact dark matter models, uncover binary stars, and exoplanets during caustic crossing events. At cosmological distances, {\it James-Webb Space Telescope} ({\it JWST}) is routinely detecting microlensed giant stars in highly magnified galaxies behind massive lensing clusters. Here, we explore for the first time microlensing in modest redshift galaxies commonly seen through local Ultra Diffuse Galaxies (UDGs). Using the UDG NGC1052-DF2 as a case study, we found that detecting UDG microlensing events through UDGs is possible. However, a low total UDG microlensing event rate of $\sim 5.6\times10^{-2}\,\textrm{yr}^{-1}$ over its five background galaxies is expected for typical {\it JWST} $\sim 29\,$mag visits, and a low Vera Rubin Legacy Survey of Space and Time (LSST) detection rate of $\sim 2\times10^{-8}\,\textrm{yr}^{-1}$ such that NGC1052-DF2 might not be a prime target given its lack of low-redshift background galaxies. {\it Euclid} is ideal for identifying samples of low-redshift star-forming galaxies seen through local galaxies for deeper cadenced follow-up, where our zeroth-order calculation estimates that $\mathcal{O}(1-10)$ events per year are expected over the whole sky under the monitoring of LSST. Finally, we postulate that UDG microlensing will allow an independent estimate of the initial mass function (IMF) and the stellar multiplicity in the low mass regime, of considerable interest for UDG galaxies, where stellar mass has been claimed to predominate over dark matter in some cases, including NGC1052-DF2.

Nuclear density functional theory provides a unified description of finite nuclei and bulk nuclear matter, and is widely used to model the neutron star equation of state. However, extrapolations to supra-saturation densities require a quantified treatment of uncertainties arising from parameter estimation and functional choices. We present an updated Bayesian inference of a Skyrme energy density functional augmented by a flexible meta-model density dependence at high density. Nuclear observables are computed using a Gaussian emulator of the publicly available Milano HFBCS-QRPA code, enabling efficient exploration of a high-dimensional parameter space. Relative to previous analyses, we extend the calibration set with isospin-sensitive data, including masses and charge radii along selected Ca and Sn isotopic chains, and updated constraints from giant monopole resonances. The resulting posteriors are further constrained by \emph{ab initio} neutron-matter calculations and astrophysical observations, including recent NICER measurements, yielding consistent crust and core properties of catalyzed NS compatible with current constraints. Bulk nuclear-matter parameters are well approximated by a multivariate Gaussian with covariance matrix provided for direct reuse, while several finite-nucleus parameters exhibit pronounced non-Gaussianity.

Long-period transients (LPTs) are a newly identified class of radio sources with burst recurrence times from minutes to hours, and their diversity suggests multiple physical origins. CHIME/ILT J1634+44, with a short period of 841 s, a long-period modulation of 4206 s, and a significant negative period derivative, strongly suggests a binary origin. For such a short-period source, Roche-lobe constraints strongly favor an ultra-compact companion, motivating a double-white-dwarf (WD--WD) interpretation. In this Letter, we show that the WD--WD channel makes a sharp timing prediction: if the burst period is the orbital clock and the long-period modulation is a spin-orbit beat, then the modulation period is not a free timescale. Instead it must evolve jointly with the orbital clock and the spin clock through gravitational-wave losses, magnetic dissipation, and tidal interaction. For CHIME/ILT J1634+44-like parameters, we find that the beat clock drift $|\dot P_b|\sim 10^{-10} \text{ s s}^{-1}$, implying an observed-minus-calculated drift of tens of seconds in one year. Joint measurements of the burst period, modulation period, and their derivatives provide a minimal and falsifiable timing test of an ultra-compact binary origin.

We present 43\,GHz VLBI observations of the radio galaxy 3C~120 during its brightest $γ$-ray outburst (March 2018), recently associated with the IceCube neutrino alert IC-180213A. Despite reaching $L_γ= 3.7 \times 10^{44}$\,erg\,s$^{-1}$, contemporaneous X-ray monitoring from INTEGRAL/ISGRI, MAXI/GSC, and \textit{Swift}/XRT revealed no variability across 0.3-200\,keV, nor in B, V, R, and I band optical observations or 37 \& 235\,GHz observations, establishing an orphan flare. High-cadence VLBI imaging identified a new jet disturbance (N) propagating at $β_{\rm app} = (2.8 \pm 1.3)$ through quasi-stationary features C1-C3. The $γ$-ray peak coincided spatially and temporally with N crossing C3 ($r \sim 0.38$\,mas), where we measured a factor-of-5 increase in fractional polarization ($m = 16\%$) and $Δχ\sim 24^\circ$ EVPA rotation, indicating localized magnetic field compression. The extreme Compton dominance ($L_γ/ L_{\rm syn,blob} \approx 160$) is naturally explained by the Ring of Fire scenario, in which N ($Γ_{\rm blob} = 6$, $B_{\rm blob} = 0.023$\,G) inverse-Compton scatters synchrotron photons from C3, reproducing the observed $γ$-ray luminosity for physically reasonable parameters. Unlike the 2014-2015 orphan flares attributed to rapid spine reorientation near the BLR, the 2018 event represents a distinct physical mechanism, a propagating disturbance interacting with stationary jet structure at $\sim10\times$ the BLR radius.This work provides the first direct observational link between VLBI-resolved jet dynamics and orphan $γ$-ray emission in a radio galaxy.

We present an updated analysis of the gigaelectronvolt (GeV) gamma-ray emission from the shell-type supernova remnant (SNR) RX J0852.0-4622 (Vela Jr) using 15 yr of Fermi Large Area Telescope (Fermi-LAT) data. We quantitatively model the GeV morphology and find that it is best described by the masked H.E.S.S. shell template, indicating that the embedded pulsar wind nebula (PWN) contributes little to the GeV flux. The 0.1-500 GeV spectrum is well fitted by a hard power law with a photon index of $1.77 \pm 0.03$ and connects smoothly to the teraelectronvolt (TeV) spectrum, confirming previous results with improved precision. We further construct an independent eROSITA shell template and derive the 1-5 keV X-ray spectral energy distribution (SED) of the whole remnant, which provides new constraints on the synchrotron emission. We model the multi-wavelength (MWL) SED with a pure leptonic model and a hybrid lepton-hadron model. While the pure leptonic model reproduces the overall broadband shape, the hybrid model provides a better statistical description of the same dataset, supporting a mixed-origin picture in which the hadronic contribution is mainly relevant in the GeV band and the TeV emission remains predominantly leptonic.

Out of all the discovered open clusters (OCs) that are located in the inner part of the Galaxy, only a small fraction has been observed with high-resolution spectroscopy. An intriguing population of inner-disc OCs at relatively high altitudes ($Z$) from the Galactic plane remains poorly studied. There are few reliable detections of such OCs, and their occurrence rate, dynamical origin and survival mechanism remain uncertain. We perform a detailed spectroscopic analysis of UBC 1052, located at a cylindrical galactocentric radius $R_{GC} = 6.1$ kpc and $Z = 340$ pc, which stands out as the oldest and highest-|$Z$| inner-disc OC studied at high resolution to date. We used FLAMES/VLT to acquire high signal-to-noise ratio UVES spectra of four red clump (RC) members ($G\sim14$ mag). From them we derived high-precision radial velocities ($v_{r}$) and local thermodynamic equilibrium chemical abundances for 23 elements through a strict line-by-line differential analysis, achieving a median precision in [X/H] of $\simeq0.06$ dex for each star. The four RC stars have fully compatible chemical abundances, with [X/H] dispersions among them <$0.03$ dex for 20 elements. We also acquired GIRAFFE spectra of other candidate members and derived their $v_{r}$. We find that UBC 1052 has an age of $2.25\pm0.25$ Gyr, a distance of $3.11\pm0.07$ kpc, an extinction $A_{V} =1.23$ mag, a mean radial velocity of $34.0\pm0.6$ $km$ ${s}^{-1}$, and a slightly super-solar [Fe/H] = $0.05\pm0.01$ dex. Such relatively low [Fe/H] at its $R_{GC}$ suggests that UBC 1052 is a rare candidate for an inward-migrated OC in the inner disc. Its detailed abundance pattern (e.g. [Ba/Zr] and [Nd/Y]) shows some interesting features that appear to be unique in the current census of OCs studied at high resolution, making it an interesting object for potential strong chemical-tagging searches for already dispersed member stars. [Abridged]

We investigate the relationships between LHAASO TeV gamma-ray sources and various kinds of objects, including pulsar wind nebulae (PWNe), supernova remnants (SNRs), HII regions, microquasars, and OB associations. We propose a Randomization-Adjusted Overlap Correlation (RAOC) method to statistically assess association probabilities and evaluate association proportions across catalogs. The results reveal statistically significant overlaps between LHAASO sources and SNRs, PWNe, and microquasars, supporting their role as important contributors to TeV gamma-ray emission. The estimated association proportions of LHAASO sources are 0.19$\pm$0.08 with SNRs, 0.20$\pm$0.04 with PWNe, and 0.027$\pm$0.008 with microquasars. The proportion of the gamma-ray sources associated with the subsample of shell-type SNRs is ~0.1. While HII regions also show potential association, particularly with the KM2A component, their large self-overlap ratio complicates precise estimation. In contrast, OB associations exhibit a high probability of chance coincidence, suggesting their limited contribution to TeV gamma-ray emission. Our analysis of TeV gamma-ray emission capabilities shows that ~60% of PWNe are gamma-ray bright in both the WCDA and KM2A energy ranges. For SNRs and microquasars, the TeV gamma-ray bright fraction is ~10%. The subsample of PWNe associated with molecular clouds (MCs) shows enhanced gamma-ray emission. Furthermore, positional analysis reveals a systematic offset of the gamma-ray sources overlapping with PWNe toward the associated MCs. These findings imply a role for MCs in PWN gamma-ray production. Additionally, self-correlation analysis indicates that about 70% of the WCDA and KM2A gamma-ray components share a common origin. The study also identifies selection effects in existing SNR catalogs and notes clustering among approximately 30% of HII regions within larger star-forming regions.

In the recent preprint arXiv:2604.07244v1, the authors introduce a novel combination of redshift, distance, and expansion rate observables for constraining cosmic backreaction. The current work presents a first application of the method, yielding the first direct constraints on the total cosmic backreaction in our Universe over a significant redshift range. The constraints are consistent with vanishing backreaction within one standard deviation. However, the constraints are fairly weak and significant backreaction cannot be ruled out.

[Abridged] We aim to study the radial and vertical extents of 12CO gas, millimeter dust thermal emission and optical/NIR scattered light by dust in disks. We analyze a sample of 14 highly inclined protoplanetary disks. We present ALMA high angular resolution band 7 (0.9 mm) continuum images and 12CO (3-2) gas moment maps as well as HST and VLT/SPHERE scattered light images. The majority of disks in our sample (11 out of 14) follow Rgas > Rdust,micron > Rdust,mm. The other 3 disks appear more extended in millimeter continuum than in scattered light. Highly inclined disks tend to appear less radially extended in CO gas line emission than in millimeter dust continuum compared to less inclined disks. This results from optical depth effects and/or radial drift. The known correlation between disk size and millimeter continuum and line fluxes are confirmed in our sample with highly inclined disks significantly fainter than disks seen at lower inclination for a given disk radius. We found that this correlation is significantly tightened once fluxes are corrected for the disk inclination, consistent with the disks being optically thick at millimeter wavelengths. Regarding the vertical extent defined as the apparent emitting height, most disks in our sample follow Hgas > Hdust, mm. This strengthens our previous findings that the millimeter dust is highly decoupled from the gas and forms a layer in the disk midplane due to vertical settling. Most disks appear more vertically extended in gas than in scattered light, suggesting that the micron-sized dust is not fully coupled to the gas. We also estimated dynamical masses using PV diagrams for the first time for most of the objects in our sample. We found an anti-correlation between the dynamical mass and the aspect ratio, emphasizing the dominant role of gravity in setting the disk vertical extent, but no correlation with the disk radius.

We investigate how resolving the inner few astronomical units of a massive protostellar disk affects the migration, disruption, and accretion signatures of an inward-moving fragment. In particular, we aim to determine whether the predicted burst strength and duration depend on the adopted sink cell size. We present a new three-dimensional radiation-hydrodynamic simulation of a $\sim$5$M_{\odot}$ protostar surrounded by a self-gravitating disk, comparing the original 30 AU sink model to a refined model with a 1 AU sink that resolves the inner disk. The resulting gas structures are post-processed with radiative transfer calculations to derive synthetic photometry and multi-band images. Both simulations produce a major accretion burst as a migrating fragment is tidally disrupted, but their detailed behavior differs markedly. The refined model shows faster migration, a complete tidal disruption of the fragment, and a shorter, sharper outburst (more consistent with observations) with nearly the same peak accretion rate as the 30 AU model, which yields a broader, smoother event. The refined run produces much stronger near- and mid-infrared emission, reflecting the formation of a compact, hot inner disk. Resolving the inner few AU qualitatively changes the dynamics and observable appearance of fragment-driven bursts. Diffuse fragment disruption can reproduce decade-long events, but the much shorter ($<$3 yr) bursts observed in some massive protostars likely require the tidal disruption of more compact objects such as second Larson cores. Our trajectory analysis indicates that second Larson cores can migrate sufficiently close to the star to be tidally destroyed, offering a plausible mechanism for the fastest FU-Ori-like bursts observed in massive protostars.

The fiducial-independent transversal BAO dataset (BAOtr) systematically prefers smaller comoving distance ratios $D_{\rm M}/r_{\rm d}$ than the DESI DR2 three-dimensional BAO measurements at $z \lesssim 0.65$, driving dataset-dependent CPL dark-energy inferences and conflicting conclusions about the Hubble tension. We investigate whether this disagreement can be attributed to the $Λ$CDM fiducial assumed in the 3D BAO pipeline, or resolved within the CPL parametrisation. We show that the published 3D BAO distances are fiducial-independent by construction, with residual effects at $\lesssim 0.3\%$ -- negligible against the 10--18\% BAOtr uncertainties. We then scan the CPL parameter space with $Ω_m$ and $H_0$ jointly determined at each $(w_0, w_a)$ by the Planck $θ_*$ constraint and optimisation against the DESI data. Two complementary tests are performed: a direct comparison of each DESI-optimized model with the BAOtr data, and an $α$-interpolation test that anchors the prediction to the DESI measurements. Both reveal an inescapable trade-off: models that fit DESI well ($χ^2_{\rm DESI} \lesssim 5$) yield $χ^2_{\rm BAOtr} \gtrsim 42$, while reducing the BAOtr tension to $χ^2_{\rm BAOtr} \sim 37$ requires $χ^2_{\rm DESI} \gtrsim 8$. No CMB-consistent CPL model fits both datasets simultaneously. The direct comparison at $z = 0.510$ -- where BAOtr and DESI disagree by $3.7σ$ (data-versus-data) -- sets an irreducible tension floor that no smooth modification of $D_{\rm M}(z)$ can remove. These conclusions are robust across analysis methods, extrapolation schemes, and substitution of SDSS for DESI. The remaining explanations are observational systematics -- most plausibly in the angular BAO measurements -- or new physics beyond CPL.

We investigate a realistic non-supersymmetric hybrid inflation model incorporating right-handed neutrinos and assess its viability in light of recent cosmological observations. At tree level, the inflaton potential yields a blue-tilted scalar spectrum, which is disfavored by current data from Planck and ACT that instead support a red tilt. We show that including one-loop quantum corrections, arising from generic couplings required for reheating, significantly modifies the potential, flattening it at large field values. This leads to a red-tilted spectral index ($n_s < 1$) and a suppressed tensor-to-scalar ratio $r$, both consistent with observational constraints. To ensure theoretical control, we focus on sub-Planckian field values, where the effective field theory description remains valid. The coupling of the inflaton to right-handed neutrinos naturally facilitates efficient reheating and enables the generation of the baryon asymmetry via non-thermal leptogenesis. We further explore the model's parameter space using a multi-output random forest classifier, achieving prediction accuracies in the range of $87.5\%$ to $98.9\%$. Our analysis shows that approximately $15\%$ of the parameter space satisfies at least one current experimental constraint, underscoring the essential role of quantum corrections in reconciling particle physics models with precision cosmology, and highlighting the effectiveness of machine learning techniques in probing complex theoretical frameworks.

We investigate how the transition density \(ρ_{tr}\) affects hybrid constructions of the neutron-star equation of state (EoS) in which a nucleonic description at low densities is matched to a model-agnostic high-density extension based on a speed-of-sound parametrization. Using four representative nucleonic models--Taylor expansion, \(\frac{n}{3}\) expansion, Skyrme, and relativistic mean-field--built from identical nuclear matter parameters, we isolate the impact of the low-density EoS and the transition density on neutron star observables. We find that, within the present smooth-matching prescription, neutron star properties such as radii and tidal deformabilities retain significant sensitivity to the choice of low-density EoS for commonly adopted transition densities around \(ρ_{tr} \approx 2ρ_0\), even when the same high-density parametrization is employed. This residual dependence arises from differences in the matching conditions at \(ρ_{tr}\), which propagate into the high-density extension, so different low-density inputs lead to different effective high-density EoSs. These findings are robust across two distinct speed-of-sound parametrizations. Quantitatively, the model spread in radius and tidal deformability at $1.4\,M_\odot$ exceeds the current observational uncertainty by factors of $\sim 1.8$ and $\sim 1.4$ at $ρ_{\mathrm{tr}} \approx 2ρ_0$, whereas these factors reduce to $\sim 1.05$ and $\sim 0.4$ at $ρ_{\mathrm{tr}} = ρ_0$. Lowering the transition density, therefore, systematically diminishes the spread among models and leads to more consistent predictions. Our results demonstrate that the widely used choice \(ρ_{tr} \approx 2ρ_0\) does not guarantee model independence in hybrid EoS constructions, and should be treated as an explicit source of systematic uncertainty when inferring dense matter properties from neutron star observations.

Radio pulsars, as highly coherent point sources, serve as powerful probes of the ionized interstellar medium (IISM). Pulsar scintillation observations have revealed inverted arclets on the secondary spectrum, indicating quasilinearly aligned images created by substructures on a scintillation screen. The density profiles of these substructures remain unconstrained but are crucial to identifying their physical nature. This work employs a plasma lensing framework to study observable features of substructure phase screens. Using three lens models, we identify the substructure properties that can be constrained by observables. The outer caustic is the most prominent feature of a lensing substructure, measurable via multiepoch or ultrawideband observations. Its location constrains the maximum column density gradient of the substructure. The inner caustic, though difficult to observe except for substructures capable of producing extreme-scattering events, directly indicates the substructure size. Even when caustic locations are not observed, the minimum span where substructure images exist can be measured and used to place a lower limit on the column density amplitude. The logarithmic brightness of individual arclets forms a concave function of the pulsar-lens angular separation, contrasting with the convex brightness distribution of all substructure images--highlighting the complementarity of individual arclets to statistical studies. These findings reveal the potential of pulsar scintillation to uncover IISM substructure and underscore the need for multiepoch and/or ultrawideband measurements to constrain discrete lensing morphologies and help reveal the nature of interstellar plasma structures.

Extreme-mass-ratio inspirals (EMRIs) carry valuable information about their surrounding astrophysical environments. Over the course of their long-term evolution, interactions between the secondary object and the accretion disk can produce observable effects on both the orbital evolution and the emitted gravitational waveform. Based on the modifications to the companion's orbital evolution induced by the accretion disk environment, we investigate the feasibility of identifying the presence of accretion disk environmental effects in EMRI systems using gravitational wave signals. Within a Bayesian framework, we analyze the capability of EMRI systems with multiple parameter configurations to distinguish accretion disk environmental effects. Our results show that, under the $α$-disk model, all injected events can successfully identify the environment in which the EMRIs reside. Furthermore, we examined the improvement in the precision of Hubble constant measurements using the dark siren method after correctly identifying the accretion disk environment and constraining the relevant disk parameters. Constraining these environmental parameters may further deepen our understanding of the host environment, thereby enabling a more reliable inference of the physical properties of the accretion disk and its associated luminosity and ultimately improving the measurement of cosmological parameters. We find that the measurement precision for a single event can improve by as much as $20\%$. This work highlights the necessity of incorporating environmental effects into future EMRI data analysis. Proper modeling of such effects not only helps identify EMRI systems embedded in accretion disk environments but also further improves the precision of gravitational wave cosmological parameter inference.

In 2021, Lei et al. claimed the equivalence between the two Lagrangians $\mathcal{L}_1 =-mc\sqrt{-g_{μν}{\dot{x}}^μ{\dot{x}}^ν}-V$ and $\mathcal{L}_2 = \frac{1}{2}mg_{μν} {\dot{x}}^μ{\dot{x}}^ν-V$ for describing particle dynamics in combined gravitational and matter fields. In the present work, we rigorously demonstrate that their equivalence depends critically on the external potential V. Both Lagrangians yield identical Hamiltonians that strictly satisfy the mass shell constraint, and are therefore equivalent when V vanishes or corresponds to an electromagnetic potential. However, they are generally not equivalent for generic external potentials excluding the electromagnetic ones. This discrepancy arises because L1 and L2 correspond to different Hamiltonian formulations. The Hamiltonian derived from L1 inherently enforces the mass shell constraint, whereas the Hamiltonian from L2 does not. When the Schwarzschild metric supplemented with an artificial mechanical potential is taken as a toy model, numerical investigations reveal that L1 leads to chaotic behavior, which signifies non-integrable dynamics. By contrast, L2 can be shown analytically to produce integrable dynamics free of chaos. In many scenarios, L1 is strongly recommended due to its theoretical superiority and universality. L2 is generally suitable for classical approximate problems involving low energy and weak gravity. Nevertheless, it is the preferred choice for strong field problems concerning the dynamics of charged (or neutral) particles near black holes with (or without) external electromagnetic fields, owing to its mathematical simplicity and computational efficiency. Moreover, it can still satisfy the mass shell constraint when an additional constraint is imposed on its corresponding Hamiltonian.

We model the Seyfert II AGN NGC 1068 within a turbulence-induced magnetic reconnection framework to explain its high-energy emission. Observations reveal a neutrino flux excess higher than the observed GeV gamma-ray emission by orders of magnitude, with no detected TeV counterpart, suggesting efficient hadronic acceleration in the nuclear region with strong gamma-ray absorption. Assuming that proton acceleration occurs in a turbulent reconnection layer via a first-order Fermi process, we use a lepto-hadronic model based on a coronal-accretion disk configuration in which magnetic field lines anchored to the $2 \times 10^{7} M_{\odot}$ black hole horizon reconnect with field lines from the inner accretion disk corona. Our model matches the observed spectral energy distribution with a magnetic field $B_{c} \sim 10^{4}$ G and magnetic reconnection power $\dot{W_{B}} \sim 10^{43}$ erg s$^{-1}$, with $\sim 50\%$ efficiency in proton acceleration. Unlike previous studies, we find that both particle acceleration and emission take place in the inner region, where protons reach $\sim 10^{14}$ eV via first-order Fermi acceleration within the turbulent reconnection layer, rather than drift acceleration. These protons interact with disk photons, coronal X-rays, and coronal protons, producing neutrinos, predominantly via $pp$ interactions, at levels consistent with IceCube detections. The associated gamma-rays are attenuated by $γγ$ annihilation, remaining below current upper limits. Turbulence-driven reconnection is thus a viable mechanism for neutrino production in the coronal region of NGC 1068 and similar sources.

Context. Nearly 150 tidal streams are known in the Milky Way, but full phase-space information exists for only a few. RR Lyrae stars (RRL), as standard candles, provide a powerful way to probe these structures, yet they have been identified in less than a dozen streams. Aims. We study the RRL population in all known stellar streams with reported proper motions in the galstreams library, performing the first systematic census of these stars. Our goals are to identify likely RRL members, map distances along streams, and compare RRL populations in streams and their progenitors. Method. We use a union of the largest RRL catalogs (Gaia DR3 SOS, PS1, and ASAS-SN-II) to construct a Bayesian probabilistic membership model and find 361 RRL in the 56 streams studied. Results. i) We find that 32 of the 56 streams have RRL in their tidal tails -- 13 with progenitors and 19 without; 13 of these contain more than 3 RRL in their tails. ii) We report new RRL detections in 31 of these streams, anchoring distances and, in particular, inferring new distance gradients for 5 of them. iii) Our method provides intrinsic dispersion estimates in distance and proper motion for each track and statistically quantifies the expected contamination. iv) The census revealed some complex origin histories, such as the new plausible origin scenario we propose for M92 with multiple progenitors. v) We find that the presence of RRL in the tidal tails is linked to the late stages of progenitor dissolution. Conclusions. This census represents a first step toward identifying which of the studied stellar streams contain a significant number of RRL based on currently reported tracks while also providing a homogeneous and robust catalog of RRL members with precise empirical distances, crucial for a full phase-space analysis of these structures and their use as probes of the Galaxy's history and gravitational potential.

It is a common miss-conception that 1E 0657-56, the "Bullet Cluster", is somehow inconsistent with MOND expectations. The argument centres on the fact that the baryonic matter distribution of this system is dominated by the X-ray emitting gas, while the total projected surface density required under General Relativity to explain the observed lensing signal, centres on the observed galaxies. This is sometimes interpreted as being in conflict with MOND, as under such an interpretation, it is naively assumed that all dark matter being absent, the gravitational potential should necessarily be dominated by the largest mass distribution, that of the gas. However, just as under General Relativity, under MOND, the total gravitational potential of a system depends sensitively upon the volume density and not just on the total mass. It is shown in this {\it letter} that the surface density which QUMOND predicts will be inferred under a standard gravity framework from the total gravitational potential of the Bullet Cluster, closely matches what General Relativity inferences of lensing observations return. The close-to-point-like galaxies imply under QUMOND a relatively much larger surface density signal than what is expected from the Mpc scale gas distribution.

Modern radio and multi-instrument astrophysical datasets are increasingly assembled from surveys with different sensitivities and selection effects. In such heterogeneous datasets, published measurement uncertainties are often incomplete, non-uniform across subsets, or missing cross-correlation information altogether. This limits reliable statistical inference, since underestimated or inconsistently modeled uncertainties can distort fitted spectral shapes, bias parameter estimates, and obscure physically meaningful structure. We introduce the Fisher Information Metric Error Reconstruction (FIMER), an information-geometric framework for reconstructing effective measurement uncertainties directly from heterogeneous astrophysical data. FIMER combines weighted Fisher-information geometry, FBET and an adaptive discrete hyperparameter search, while incorporating prior statistical knowledge of detector behavior into the weighting procedure. The priors used are not chosen as arbitrary tuning prescriptions or uninformative regularizers; they are motivated by statistical properties of the underlying detection process. Poisson priors represent counting-statistics behavior, while extreme-value priors allow tail-dominated fluctuations to be incorporated when rare or asymmetric excursions are expected to influence the inferred uncertainty distribution. We apply FIMER to radio SEDs of RxAGN using COSMOS VLA data at 1.4 and 3 GHz together with GMRT data at 325 and 610 MHz. The results show that FIMER provides a practical route to uncertainty reconstruction in heterogeneous survey combinations, especially when reported uncertainties are unavailable, underestimated, or strongly correlated. The method is particularly relevant for archival and multi-survey astrophysical datasets, where full covariance information is rarely available but reliable statistical inference remains essential.

We introduce a framework connecting dark matter self-interactions (SIDM) to the large-scale topology of cosmic reionization in this paper. SIDM core formation reduces the gas binding energy in high-$z$ halos, enhancing supernova-driven clearing of ionizing-photon escape channels. We decompose the observable signatures into two scale-dependent levers a percent-level shift in the emissivity-weighted halo bias $b_γ$ that modifies large-scale 21\,cm power, and a factor of 2--4 suppression of emissivity shot noise from increased duty cycles that reshapes intermediate-scale ionization morphology. We derive analytic predictions for the 21\,cm power spectrum ratio and validate them with a halo-by-halo semi-numerical excursion-set framework at $128^3$ resolution where individual halo duty-cycle stochasticity is resolved. For $σ/m = 1$--$10\;\mathrm{cm^2/g}$, we find $\sim 60$--$80\%$ suppression of emissivity shot-noise power, a $12$--$21\%$ reduction in the emissivity variance ratio $\langle \dot{N}^2\rangle/\langle\dot{N}\rangle^2$, and a $50$--$110\%$ increase in the Euler characteristic of the ionization field at fixed $\bar{x}_{\rm HI} = 0.5$. SIDM produces more numerous, more uniformly distributed HII bubbles compared to CDM's topology of fewer large bubbles around rare bright sources. Theintermediate scale signatures are potentially detectable by SKA1-Low in $\sim 1000$ hours. Our results establish reionization topology as a new, complementary probe of dark matter microphysics at mass scales $M \sim 10^{10}$--$10^{11}\,M_\odot$ and redshifts $z \sim 6$--$10$.

Recent observations of hydrostatic structure and virial equilibrium in supersonically turbulent, self-gravitating molecular clouds imply a stability that contrasts with the transcience of turbulent structure. To investigate this contradiction, we model a molecular cloud as a turbulent eddy and study its evolution as a dynamical system. In a two-dimensional phase space of structure and energy, we find that the dynamical equilibrium is a saddle point, stable in the direction aligned with force balance, but unstable in the direction of energy balance because of the combination of the turbulent dissipation and the negative heat capacity of self-gravitation. Near the saddle point, evolutionary trajectories follow a characteristic pattern that first approaches the equilibrium before departing in the direction of instability. Since the phase-space speed is proportional to the virial and energy imbalance, trajectories slow near the equilibrium resulting in a local overdensity of clouds. Also, near equilibrium, the relaxation to force balance is faster than the growth rate of the instability in energy. Consequently, more clouds are observed in near equilibrium states with hydrostatic structure even though the equilibrium is metastable. This resolves the apparent contradiction of equilibrium structure observed in dynamically unstable, self-gravitating turbulence.

The vertical structure of galactic disks is an important probe of disk assembly history. We investigate a sample of 79 local disk galaxies within 50 Mpc using data from the DESI Legacy Imaging Surveys. Vertical luminosity profiles as a function of radius in the g, r, and z bands are extracted and fitted with a single-component sech^2 model to determine the scale height and its radial variation (flaring). Our results show that local galactic disks are generally thin with negligible flaring. The median scale heights at one effective radius are 0.21, 0.22, and 0.22 kpc in the g, r, and z bands, respectively, while the corresponding median radial gradients are -0.006, 0.003, and 0.001. These values are consistent with those of the geometric thin disk of the Milky Way, represented by metal-rich, low-[alpha/Fe] populations, supporting the conclusion of weak flaring. We also find a clear positive correlation between scale height and stellar mass, extending down to 10^7 solar masses. These results provide a homogeneous benchmark for the vertical structure of nearby disk galaxies and support the cosmological representativeness of the Milky Way thin disk.

A method of determining ages and masses of asymptotic giant branch (AGB) stars between 0.8 and $\sim$6 M$_\odot$ is demonstrated, based on comparing the star's position in the period--absolute-magnitude diagram to theoretical evolutionary models. For samples of Milky Way stars, the method provides errors (statistical and systematic, respectively) of order of $^{+29}_{-35} \pm 15$ per cent in age, $^{+14}_{-7} \pm 7$ per cent in initial mass and $^{+17}_{-11} \pm 27$ per cent in current mass. However, its applicability to individual stars depends strongly on both their position in the $P-L$ diagram and the uncertainty of that position. This method is applied to published samples of AGB stars from the \emph{Gaia}, NESS, DEATHSTAR and ATOMIUM surveys. These surveys' statistical ensembles are compared to expectations from stellar evolutionary models, finding that most AGB samples are biased towards stars of younger ages and higher masses. An average mass for Milky Way AGB stars is found to be $\sim$1.1 M$_\odot$, while mass returned to the interstellar medium by AGB stars typically comes from $\sim$1.2 M$_\odot$ stars with mass-loss rates of order $2-3 \times 10^{-6}$ M$_\odot$ yr$^{-1}$.

Fast radio bursts (FRBs) and gamma-ray bursts (GRBs) are both linked to compact-object activity, yet their possible connection remains unclear. Here we perform a systematic search for spatial and temporal associations between FRBs in the second CHIME/FRB catalog and Swift GRBs. Instead of using the positional ellipses reported in the catalog, the full CHIME localization probability maps are adopted for spatial cross-matching. This yields 130 candidate pairs and increases the number of spatially consistent matches by a factor of several. A redshift consistency requirement reduces the sample to 45 pairs. Applying an additional temporal criterion, requiring long GRBs to precede FRBs and short GRBs to follow them, further reduces the sample to 26 candidates. Monte Carlo simulations show that the overall excess of associations is not statistically significant, and the distribution of matches across localization confidence levels is consistent with random expectations. However, potential associations may be diluted by localization uncertainties and a dominant background of chance coincidences. These results place constraints on any FRB-GRB connection and highlight the need for improved localization and larger samples.

Low-ionization metal absorption lines provide a primary probe of cool gas in and around galaxies. We confront observations of metal-line absorption in quasar spectra with predictions from the IllustrisTNG cosmological simulation in order to benchmark how well current galaxy formation models reproduce the observed circumgalactic medium (CGM) and intergalactic medium (IGM) absorption signatures. We implement two ionization prescriptions: a purely collisional model and a model including photo-ionization by a uniform ultraviolet background (UVB). Using a grid-based framework, we compute MgI, MgII and FeII column densities and construct column density probability distribution functions (PDFs) and equivalent width (EW) statistics for comparison with observations. The observational samples considered here are based on the High Resolution Echelle Spectrometer (HIRES), the Ultraviolet and Visual Echelle Spectrograph (UVES), the Sloan Digital Sky Survey (SDSS) and the Dark Energy Spectroscopic Instrument (DESI). The computed PDFs broadly reproduce the observed ones across the sampled column density range of $10^{11.4}\lesssim \text{N}\lesssim 10^{16}\ \rm{cm^{-2}}$, indicating that the simulation captures the dominant physical drivers of low-ionization absorption. We then compute the cosmic incidence of MgII systems, namely the evolution of their number with redshift $d\mathcal{N}/{dz}$. The model that includes UVB accurately produces $d\mathcal{N}/{dz}$ up to equivalent widths (EW) of $\rm W_0^{2796} < 0.6\ \mathring{A}$, consistent with low-density photo-ionized gas in the outer CGM. At high EWs of $\rm W_0^{2796} > 1\ \mathring{A}$ TNG underestimates $d\mathcal{N}/{dz}$ and fails to capture its rise toward $z\sim2$.

$SU(\infty)$-QGR is a recently proposed fundamentally quantum approach to gravity and cosmology. In this model the Hilbert space of the Universe represents $SU(\infty)$ symmetry. Its fragmentation generates approximately isolated subsystems (particles) representing, in addition to $SU(\infty)$, finite-rank local symmetries. The common $SU(\infty)$ is associated to quantum gravity, and at lowest quantum order the effective action for all symmetries is Yang-Mills on a 4D parameter space $Ξ$. Nonetheless, physical processes and measurables must be independent $Ξ$'s geometry. In previous works we demonstrated that diffeomorphism of $Ξ$ can be neutralized by $SU(\infty)$ gauge transformation. In this work we show that invariance of action under metric change leads to a constraint resembling Einstein equation. It consists of energy-momentum tensors for all components of the model, including the spin-1 gravitons. In addition, through calculation of quantum information measures we study the effect of Hilbert space fragmentation on the evolution of emergent classical spacetime and cosmological phenomena, namely inflation and late time accelerating expansion. The results show that fields associated to these processes may be order parameters collectively presenting the evolution of quantum states of the contents of the Universe.

The exceptionally high X-ray polarization recently detected in the magnetar 1E 1547.0-5408 is considered a strong candidate signature of quantum electrodynamic vacuum birefringence, an interpretation that hinges critically on the source's viewing geometry. This stark contrast to the typically lower polarization degrees seen in other magnetars prompts a fundamental question: to what extent does viewing geometry, rather than intrinsic physics, drive the observed polarization diversity? To answer this, we perform a systematic, comparative geometrical analysis of two magnetars representing opposite extremes: the high-polarization source 1E 1547.0-5408 and the low-polarization source 1E 2259+586. The data are modelled within a unified Bayesian framework with both the classical rotating vector model (CRVM) and a twisted-magnetosphere extension (MRVM). For 1E 2259+586, both models favour a geometry with moderate magnetic inclination and viewing angles but a small impact angle. By combining the single-epoch phase-resolved fit with three-epoch phase-averaged position angle measurements, we find no significant secular evolution of the twist parameter λand derive a conservative upper limit of |Δλ|<0.79 at the 95 per cent level over 26 day. For 1E 1547.0-5408, the observed position angle curve is already well reproduced by the CRVM, while the MRVM shows no statistically significant advantage. When radio-informed priors are imposed, the posterior shifts towards a nearly aligned configuration consistent with the radio constraints.Both sources show no evidence for strong, static global twists in the current epoch. The observed polarization dichotomy arises from the confluence of viewing geometry, intrinsic surface emission physics, and magnetospheric propagation effects.

The massive binary common envelope (CE) phase plays a pivotal role in the formation of close black hole/neutron star (BH/NS) binaries, yet significant uncertainties remain in our understanding of this process. In this study, we aim to constrain the massive binary CE phase by systematically reconstructing three observed BH X-ray binaries (BHXBs): GRO J1655-40, SAX J1819.3-2525, and 4U 1543-47. Through comprehensive binary evolution simulations and parametric supernova (SN) modeling, we establish lower limits for the CE efficiency parameters under different energy considerations within the standard energy formalism. Specifically, we derive minimum values for three cases: $α_{\rm 0.5U}$ and $α_{\rm U}$ representing CE efficiencies with half and all of the internal energy contributing to the envelope ejection, respectively, and $α_{\rm H}$ accounting for the envelope's enthalpy. Our analysis reveals that the self-consistent formation of these three BHXBs requires CE efficiency parameters satisfying: $α_{\rm 0.5U}\gtrsim 6.7$, $α_{\rm U}\gtrsim 4.2$ and $α_{\rm H}\gtrsim 1.7$. Notably, we find no viable solutions with CE efficiency values below unity, even when considering the most extreme scenarios in which the envelope binding energy is significantly reduced through enthalpy inclusion. {Our results strongly imply that either additional energy sources are required, or the formalism itself must be revised.} Furthermore, we quantitatively assess the impact of BH natal kicks on our results. A key finding is that 4U 1543-47's formation requires substantial natal kicks ($\gtrsim 50 \;\rm km/s$), as lower kick velocities are incompatible with isolated binary evolution.

The detection of $γ$-ray candidates up to the energy of $\approx$13 TeV from the exceptionally bright $γ$-ray burst GRB 221009A by the Large High Altitude Air-Shower Observatory (LHAASO) has raised considerable interest in the astrophysical community. The $γ$-ray dataset resulting from the LHAASO observations allows one to reconstruct the intrinsic spectrum of GRB 221009A with an unprecedented precision. This intrinsic spectrum reveals a downturn at the energy of several TeV (statistical significance $\gt 5 σ$), i.e. the reconstructed intensity is below the intensity expected for a power-law spectrum. We show that a significant TeV $γ$-ray component may be produced by neutrons from photohadronic interactions inside the fireball. These neutrons escape the fireball and interact with the surrounding matter, giving rise to a flux of electrons and positrons, eventually resulting in an observable flux of GeV--TeV synchrotron photons -- a high energy "echo" of the GRB prompt emission phase. Finally, we show that at multi-TeV energies the contribution of $γ$ rays from intergalactic electromagnetic cascades initiated by primary ultra high energy protons is severely limited during the early afterglow phase for the typical magnetic field strength in the intergalactic filaments above 1 nG.

Black hole spacetimes contain several geometrically distinguished hypersurfaces, including event and Cauchy horizons, stationary-limit surfaces, curvature singularities, and asymptotic infinity. These structures are usually identified by different geometric or causal criteria. Here, we show that for the Kerr-Newman black hole, a single scalar function encodes all of them at once. The function arises by analytically continuing the membrane-paradigm pressure of the stretched horizon into the full spacetime. In fully factorized form, its zeros locate the outer and inner horizons, its poles locate the outer and inner stationary-limit surfaces, its higher-order divergence identifies the ring singularity, and its decay at large $r$ captures the asymptotic region. Thus, the analytically continued membrane pressure serves as a unified global detector of the critical surfaces in the Kerr-Newman geometry. We further note that the same analytic structure admits a secondary interpretation as an effective generalized multi-component van der Waals-type equation of state, whose intrinsic scales are fixed by the distinguished radii of the spacetime itself.

Sagittarius A* (Srg A*), the supermassive black hole at the center of the Milky Way, provides a unique laboratory to study accretion dynamics and plasma processes near the event horizon. We investigated the variability and polarization properties of Srg A* using ALMA observations during the 2018 Event Horizon Telescope campaign. We analyzed high-cadence full-polarization light curves from ALMA at millimeter wavelengths, performed time-series analysis, and investigated the temporal behavior during an X-ray flare observed by Chandra on 2018 April 24. The variability characteristics are compared with expectations from standard accretion flow models. We find low variability in total intensity ($σ/μ< 10\%$), but significantly higher variability in linear and circular polarization (~ 30% and ~ 50%, respectively). A time-series analysis reveals red-noise variability, with power spectral densities between -2 and -3 across all Stokes parameters. Polarized intensity shows stable intra-day timescales, while total intensity exhibits more variable timescales, suggesting distinct emission regions, with polarization likely arising from a coherent structure. On April 24, a statistically significant inter-band delay in polarized intensity coincides with a near-simultaneous X-ray and millimeter peak that deviates from the typical delayed flare scenario. This event also features enhanced millimeter variability and coherent polarization loop evolution. The observed simultaneity challenges standard models of transient synchrotron emission with cooling delays, favoring instead a scenario of continuous energy injection in an optically thin region. Our results offer new constraints on the physical mechanisms driving variability in Srg A*, and provide key observational input for refining theoretical models of accretion and plasma behavior in the vicinity of supermassive black holes.

We report the discovery of diffuse stellar substructure in the Milky Way's outer halo toward Boötes, unveiled by deep imaging data of the Subaru/Hyper Suprime-Cam. This substructure is detected as an excess of faint main-sequence stars, at heliocentric distances beyond 30 kpc, extending over at least 100 $\mathrm{deg^2}$. To infer its origin, we compare the projected spatial distribution of these stars to that of simulated tidal debris from the Large Magellanic Cloud (LMC), under the assumptions that the LMC is on either its first or second passage of the Milky Way. We found that the observed overdensity lies in a region of the halo where debris from the LMC is expected if it is on its initial pericentric phase 7-8 Gyr ago, which is predicted in the second-passage model, while the first-passage model is unable to explain the observed substructure. Chemo-kinematical data are required to further constrain its past orbit and to understand the origin of this new halo substructure, as should be obtained in the near future with deep photometric surveys such as UNIONS and LSST, and wide-field spectroscopy such as possible with PFS and DESI.

We investigate the structure of the Milky Way's stellar halo within 70 kpc of the Sun using a wide-field photometric catalog obtained from the Hyper Suprime-Cam (HSC) Subaru Strategic Program (HSC-SSP). We employ a large sample of main-sequence turn-off stars as distance tracers. To robustly derive the structural parameters of the stellar halo, we develop a forward-modeling framework that explicitly accounts for distance uncertainties, the solar position, and the limited sky coverage of the survey. Applying this method to the HSC-SSP catalog, we found that the smooth stellar halo is well described by a double power-law density profile, with inner and outer slope of approximately -3.3 and -4.8, respectively, with a break radius of 17.4 kpc. The outer steep density slope derived in this work supports a picture in which the present-day structure of the Milky Way's stellar halo is influenced by early massive accretion events, consistent with inferences from kinematic substructures such as Gaia Enceladus/Sausage. Ongoing wide-field imaging surveys, including UNIONS and LSST, will provide further constraints on the structure of the stellar halo and key insights into its formation history.

The rapid variability observed in very-high-energy (VHE) sources-such as pulsars, gamma-ray bursts (GRBs), and flares from active galactic nuclei (AGN)-can be used to detect or constrain a potential violation of Lorentz invariance (LIV). These effects can be investigated by measuring time lags in the arrival of VHE photons. However, an important source of uncertainty arises from intrinsic processes within the sources themselves that may induce photon delays unrelated to LIV. To address this challenge, we aim to combine observations of different sources, located at different redshifts. In this study, we present the results of a standardized analysis applied to all AGN observations conducted by the first Large-Sized Telescope of the upcoming Cherenkov Telescope Array Observatory. Our analysis includes a systematic search for intra-night variability in archival data from nights with significant excess detections for target candidates. By combining these observations, we derive constraints on the characteristic energy scales at which LIV deterministic or stochastic effects are expected to manifest.

One of the fundamental objectives of contemporary cosmology is to understand the physics of the inflationary universe, owing to its observably verifiable predictions about the very early universe with an energy scale of $\sim 10^{16}$ GeV. Recent observations from the ACT and the Planck mission, constrain the values of the scalar spectral index, $n_s$, and the tensor-to-scalar ratio, with state-of-the-art accuracy and upper limits, respectively. In the current work, a type of non minimally coupled inflationary model in which the gravity and the background scalar field interact through a covariant product of the Ricci tensor and derivatives of the scalar field. With this interaction at the backdrop, we estimate $n_s$ and $r$ for a wide range of inflaton self-interaction potentials, including power law, exponential $α$ attractor, Arctan, Hilltop, and polynomial model. We show that the higher derivative terms involving the scalar field resulting from the derivative coupling term can be handled without facing any singularity within the slow-roll regime. We show that it is possible to produce $n_s$ and $r$ values consistent with ACT and Planck observations for each of the chosen sets of potentials for the derivative coupled action.

We investigate inflationary magnetogenesis in a scenario where conformal invariance of electromagnetism is broken through a \emph{non-minimal Yukawa-like coupling between the inflaton and the Ricci scalar}. We account for electromagnetic backreaction and the Schwinger effect, analyzing both standard single-field inflation and a generalized K-essence framework, \emph{dubbed quasi-quintessence}. We consider inflationary potentials compatible with Planck satellite constraints, including Starobinsky and $α$-attractor models for large fields, as well as hilltop scenarios for small fields. Moreover, we explore very different functional electromagnetic couplings, introducing a novel ansatz modeled for small-fields. We show that the non-minimal coupling plays a central role in controlling the dynamics, \emph{acting as a timing parameter that regulates the onset of electric backreaction and the Schwinger regime}. This leads to a deep modification of the magnetogenesis process. Indeed, the amplitude of the generated magnetic fields can be enhanced by several orders of magnitude with respect to the minimally coupled case, reaching present-day values up to $B_0 \sim 10^{-13}\,\mathrm{G}$ in large-field scenarios, \emph{which appear as the only ones compatible with observational bounds}. Conversely, small-field models yield negligible magnetic amplitudes and appear non-predictive within our non-minimal framework.

SGR J1935+2154 is the unique magnetar so far from which fast radio bursts have been detected. In October 2022, it resumed its burst activity, and we implemented a dedicated target-of-opportunity (ToO) observation on it from Oct. 13th to Nov. 1st, 2022 (about 940 ks in total) with \textit{Insight}-HXMT, while the KM40m radio telescope observed this source for about 1400 hours since Oct. 15th. We searched the LE, ME, and HE data of \textit{Insight}-HXMT in the overlapping observation time windows with the KM40m radio telescope and revealed 60 magnetar X-ray bursts (MXBs), while KM40m only detected 1 radio burst. In particular, we find that there is an X-ray burst on October 21 (denoted as MXB 221021) temporally associated with this radio burst. Interestingly, this association event shows very different morphology from those X-ray and radio association events from this source reported before (e.g., MXB/FRB 200428). Moreover, we systematically analyzed the temporal and spectral properties of the sample of MXBs during this observation and found that % the (radio-associated) MXB 221021 shows some different properties from other MXBs without associated radio bursts. These findings shed new light on the physical mechanisms of X-ray bursts and radio burst emission in magnetars.

In the standard model of cosmology ($Λ$CDM) the first stars, star clusters, and galaxies are expected to have formed in low-mass dark matter halos at high redshifts ($z \sim 6 - 30$). Attempts to predict the properties and abundances of these objects have mainly relied on numerically expensive cosmological simulations, which often lack the sub-parsec resolution needed to resolve compact star clusters and/or neglect potentially important stellar feedback processes. Motivated by this, I introduce Anaxagoras, a detailed analytical ab initio model of starbursts in low-mass halos. The model includes gas cooling, central gas accretion and disk formation, and stellar feedback from direct radiation pressure, Ly$α$ scattering and IR photons, stellar winds, expanding H II regions, and (crudely) supernovae. I apply Anaxagoras to star formation at $z > 6$ in satellite halos of the Milky Way, as well as to Population III (Pop III) star formation in minihalos. For the Milky Way setup, hundreds of galaxies are predicted to form with luminosities, half-mass radii, mass-to-light ratios, and ages in good agreement with the observed local population of Ultra-Faint Dwarfs. Furthermore, at least $\sim 40$ old globular cluster candidates with initial stellar masses $10^5 - 10^6\,M_\odot$ are predicted to form at the centers of low-mass halos. Finally, if Pop III stars are not overly massive ($25\,M_\odot$), between $\sim 1 - 30$ stars could form per minihalo at $z > 20$, increasing to $\sim 10 - 500$ at $z < 15$ as Lyman-Werner feedback delays star formation until halos reach larger masses; if Pop III stars are more massive ($140\,M_\odot$), most minihalos form just a single star.

The Intergalactic Magnetic Field (IGMF), permeating cosmic voids, is thought to be a relic of primordial magnetic fields generated in the early Universe and that gave rise to all astrophysical magnetic fields. While it has escaped direct detection, lower limits on its intensity can be derived by characterising the time-delayed secondary emission initiated when primary very high-energy (VHE) photons from gamma-ray bursts (GRBs) produce lepton pairs that are deflected by the IGMF before generating a secondary gamma-ray flux. Most current studies exclude IGMF values below $10^{-18}\;\mathrm{G}$, however, they are typically performed under idealised conditions. Focusing on the impact of modelling and observational choices, we simulate CTAO observations of GRBs 190114C and 221009A under varying conditions. For GRB 190114C-like sources, we establish a stable lower limit of $2\times10^{-16}\;\mathrm{G}$, robust against most variations in source properties and detection strategies. For more extreme GRB 221009A-like events, we demonstrate that CTAO could probe fields up to at least $10^{-16}\;\mathrm{G}$ under harsh conditions, improving significantly the current IGMF constraints.

The structure of dark matter admixed neutron stars (DANSs) are investigated, adopting a non-annihilating self-interacting fermionic dark matter (DM) model, with a particular focus on the case of the light DM particle mass $m_D \in [10^{-10}, 1]$ GeV. The DANSs become DM-dominated configurations when $m_D <10^{-1}$ GeV, where a compact neutron star core becomes embedded within an extremely large DM halo. It is found that the maximum mass of DANSs is inversely proportional to $m_{ D}$, approximately as $ 0.627 (\mathrm{GeV/} m_{\rm D})^2 ~\mathrm{M_{\odot}}$, which implies that extremely large masses can be achieved for small $m_{\rm D}$. For $m_D \sim5\times10^{-4}$ GeV, the calculated mass and size of the DM halo can be comparable to those of supermassive black holes such as Sgr A*. Our findings hint at a scenario where neutron stars might serve as strong gravitational seeds for such supermassive dark objects.

The luminosities of protostars provide one of the only indirect methods of measuring their masses and mass accretion rates in their earliest stages of evolution. Accurate measurements of protostellar luminosities traditionally requires assembling complete spectral energy distributions (SEDs) from the near-infrared through millimeter wavelengths. In this work, we use published evolutionary radiative transfer models of collapsing protostellar cores to evaluate the extent to which protostellar luminosities can be estimated from a limited number of infrared photometric measurements. We confirm previous results showing a tight correlation (in log-log space) between the luminosity of a protostar and its flux at 70 microns, although we demonstrate that these previous results yield luminosity estimates that are too low by factors of 2-3. We expand this work to additional wavelengths, finding that single wavelengths at 40 - 350 microns provide luminosity estimates with a 1sigma uncertainty of a factor of 3 (0.477 dex of solar luminosities) or lower, with the uncertainty reduced to a factor of 2 (0.301 dex of solar luminosities) or lower at 70 - 160 microns. While the shorter wavelengths observed by JWST (0.6 - 27.9 microns) do not correlate as well with luminosity, we demonstrate that using a single photometric measurement in two different JWST filters simultaneously can result in luminosity estimates that are less uncertain than even the best estimates obtained using a single JWST filter. Using a single photometric measurement in three different JWST filters simultaneously can result in luminosity estimates that are comparable in accuracy to those obtained using single far-infrared photometric flux measurements.

The radiative mechanism of coherent radio emission has remained an enigma since the discovery of pulsars, even the emergence of fast radio bursts (FRBs), which exhibit similarities to the single-pulse behavior of pulsars and have opened a new view for deciphering the long-standing mystery. Besides tremendous efforts in modelling, advanced facilities matter for solving the problem. The authors review the observational breakthroughs from the Five-hundred-meter Aperture Spherical radio Telescope (FAST), which are providing pivotal insights to unravel the underlying physics of pulsars and FRBs. This study offers a novel perspective in the era when pulsars meet FRBs, and further investigations are encouraged to utilize the highly sensitive telescope, the FAST.

We investigate the morphological and physical properties of ring galaxies residing within cosmic voids. Using void catalogs from VoidFinder, ring candidates identified via the Galaxy Zoo 2 decision tree, and morphological classifications from the Buta (2017) CVRHS based catalog, we analyze a sample of 341 void ring galaxies and find a radial preference, with 91.5% located away from the void cores. Morphologically, inner rings and inner pseudorings account for 45.2% of the sample while outer pseudorings are even more common(56.3%) and outer rings appear in 17.9% of galaxies, which points to secular evolution driven by internal dynamics as the primary formation mechanism. Compared to the general void population, our sample ring galaxies are found to be more massive, redder, and have lower specific star formation rates. Subtle gradients in stellar mass and sSFR from void centers to edges also shows a gentle, density-dependent evolutionary progression. Our results show ring galaxies as a distinct, secularly evolved population shaped by the weakest large-scale environmental gradients

In this work we will create the Secular Light Curve (SLC) of exocomet 3I/ATLAS, using the SLC-Methodology (Ferrin 2010-2023). The SLCs give a throve of new information and allow the comparison of exo-comets with comets of our own solar system. We arrive at the following conclusions: The colors of 3I are consistent and lie inside the area of colors of other comets in our solar system. The SLC of this comet exhibits a photometric anomaly, a region from -120 to -45 days before perihelion that we interpreted as an eclipse, suggesting that 3I might also be a binary. At -45 days, the SLC changes abruptly its slope, reaching a maximum absolute magnitude of mV(1,1,α) = 6.8+-0.1. Using reported estimates derived from 97 papers in the arXiv.org depository for the size, dust, H2O, CO2, and CO production rates, we calculate the total mass loss. We use the inverse total mass loss, as a proxy for age. The Mass-Loss Age = 0.16 comet years will be plotted in the horizontal axis of a Comet Evolutionary Diagram (CED) while the number of Remaining Returns defined as RR = r/Δr = 24, will be plotted in the vertical axis of the CED. 3I/ATLAS exocomet lies among the comets of our Oort comet family. We conclude that 3I is a comet of the Oort Cloud, but from a different stellar system. The Evolutionary Diagram presented in this work shows complexity beyond current understanding

In the course of our ongoing survey of faint planetary-nebula nuclei (PNNi), we obtained optical spectroscopy of the central star of the little-studied PN Kohoutek~1-9 (K 1-9). Its spectrum is found to be that of a G-type dwarf with strong absorption features of carbon molecules and s-process elements such as Sr and Ba--a dwarf barium star. K 1-9 thus joins a very small group of PNe with barium-star nuclei. Their likely progenitors are wide binaries in which the primary star reached the thermally pulsing asymptotic-giant-branch (AGB) phase, dredged up C and s-process elements from its interior, and transferred enriched material to the companion through a dense stellar wind. The remnant core is now a hot, optically inconspicuous (pre-)white dwarf, responsible for ionizing the AGB ejecta, and leaving the optical spectrum dominated by the cool barium star. We present deep narrow-band images of K 1-9, obtained by accumulating long exposure times using amateur telescopes. The PN shows a thin-ring morphology, remarkably similar to the "wedding-ring" shapes seen around other members of this class of binary PNNi. The thin ring probably represents material preferentially ejected into the orbital plane of the binary; we note that the PNN is slightly off-center within the ring, as has been predicted theoretically. We suggest several follow-up studies, including precision photometry to search for periodic variations due to starspots on the rotating barium star, and high-resolution spectroscopy to determine atmospheric parameters of the star, chemical abundances, and its rotation velocity.

Numerical studies of core-collapse supernovae have demonstrated the importance of non-radial motions in pre-collapse progenitors on the explosion outcome. We use the CHIMERA neutrino radiation hydrodynamics code running seven two-dimensional simulations of 15 solar mass progenitors with different progenitor structures introduced by different one and two-dimensional pre-collapse stellar evolution environments to examine the impacts of stellar structure and non-spherical motion in the pre-collapse progenitor on the development of explosions in 2D core-collapse supernova simulations. We compare the explosion evolution of these models in terms of shock dynamics, diagnostic energy, neutrino heating, accretion, explosion geometry, nuclear abundances, and turbulent convection. We also analyze how stochasticity impacts our simulations. Contrary to results reported by other groups examining the impacts of multi-dimensional progenitors, we observe similar shock revival times and explosion development in our simulations despite differences in initial compositions and structures. We find no discernible impact from turbulent energy introduced by the multi-D structures in the progenitor as the models evolve from the stalled shock to explosion. We attribute this to the turbulence generated in the post-shock region by shock deformation and standing accretion shock instability to a saturation level before the neutrino-driven convection dominates the post-shock dynamics. An examination of model stochasticity shows that any prior expected impacts on explosive outcome due to convection-related perturbations lie below the detectable threshold of numerical variation.

The recently commissioned Subaru `Ōnohi`ula Prime Focus Spectrograph (PFS) will obtain spectra from nearly 2,400 fibers that cover 1.24 square degrees. The 360 night Subaru Strategic Program for PFS is dedicating approximately one-third of its allocation (130 nights) to study the structure and evolution of galaxies in the Local Group. This Galactic Archaeological survey has three pillars. (1) We will determine whether the mass density profiles of dwarf galaxies are consistent with cusps, as expected for cold dark matter, or cores, as expected from alternative dark matter theories or baryonic feedback. We will deduce the density profiles as a function of radius from modeling of the full line-of-sight velocity and abundance distributions for six dwarf galaxies. Our total sample will consist of 18,000 member stars to beyond the nominal tidal radius of each system. (2) From measurements of the [alpha/Fe] abundance ratio, we will learn the difference in assembly history of the two most massive galaxies in the Local Group: M31 and the Milky Way. We will observe 30,000 member stars over 45 square degrees of M31's halo and outer disk. (3) We will uncover how the most fragile (outer) part of the Milky Way responded to accretion events both in the distant past (such as Gaia-Sausage Enceladus) and in more recent history (such as the Sagittarius dwarf spheroidal galaxy). To support this study, PFS will provide velocities and metallicities--from which, in combination with photometry, we will deduce ages--for tens of thousands of main-sequence stars out to a Galactocentric distance of ~30 kpc.

Foreground mitigation remains a central challenge for 21 cm intensity mapping experiments, which require precise, wideband calibration of telescope beams and gains. We present the Precision Emitter for 21 cm Array Coherent Calibration (PEACC), a digitally synthesized calibration source that generates Gaussian noise across a 1.2 GHz bandwidth, time-synchronized to a 1 pulse-per-second output from a GPS-disciplined oscillator, and optimized for aerial deployment. PEACC uses a dual-source architecture with one unit mounted on an aerial platform and a second reference unit connected directly to the radio data acquisition system; this configuration enables improved sensitivity in the low-SNR regime and direct phase measurement. The system further supports configurable band selection, allowing adaptation to various 21 cm intensity mapping telescopes. We validated PEACC through anechoic chamber measurements and by integrating the source on a drone flown over a local radio dish testbed. In both settings, the correlated channel substantially outperformed the auto-correlation channel across all signal-to-noise regimes of interest, confirming the key advantage of the dual-source architecture. To our knowledge, this is the first published demonstration of a free-space coherent calibration signal synchronized only by clocks, the first deployment of such a source on a drone, and the first published beam measurements made with such a source. Given the growing interest in drone-based calibration for 21 cm arrays, this work establishes the feasibility of high-fidelity digital calibration for next-generation 21 cm instruments, and provides a practical path towards improved foreground control and beam calibration in future arrays.

The supermassive black hole (Sgr A*) in the Galactic center is surrounded by the S-star cluster consisting of young stars on eccentric orbits. Recently, the S-star binary, called D9, was found to be orbited by a circumbinary disk. Due to the gravitational interaction between Sgr A* and the binary, the disk could be short-lived. We investigate the evolution of the disk around a stellar binary while orbiting Sgr A*. We use the \texttt{AMUSE} framework for coupling a gravity solver (for the binary and Sgr. A*) with a hydrodynamics solver (for the disk). We find that, the disk eventually settles between 5.2$a_{\rm in}$ and 0.28 Hill radii of the binary. Here, $a_{\rm in}$ is the semi-major axis of D9. The inclination of the circumbinary disk follows the binary's, which evolves due to the von Zeipel-Lidov-Kozai (vZLK) mechanism induced by Sgr A*. The mean eccentricity of the disk is approximately in anti-phase with the eccentricity evolution of the binary. We find a vZLK timescale of $T_\text{vZLK}\approx62.5\,$kyr, which is two orders of magnitude shorter than the value reported by Peisker etal. (2024). As a consequence, D9 has undergone multiple vZLK oscillations in its lifetime of 2.7 Myr. We find the disk shows periodic bursts of mass loss on the vZLK timescale, suggesting that the mass loss itself is in part driven by the vZLK mechanism. The secular evolution observed in both the binary and the disk are consistent with theoretical predictions. We find the disk loses $\sim$7\% $\pm$ 2\% of its mass every vZLK cycle. If we extrapolate this mass loss, the disk will have 1\% of its current mass left after another $\sim$4 Myr. D9 will then be $\sim$6.7 Myr old, which is on the same order as the current average age of S cluster members. The vZLK-driven mass loss could, therefore, explain the absence of Br$γ$ emission from other S cluster members.

We study non-minimally coupled scalar-induced multi-phase inflation in metric and Palatini gravity, considering linear, Brans-Dicke-like, and Higgs-like sectors. The scalar spectral index lies in the range \( n_s \simeq 0.93 \ \text{--} \ 0.98 \), consistent with \textit{Planck} and combined \textit{Planck}+ACT data. The tensor-to-scalar ratio can reach \( r \sim 0.03 \) in metric, whereas Palatini models generically predict \( r \lesssim 10^{-5} \). In the Palatini case, field excursions remain sub-Planckian, and the perturbative unitarity cutoff is raised. Reheating proceeds via perturbative inflaton decays into Higgs bosons and fermionic dark matter (DM) through the portal coupling \( λ_{12} \) and Yukawa coupling \( y_χ\). Radiative stability of the inflationary plateau constrains the couplings to \( y_χ, λ_{12} \sim 10^{-7} \ \text{--} \ 10^{-3} \), implying \( 4\,\mathrm{MeV} \lesssim T_{\rm rh} \lesssim 10^{15}\,\mathrm{GeV} \). Palatini realizations require smaller couplings and thus a narrower reheating window. Non-thermal DM production $χ$ from inflaton decays is viable for DM mass \( m_χ\sim \mathrm{keV} \ \text{--} \ \mathrm{PeV} \) with \( y_χ\lesssim 10^{-6} \) over large parameter regions. We estimate the inflaton-right-handed neutrino (RHN) Yukawa coupling \( y_N \) required for successful baryogenesis via non-thermal leptogenesis within a Type-I seesaw framework, for the lightest RHN mass \( M_{N_1} \sim 10^{9} \ \text{--} \ 10^{14}\,\mathrm{GeV} \), provided \( M_{N_1} > T_{\rm max} \), where \( T_{\rm max} \) follows from radiatively consistent reheating. In Palatini scenarios, the lower maximal temperature and tighter stability bounds further restrict the leptogenesis parameter space.

Nova-likes such as BZ Cam are high state Cataclysmic Variables showing hard X-ray emission that can be characterized with advective hot flows in the inner accretion disk. We explore Chandra High Energy Transmission Grating (HETG) observations of BZ~Cam for detailed line diagnosis and ionization conditions in the X-ray regime. We mostly find H- and He-like emission lines of Mg, Si, S, and Fe. All He-like line components of forbidden, intercombination and resonance lines are present. The R ratios of selected lines indicate plasma densities of a few $\times$10$^{12-14}$ cm$^{-3}$ and G ratios reveal temperatures (3-6)$\times$ 10$^6$ K where the Fe lines yield (1-3)$\times$ 10$^7$ K. The H to He line ratios and the R and G ratios show that the plasma is in a nonequilibrium ionization condition, which is consistent with our previous X-ray results and the accretion flow in the X-ray region being an ADAF-like (advective) hot flow. Simultaneous fits of the HEG and MEG spectra or the broadband joint spectra of ROSAT, Chandra zero order and NuSTAR yield temperatures 3.4-6.3 keV using a VNEI model of plasma emission (in XSPEC) or Bremsstrahlung emission. An additional power law is detected above 98\% Confidence Level in the broadband analysis. The orbital variations and the broadband spectra show dipping/veiling of the X-rays and an additional warm absorber model with an ionization parameter log($ξ$) = 2.7 is required at the 3$σ$ level, along with the VNEI model where the HEG and MEG simultaneous fits yield the log($ξ$) = 3.6 .

All multipole moments of a Kerr black hole are uniquely determined by its mass and spin. Gravitational wave observations can test this prediction by measuring spin-induced multipole moments imprinted on the inspiral phase of compact binary mergers. In this Letter, we show that the recently reported compact binary coalescence GW241011 enables a simultaneous test of deviations in the spin-induced quadrupole and octupole moments of the binary components from their black hole values. We find no evidence for deviations from the Kerr prediction and place the first constraints on spin-induced octupole moments of the compact binary. This approach complements tests of the Kerr nature of compact binary merger remnants based on quasinormal mode measurements in the ringdown phase.

We present O$-$C (``observed minus calculated'') timing analysis of the quasi-periodic eruption (QPE) source eRO-QPE2 with a multi-mission X-ray campaign, which includes 32 observed eruptions spanning a month (i.e. 325 QPE cycles). In relation to accretion (e.g. disk instability) models, the O-C is consistent with a damped random walk of the QPE recurrence, albeit with highly uncertain parameters. If instead an underlying orbital clock is present, eRO-QPE2 is consistent with a period of $P \sim 2.24$\,h and two hierarchical super-periodic modulations, with periods of $\sim 4.4$\,d ($\sim47$\,P) and $\approx 95$\,d ($\approx 1000$\,P). We found no negative period derivative, with $|\dot{P}| \lesssim 2 \times 10^{-6}$\,s/s at $3σ$. This disfavors high-eccentricity WDs and high-mass/eccentricity IMBHs via GW decay. For disk-collision models, where the $\dot{P}$ from gas drag and the QPE integrated energy provide bounds on the local disk density, a main-sequence star is disfavored as EMRI secondary unless stellar debris streams are present, while stripped stars remain allowed. The correlated odd/even O-C disfavors both disk crossings per orbit being observed. Interpreting the data with one \emph{observed} event per orbit, the short modulation is consistent with apsidal precession for $a \sim 140\,R_g$, $e \approx 0.1$, and $M_{\rm BH} \approx 1.5 \times 10^{5}\,M_\odot$. The longer modulation (much less constrained) is inconsistent with EMRI nodal precession and disk precession is allowed for a limited parameter volume, while there is a solution with a stable hierarchical triple system with an outer massive black hole at $\sim 0.4\,\mathrm{mpc}$ and mass $\sim(0.1-1) \times M_{\rm BH}$. However, no reliable solution can be found with more robust EMRI trajectory models, possibly due to narrow likelihood peaks in a multi-dimensional parameter space with sparse data.

We present panchromatic observations of the Type Ia supernova (SN Ia) 2023qov, ranging from $\sim$2 weeks before to $\sim$1 year after maximum light. \textit{JWST} near- and mid-infrared spectra at $+$276 and $+$363~days show $\sim$400 K dust emission that cools by $\sim$75 K between epochs, the first unambiguous spectroscopic detection of dust emission in a normal SN Ia. We find that the emission is well described by models of carbonaceous dust placed within $\sim$1 light year of the SN, with a dust mass of $\sim$$10^{-4}$ M$_{\odot}$. We do not see evidence of active dust creation, suggesting an infrared light echo by pre-existing circumstellar dust as the likely source of the emission. The \textit{JWST} nebular line profiles suggest asymmetric, stratified ejecta, similar to other normal SNe Ia, though a slight double-horn structure in the argon lines indicate a toroidal enhancement. SN 2023qov exhibits a slightly red, fast-declining early light curve ($Δm_{15}(B) = 1.47 \pm 0.05$ mag), from which we determine a $^{56}$Ni mass of $M_{56} = 0.21 \pm 0.04$ M$_{\odot}$, and a distance of $d = 36.0 \pm 1.8$ Mpc to the SN and its host, NGC 7029.

We model the formation and retention of the most massive black hole (BH) in 47~Tuc using the semi-analytical code \texttt{cBHBd}, coupling cluster evolution with binary BH dynamics and computing merger-remnant masses, spins, and gravitational-wave recoil kicks via numerical-relativity surrogate prescriptions. We evolve 80\,000 cluster realisations spanning initial masses, densities, IMFs, and metallicities, in both a baseline scenario ($m_{\rm max} = 130\,\mathrm{M}_{\odot}$) and an extended-IMF scenario with ${\sim}\,50-110$ primordial BH seeds above the pair-instability gap ($M_{\rm BH} \sim 130-700\,\mathrm{M}_{\odot}$). Selecting models reproducing 47~Tuc's present-day mass and half-mass radius, we find hierarchical mergers alone yield a most massive retained BH of $M_{\rm BH} \sim 45-70\,\mathrm{M}_{\odot}$ with spin $χ_{\rm BH} \sim 0.65$, limited to ${\sim}\,1-3$ mergers, as second-generation remnants acquire spin $χ\sim 0.7$ that amplifies recoil kicks in subsequent generations. When primordial seeds are included, the retained-mass distribution becomes bimodal -- in ${\sim}\,90\%$ of realisations all seeds are ejected, but in ${\sim}\,10\%$ a massive seed ($M_{\rm BH} \gtrsim 450\,\mathrm{M}_{\odot}$) survives -- while the joint mass-spin distribution is trimodal; seeds surviving via stellar-mass BH mergers retain low spin ($χ\lesssim 0.3$), whereas seed-seed mergers produce high-mass, high-spin remnants ($χ\sim 0.65-0.7$), yielding 90th-percentile retained masses of ${\sim}\,500-1100\,\mathrm{M}_{\odot}$. Both scenarios are consistent with the $3σ$ dynamical upper limit of $578\,\mathrm{M}_{\odot}$. Our results favour a dark-remnant subsystem over a single massive IMBH and provide a spin-mass diagnostic testable with LIGO-Virgo-KAGRA, the Einstein Telescope, Cosmic Explorer, and LISA.

In a magnetically-dominated model of star formation, we expect to see alignments between the magnetic field orientation of star-forming dense cores and the cloud-scale magnetic field. Pandhi et al. (2023) showed instead, however, that the orientation of cores and their angular momentum vectors appear random with respect to the larger-scale magnetic field, implying that magnetic fields may play a diminished role in core formation and evolution. Here, we use higher-resolution dust polarization data from the B-Fields In Star-forming Region Observations (BISTRO) survey on the James Clerk Maxwell Telescope (JCMT) to investigate the change in the magnetic field orientation from cloud scales to core scales, and reassess any correlations between core-scale magnetic fields, core orientations and core velocity gradients. We produce a catalog of 79 cores over 14 star-forming regions with averaged core-scale magnetic field orientations. We find that the core-scale magnetic field is more disordered compared to the cloud-scale field, as measured by an increased standard deviation in the magnetic field vector orientations. Alignment between the core-scale and cloud-scale field varies greatly between regions. Our results are consistent with random alignments between the core-scale magnetic field, core orientation, and core velocity gradient, in agreement with the results by Pandhi et al. (2023) for the cloud-scale field. We conclude that there is a clear change in the magnetic field in the transition from cloud- to core-scales. Our results suggest that the magnetic field may not play a dominant role in the evolution of dense cores on core scales.

One channel to form millisecond pulsars with CO white dwarf companions is through the stable Roche-lobe overflow of intermediate-mass ($3\,{\rm M}_\odot\lesssim M\lesssim 5\,{\rm M}_\odot$) stars at the end of the main sequence (Case A) or the beginning of the hydrogen shell burning phase (Case B). We reproduce previous numerical calculations of this channel and supplement them with a simple analytical model that relates the final orbital period $P(M,m_{\rm wd})$ to the white dwarf's mass and to its progenitor's initial mass $M$. We also theoretically calculate for the first time the eccentricity $e$ in this process, which is set by the fluctuating gravitational quadrupole moment of the progenitor's convective envelope during Roche-lobe detachment. Intermediate-mass progenitors detach when their non-degenerate cores ignite helium, in contrast to low-mass ($M\lesssim 2\,{\rm M}_\odot$) stars with degenerate cores that detach when their envelopes become too light to support a burning shell. Despite the order of magnitude higher envelope mass at detachment $m_{\rm e}$ in our case, the eccentricity is barely affected because $e\propto m_{\rm e}^{1/6}$, explaining why intermediate-mass ($m_{\rm wd}\lesssim 0.6\,{\rm M}_\odot)$ CO white dwarfs have similar eccentricities to lower mass helium white dwarfs. Massive CO and ONe white dwarfs ($m_{\rm wd}\gtrsim 0.6\,{\rm M}_\odot)$, on the other hand, probably formed through a different channel of unstable Roche-lobe overflow during helium shell burning (Case C), followed by common envelope inspiral. The measured eccentricities of these massive white dwarfs remain to be explained.

Dust attenuation strongly affects the observed spectral energy distributions of galaxies, introducing significant uncertainties in the derivation of key physical properties such as star formation rates, stellar masses, and metallicities. While attenuation curves have been extensively studied in the local Universe and at intermediate redshift, direct spectroscopic constraints at earlier cosmic epochs have remained limited prior to JWST. We aim to derive the average dust attenuation law of star-forming galaxies over the redshift range 2<z<7. We combine NIRSpec spectroscopy from the JADES survey with deep multi-wavelength photometry from the ASTRODEEP catalogs. Using a mass-selected sample (log(M_\star/M_\odot) > 9) of 120 galaxies with reliable Balmer decrement (Ha/Hb), we construct stacked spectral energy distributions in bins of Balmer optical depth and derive the selective attenuation curve following the empirical methodology introduced by Calzetti et al. (2000). The wavelength coverage is further extended toward the near-infrared using MIRI photometry. The resulting attenuation curve spans the rest-frame range 0.16-1.14mu and is well described by a smooth function. We derive a normalization factor R_V=3.98, finding that the average attenuation law is consistent with the local starburst relation in both slope and normalization. Compared to several determinations at intermediate redshift, however, our curve appears systematically flatter in the ultraviolet. We find no significant evidence for a 2175A UV bump in the average attenuation curve. Our results provide the first empirical determination of the average dust attenuation law for star-forming galaxies at 2<z<7 based on JWST spectroscopy. Despite the diversity of attenuation properties observed in individual systems, the ensemble-average behavior remains consistent with the local starburst relation.

Primordial black holes (PBHs) are one of the most appealing dark matter candidates over a wide range of masses and abundances. This broad parameter space has been constrained by a variety of observational probes. In this work, for the first time, we use data from high-energy neutrino telescopes, like IceCube and ANTARES, to constrain sub-asteroid mass ($\lesssim 10^{18}\,\mathrm{g}$) Schwarzschild PBHs with extended mass functions. We derive limits from the diffuse high-energy neutrino flux produced by the direct evaporation of PBHs, as well as from the transient signatures associated with PBHs passing in the vicinity of the Earth. While our bounds are slightly weaker than existing constraints from gamma-ray observations, they provide an independent and complementary probe based on observational high-energy neutrino data. We further show that future detectors such as IceCube-Gen2 and KM3NeT can significantly improve these constraints, potentially excluding PBHs with masses up to $\sim \mathrm{few} \times 10^{18}\,\mathrm{g}$ composing the entirety of dark matter.

We present a time-resolved analysis of high-resolution spectra of the AGN MCG-6-30-15 obtained by XRISM alongside broadband spectra from NuSTAR and XMM-Newton during a coordinated observing campaign in February 2024. These observations provide some of the most detailed measurements of X-ray reflection from the innermost regions of the accretion disc around a supermassive black hole, and its evolution during periods of significant variability. We find that both the X-ray spectrum and its variability can be described by a self-consistent model of the reflection of the coronal X-ray emission from the accretion disc around a rapidly-spinning (a > 0.93) black hole, in which the observed variability arises from underlying changes in the luminosity, spatial extent and motion of the corona. While the corona is compact, residing within 10rg of the black hole for the majority of the observations, finite spatial extent is required to fully explain the shape of the reflection spectrum. A flare was observed in the X-ray emission during which the corona expanded to around 15rg and was accelerated away from the black hole reaching a velocity of 0.27c. Around the flare were short dips in the observed flux, during which the corona was found to have collapsed to a confined region, within 2.5rg of the black hole, enhancing the relativistic effects observed from the inner accretion disc. We find it is necessary to account for such significant spectral variation in order to obtain accurate measurements of the spin of the black hole via X-ray reflection spectroscopy.

We carry out a multi-spin perturbation-theory study of the four-dimensional Einstein-Skyrme (ES) anti-de Sitter (AdS) black hole (BH), whose lapse $f(r)=1-8πK-2M/r+4πKλ/r^{2}$ inherits two couplings from the hadronic model -- the pion combination $K=F_π^{2}/4$ and the Skyrme coupling $e$ -- with $Kλ=1/e^{2}$ pinned by the theory rather than being a free integration constant. After deriving the Klein-Gordon, Maxwell and Dirac effective potentials on this background, we compute the quasinormal modes (QNMs) with the sixth-order WKB formula and cross-check them against the thirteenth-order Padé-improved expansion and the eikonal limit set by the unstable photon sphere. The first overtone $(n=1)$ of the scalar and electromagnetic channels reveals a mild Konoplya-Zhidenko anomaly: the ratio $|\mathrm{Im}\,ω_{1}|/|\mathrm{Im}\,ω_{0}|$ drifts monotonically from $2.42$ to $2.54$, sitting noticeably below the Schwarzschild value near $3$. The dominant scalar mode is independently reproduced to better than $0.2\%$ by a time-domain Prony fit. Greybody factors for all three spins follow the ordering $T_{\rm EM}<T_{\rm scalar}<T_{\rm Dirac}$. Testing strong cosmic censorship at the Cauchy horizon, we find the Christodoulou parameter $β\lesssim 4\times 10^{-3}$ across the admissible $(K,e)$ window -- more than two orders of magnitude below the threshold $1/2$ -- with the margin protected by the theory itself.

This paper investigates the shadow and polarization images of a Kerr-Sen black hole illuminated by geometrically thick and optically thin accretion disks. We adopt two classes of accretion models, namely the phenomenological radiatively inefficient accretion flow (RIAF) model and the analytical ballistic approximation accretion flow (BAAF) model. Based on radiative transfer theory, we examine the effects of the spin parameter $a$, black hole charge $Q$, and observer inclination angle $θ$ on the shadow images. Both models show that, as the charge $Q$ increases, the photon rings and the central dark regions shrink simultaneously. Meanwhile, frame dragging gives rise to a pronounced brightness asymmetry, which becomes more significant with increasing $a$ and $θ$. The main difference between isotropic and anisotropic radiation is that, in the latter case, the higher order images are brighter in the upper and lower polar regions. For the BAAF model, because the conical approximation renders certain regions geometrically thinner, the spatial extent of the higher order images is narrower than that in the RIAF model, and the separation between the direct image and the higher order images is more distinct. In the polarization images, the spatial distribution of the polarization vector directions is mainly determined by gravitational lensing and frame dragging, whereas the intensity near the photon ring and the scale of the higher order images are significantly influenced by $Q$.