AX·NEWS
Sign in

Artificial Intelligence

Artificial Intelligence news — Friday 09 October 2026

150 stories from 10 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. Antropic 'Callenges', the first ultraviolet full sky map of humanity
    KRAI Times (Korea) Translated from Korean
  2. OpenAI, Researcher Hazardous Disasters..."Safe Raised Non-Security Violation"
    KRAI Times (Korea) Translated from Korean
  3. SpaceX, Wireless Communication...Mi·Europe Communications Co., Ltd.
    KRAI Times (Korea) Translated from Korean
  4. Data Center Professional Pourman, Australia IPO Retractable...'Nvidia Hug's also
    KRAI Times (Korea) Translated from Korean
  5. '; // $('.dropdown list').append('').append(); // $('.dropdown list').append('
    KRAI Times (Korea) Translated from Korean
  6. Apple released the first touchscreen MacBook... Professional Market Summary With M6 Chip
    KRAI Times (Korea) Translated from Korean
  7. Amazon Launches Android Tablets...Premium Market Challenge
    KRAI Times (Korea) Translated from Korean
  8. "Open source vulnerable AI catches". .Antropicture, Free Security Scan Tool Public
    KRAI Times (Korea) Translated from Korean
  9. "Model abuse prohibition". .&tropic, 'Clip Rod' revised the use policy to prevent misuse
    KRAI Times (Korea) Translated from Korean
  10. Can TypeSafe's Jev Make AI Agents Safer Without Another LLM?
    USTowards Data Science

    How a decision-first model can catch risky tool calls before they turn into real-world actions The post Can TypeSafe's Jev Make AI Agents Safer Without Another LLM? appeared first on Towards Data Science .

  11. Impactful scheduling for GPU clusters
    USHugging Face Blog
  12. Where Does the Money Go Across Long-Running Coding Agents?
    USTowards Data Science

    The cost of control The post Where Does the Money Go Across Long-Running Coding Agents? appeared first on Towards Data Science .

  13. OpenAI Hwago Research Institute “AI Churon monitoring function must be preserved”
    KRAI Times (Korea) Translated from Korean
  14. Build Your First AI Agent with One Tool Call
    USTowards Data Science

    AI Agents for Beginners, Part 1 The post Build Your First AI Agent with One Tool Call appeared first on Towards Data Science .

  15. The Download: AI’s refusal problem and weight-loss drug side effects
    USMIT Technology Review Biotech

    This is today’s edition of The Download , our weekday newsletter that provides a daily dose of what’s going on in the world of technology. We’re putting too much faith in AI’s ability to say no Today’s AI models are trained to refuse a…

  16. Awarded the decoration on Trump, Big Tech.. . "Version and Safety Press"
    KRAI Times (Korea) Translated from Korean
  17. United States, foreign workers are strengthened...MS·Advertisement of foreign residents
    KRAI Times (Korea) Translated from Korean
  18. OpenAI, Annual Sales Estimated 27 trillion Descent...BikTech's first-ever
    KRAI Times (Korea) Translated from Korean
  19. Stop Using AI. Start Hiring It.
    USTowards Data Science

    AI agents can now write more code than any of us can read. That changed how I think about AI. The post Stop Using AI. Start Hiring It. appeared first on Towards Data Science .

  20. Google Launches ‘Jeminai Agent’ for Business... ‘Argon’· ‘Chill Road’
    KRAI Times (Korea) Translated from Korean
  21. Korean financial rights attacked Chinese 26-year-old hacker...'Holding Shinwon exposure'
    KRAI Times (Korea) Translated from Korean
  22. Japan's solar support system is...Intensive of public areas, carports and infra
    KRAI Times (Korea) Translated from Korean
  23. Autumn, 23~24 days Reading Culture Festival...External reading and island experience operation
    KRAI Times (Korea) Translated from Korean
  24. Copyright (c) 2015 SHINSEGAE. All Rights Reserved.
    KRAI Times (Korea) Translated from Korean
  25. Incheon’s first ‘Sunlight Souk Village’...2MW Water Solar
    KRAI Times (Korea) Translated from Korean
  26. Korean financial rights attacked Chinese 26-year-old hacker...'Holding Shinwon exposure'
    KRAI Times (Korea) Translated from Korean
  27. We’re putting too much faith in AI’s ability to say no
    USMIT Technology Review Biotech

    Ever since people first seriously contemplated giving machines an intelligence modeled on our own, there has never been any question that they would, like us, be able to say no. The sci-fi canon is full of stories of robotic disobedience…

  28. Job titles of the future: Delivery drone air traffic controller
    USMIT Technology Review Biotech

    The moment Trevor Wischnewsky heard that drones were delivering pizza and sushi around his Texas neighborhood, his mind was made up. “It was super fascinating,” he says. “I just immediately wanted to be a part of it.” Wischnewsky is now an…

  29. We’re still figuring out the side effects of GLP-1 weight-loss drugs
    USMIT Technology Review Biotech

    This week my colleague Antonio Regalado had an interesting update on GLP-1 weight-loss drugs. According to research presented at an aging meeting in Boston, these drugs seem to affect at least some measures of biological age . Overweight…

  30. [AIReader's Book] INER Prompt
    KRAI Times (Korea) Translated from Korean
  31. Japanese Scientist Kenso Soai Wins Chemistry Nobel Prize With French Scientist Henri B. Kagan
    SGAsian Scientist

    The researchers won the award for solving a long standing mystery in chemistry: homochirality.

  32. Sophos cuts threat investigation time by 96% with OpenAI Daybreak
    USOpenAI Blog

    Discover how Sophos uses OpenAI’s Daybreak to cut cyber-threat investigation time by 96% and automate 52% of MDR cases while preserving human oversight.

  33. Asana cuts model costs 76x in browser tests with GPT-6.1 Sol
    USOpenAI Blog

    Using GPT-6 Astra in Codex, Asana made its browser agent 76x cheaper and 5x faster in tests to offer customers more capable models.

  34. [Oct. 8] “Achievement of AI in 4000 Issues of Mathematics”... Change of the scientific research shown by OpenAI
    KRAI Times (Korea) Translated from Korean
  35. Diffu-LoRA: A Novel Low-Rank Adaptation for Personalized Diffusion Models
    USarXiv cs.CL

    arXiv:2610.10550v1 Announce Type: new Abstract: Personalizing text-to-image diffusion models from a few reference images requires preserving subject identity while following prompts that describe new contexts. Full-model fine-tuning is…

  36. Large Language Model-Assisted Preparation of Transportation Management Plans: A Case Study with WisDOT WisTMP System
    USarXiv cs.CL

    arXiv:2610.10650v1 Announce Type: new Abstract: Work zones are critical yet hazardous components of transportation infrastructure, requiring carefully designed Transportation Management Plans (TMPs) to ensure safety and mobility. However…

  37. Stochastic Teacher Intervention for Agentic On-Policy Distillation
    USarXiv cs.CL

    arXiv:2610.10878v1 Announce Type: new Abstract: On-policy distillation (OPD) efficiently transfers capabilities from a stronger teacher to a student language model through dense token-level supervision on student-generated rollouts and has…

  38. Clinician use of language models diverges from how the models are evaluated
    USarXiv cs.CL

    arXiv:2610.11069v1 Announce Type: new Abstract: Large language model (LLM) assistants are being deployed to clinicians across health systems, and judgments about their readiness rest largely on benchmark scores, most of them derived from…

  39. Can a System-One LLM Perform Knowledge Tracing When Few or No Learners Are Logged?
    USarXiv cs.CL

    arXiv:2610.11135v1 Announce Type: new Abstract: Knowledge tracing (KT) models need many logged learners, so a new course or platform starts without a usable model. In LLM-based KT the LLM generates the answer, which we call System-Two; it…

  40. The "10th Juror": Open-Set Standpoint Screening for Bureaucratic Bias Detection
    USarXiv cs.CL

    arXiv:2610.11136v1 Announce Type: new Abstract: Presupposing the boundaries of bias is itself a form of bias. We study closed-loop bias governance for Dutch government documents, where a system must detect biased language, ground decisions…

  41. Learning the Loop, Not Just the Page: Execution-Grounded Loop Learning for Web Generation
    USarXiv cs.CL

    arXiv:2610.11543v1 Announce Type: new Abstract: Functional Web generation is increasingly optimized with executable rewards, yet existing methods largely focus on the quality of the final page and leave the process of diagnosing and…

  42. Phonologically Informed Tokenization for German Speech Recognition: A Cross-Domain Study
    USarXiv cs.CL

    arXiv:2610.11646v1 Announce Type: new Abstract: German is a morphologically rich language whose syllable structure is exceptionally well-predicted by the Knuth--Liang hyphenation algorithm. We ask whether phonologically informed…

  43. From Sparse Representations to Behavioral Insights for Multimodal Depression Assessment
    USarXiv cs.CL

    arXiv:2610.11787v1 Announce Type: new Abstract: Multimodal depression assessment offers a promising approach to analyzing behavioral patterns associated with depression. However, existing methods often rely on dense and opaque multimodal…

  44. When History Helps and Hurts: Selective History Use across Multimodal Turns
    USarXiv cs.CL

    arXiv:2610.11948v1 Announce Type: new Abstract: Reliable multimodal interaction depends on selective use of conversational history: an earlier question may remain relevant while its previous answer is outdated, whereas a current request…

  45. Language-Specific Effects of Tokenizer Choice in Multilingual Language Models
    USarXiv cs.CL

    arXiv:2610.12144v1 Announce Type: new Abstract: Tokenizer choice affects multilingual language modeling, but vocabulary capacity is finite and vocabulary size is often constrained: improving representation for some languages often comes at…

  46. HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments
    USarXiv cs.CL

    arXiv:2610.12274v1 Announce Type: new Abstract: Text-to-SQL models are commonly trained to map questions directly to static queries, whereas real-world database agents operate through stateful, multi-turn interaction with live databases …

  47. Beyond Speech Captions: Speech-Rewarded Style Planning for Conversational Text-to-Speech
    USarXiv cs.CL

    arXiv:2610.11461v1 Announce Type: cross Abstract: Natural-language style descriptions provide an interpretable interface between large language models (LLMs) and controllable text-to-speech (TTS). However, using descriptions as…

  48. Chronos Enables Code Agents to Reason over Software Evolution
    USarXiv cs.CL

    arXiv:2610.11578v1 Announce Type: cross Abstract: Historical pull requests record the design decisions, compatibility constraints, and implementation patterns behind a codebase's current state. Experience relevant to a new task can span…

  49. RouterInterp: Understanding Superposed Specialisation in Mixture of Experts Routing
    USarXiv cs.CL

    arXiv:2610.11775v1 Announce Type: cross Abstract: Sparse Mixture of Experts (MoE) models scale more efficiently than dense models by routing tokens to modular expert networks that are only active for processing a fraction of tokens. A…

  50. MindFlow: Mind Supernet Powered Thinking Flows for Research Idea Innovation
    USarXiv cs.CL

    arXiv:2610.11966v1 Announce Type: cross Abstract: Research idea innovation is a fundamental engine of scientific progress, yet it remains difficult to generate and evaluate in a scalable and controllable way. This challenge lies in its…

  51. Overcoming Prior Barriers: Supervised Fine-Tuning under Long-Tail Distribution
    USarXiv cs.CL

    arXiv:2610.12345v1 Announce Type: cross Abstract: Supervised fine-tuning (SFT) adapts pretrained large language models (LLMs) to downstream tasks, but the required concepts can receive substantially different levels of pretrained support…

  52. JevOut: Natural Context Can Flip Decision Models
    USarXiv cs.CL

    arXiv:2609.30243v2 Announce Type: replace Abstract: An ordinary-looking background detail can turn a correct model decision into a confident mistake. We demonstrate this fragility in four decision systems, including Jev, across seven…

  53. Single-Pass Uncertainty Heads for Claim-Level Hallucination Detection in Persian Medical Language Models
    USarXiv cs.CL

    arXiv:2610.03482v3 Announce Type: replace Abstract: Hallucination detection is particularly important for medical language models, but repeated-sampling approaches are computationally expensive. A faster alternative is a single-pass…

  54. SkillContrast: Difference-Guided Text Selection for Agent Skill Reranking
    USarXiv cs.AI

    arXiv:2610.11650v1 Announce Type: cross Abstract: Similar agent skills can share instructions but differ in their conditions of use. Query-based text selection may retain shared instructions and omit these distinctions. We introduce…

  55. GPF-Net: Gated Progressive Fusion Learning for Polyp Re-Identification
    USarXiv cs.AI

    arXiv:2512.21476v2 Announce Type: replace-cross Abstract: Colonoscopic Polyp Re-Identification (ReID) aims to match the same polyp across a large gallery of images captured from different viewpoints and with different cameras, playing a…

  56. Efficient Dense Crowd Trajectory Prediction Via Dynamic Clustering
    USarXiv cs.CV

    arXiv:2603.18166v2 Announce Type: replace-cross Abstract: Crowd trajectory prediction plays a crucial role in public safety and management, where it can help prevent disasters such as stampedes. Recent works address the problem by…

  57. Easy to anticipate, hard to compute: boundary dependence finds the computed outputs that entropy patching misses
    USarXiv cs.CL

    arXiv:2610.11790v1 Announce Type: new Abstract: Byte-level language models such as the Byte Latent Transformer (BLT) group bytes into patches and run their large global model once per patch. BLT starts a patch where a small model's…

  58. ActiveMedAgent: Cost-Aware Trajectory Learning for Multimodal Medical Diagnosis
    USarXiv cs.LG

    arXiv:2610.11140v1 Announce Type: new Abstract: Clinical diagnosis is inherently sequential: clinicians escalate from cheap to costly tests only when additional evidence is expected to resolve diagnostic uncertainty. We present…

  59. Scaling Verifiable Environments for Long-horizon Work Agents
    USarXiv cs.CL

    arXiv:2610.04906v2 Announce Type: replace Abstract: Work agents operate over digital artifacts to execute professional knowledge-intensive work, requiring training environments that support long-horizon interaction and trustworthy…

  60. Grokking or Glitching? How Low-Precision Drives Slingshot Loss Spikes
    USarXiv cs.CL

    arXiv:2605.06152v4 Announce Type: replace-cross Abstract: Deep neural networks exhibit periodic loss spikes during unregularized long-term training, a phenomenon known as the "Slingshot Mechanism." Existing work usually attributes this to…

  61. The Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth Probes
    USarXiv cs.CL

    arXiv:2609.10739v2 Announce Type: replace-cross Abstract: Linear probes that decode the truth from a language model's activations have been proposed as deception monitors, and a truth probe scoring below chance on a model trained to…

  62. Budgeted Cache Repair for Cross-Context KV-Cache Reuse
    USarXiv cs.CL

    arXiv:2610.02233v2 Announce Type: replace-cross Abstract: Cross-context KV-cache reuse predicts a shared segment's keys and values under a new prefix instead of recomputing them, and has been reported to do so without quality loss. We find…

  63. Towards Scalable Meta-Learning of near-optimal Interpretable Models via Synthetic Model Generations
    USarXiv cs.AI

    arXiv:2511.04000v2 Announce Type: replace-cross Abstract: Decision trees are widely used in high-stakes fields like finance and healthcare due to their interpretability. This work introduces an efficient, scalable method for generating…

  64. Beyond Distributional Fidelity: Causal-Penalized Diffusion for Synthetic Tabular Data
    USarXiv cs.LG

    arXiv:2610.11407v1 Announce Type: cross Abstract: Synthetic tabular generators are commonly optimized for distributional fidelity, but statistical similarity alone does not guarantee preservation of causal effects. In this paper, we study…

  65. SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference
    USarXiv cs.AI

    arXiv:2604.21231v3 Announce Type: replace-cross Abstract: Efficient inference for on-device Large Language Models (LLMs) remains challenging due to limited hardware resources and the high cost of the prefill stage, which processes the full…

  66. Risk-Conditioned Fine-Tuning of Large Language Models
    USarXiv cs.LG

    arXiv:2609.08064v3 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly deployed in settings where rare but severe harmful generations can have significant consequences. Existing Risk-Averse RLHF addresses this…

  67. Improving Image-Based Nutrition Estimation Through Multimodal Food-Item Verification and Recovery
    USarXiv cs.CV

    arXiv:2610.11144v1 Announce Type: new Abstract: Single-image nutrition estimation can fail silently when visible foods are missed. Even when a food is correctly identified, its proposed region may not support portion estimation. We propose…

  68. A Unified Score Matching Paradigm for Video Anomaly Detection and Anticipation
    USarXiv cs.CV

    arXiv:2610.11149v1 Announce Type: new Abstract: Video anomaly detection (VAD) is a fundamental and safety-critical task in computer vision. Recent generative approaches detect anomalies from a distributional perspective, but remain limited…

  69. Pose-Free Feed-Forward 3D Inpainting via Learnable Mask Attention and Support Token Refinement
    USarXiv cs.CV

    arXiv:2610.11857v1 Announce Type: new Abstract: 3D scene inpainting aims to recover missing or occluded regions in edited 3D scenes, while ensuring geometric and textural consistency. Existing approaches, however, typically require…

  70. Bernoulli Flow Models: Self-Consistent Generative Modeling for Binary Data
    USarXiv cs.CV

    arXiv:2610.11362v1 Announce Type: cross Abstract: Binary diffusion models typically require a large number of function evaluations (NFEs) to generate high-quality samples, making practical inference computationally expensive. Reducing NFEs…

  71. Optimally Pacing Budget Spending and Learning
    USarXiv cs.LG

    arXiv:2610.11074v1 Announce Type: new Abstract: We establish near-optimal regret bounds for budget-constrained online learning against arbitrary classes of budget-pacing experts in the adversarial setting. In particular, given any class of…

  72. Best of Both Worlds in Federated LSA: Speedup When Possible, Personalization Always
    USarXiv cs.LG

    arXiv:2610.11555v1 Announce Type: new Abstract: We study personalized federated linear stochastic approximation (LSA), a framework which notably encompass personalized temporal difference learning. In this setting, heterogeneous agents…

  73. Self-sufficient Independent Component Analysis for Demixing Flows
    USarXiv cs.LG

    arXiv:2512.00665v2 Announce Type: replace-cross Abstract: We study the problem of learning disentangled signals from data using non-linear Independent Component Analysis (ICA). Motivated by advances in self-supervised learning, we propose…

  74. Velocity Scaling in Flow Matching
    USarXiv cs.CV

    arXiv:2610.10823v1 Announce Type: new Abstract: Scaling a learned flow-matching velocity field $v_\theta$ by a gain $\gamma(t)$ was recently shown to greatly improve generation quality. Prior work argued that velocity fields trained with…

  75. Cognitive Thermometers: Machine Learning and Logical Complexity
    USarXiv cs.CL

    arXiv:2610.10724v1 Announce Type: new Abstract: How does the human mind represent semantic categories? Why do natural languages favor certain meanings over others? Prior explanations have relied on logical definability and complexity, but…

  76. SkillGraph: Skill-Augmented Reinforcement Learning for Agents via Evolving Skill Graphs
    USarXiv cs.CL

    arXiv:2605.12039v2 Announce Type: replace Abstract: Skill libraries enable large language model agents to reuse experience from past interactions, but most existing libraries store skills as isolated entries and retrieve them only by…

  77. Uncertainty-Aware Budget Allocation for Adaptive Test-Time Reasoning
    USarXiv cs.CL

    arXiv:2605.26849v2 Announce Type: replace Abstract: Sampling multiple responses improves language model reasoning, but uniform compute allocation is inefficient because easy questions are over-sampled while hard questions remain…

  78. Measuring and Mitigating Solution Mode Collapse in RLVR
    USarXiv cs.LG

    arXiv:2610.11064v1 Announce Type: new Abstract: A language model (LM) can usually answer the same question in more than one way, but reinforcement learning with verifiable rewards (RLVR) is indifferent to which correct answer a model…

  79. Towards Scalable Meta-Learning of near-optimal Interpretable Models via Synthetic Model Generations
    USarXiv cs.LG

    arXiv:2511.04000v2 Announce Type: replace Abstract: Decision trees are widely used in high-stakes fields like finance and healthcare due to their interpretability. This work introduces an efficient, scalable method for generating synthetic…

  80. LadderEdit: Edit-Level Residual Compression for Memory-Efficient Lifelong Editing of LLMs
    USarXiv cs.LG

    arXiv:2610.11160v1 Announce Type: cross Abstract: Lifelong editing of LLMs requires storing thousands of edits after acquisition. A widely used family of approaches attaches one LoRA adapter per edit, which preserves behavior but grows…

  81. Self-Supervised Keyframe Discovery for Horizon-Invariant Behavior Cloning
    USarXiv cs.AI

    arXiv:2610.10857v1 Announce Type: new Abstract: Behavior cloning (BC) in non-Markovian environments is a challenging problem because policies have to reason over contextual information over long horizons. Existing policy architectures rely…

  82. The Distillation Game: Adaptive Evaluations ＆ Efficient Defenses
    USarXiv cs.AI

    arXiv:2605.22737v4 Announce Type: replace-cross Abstract: Distillation attacks create a deployment trade-off for model providers: the same outputs that make a model more useful can also make it easier to imitate. We study this trade-off…

  83. SP-DocReader: Difference-Aware Self-Play for Precise Document OCR
    USarXiv cs.AI

    arXiv:2610.11148v1 Announce Type: cross Abstract: Accurate page transcription remains difficult for vision language models under limited input and training budgets. We present SP-DocReader, a self-play framework for optical character…

  84. Finsler Flow Matching: Dynamics-Aware Geodesic Interpolation for Single-Snapshot Trajectory Inference
    USarXiv cs.AI

    arXiv:2610.11318v1 Announce Type: new Abstract: Single-cell snapshot data can resolve a continuum of cellular states but do not uniquely determine the dynamics governing transitions between them. However, additional dynamical information…

  85. RL-ARC: Calibrating Large Reasoning Models via Reasoning-guided Uncertainty
    USarXiv cs.AI

    arXiv:2610.11352v1 Announce Type: new Abstract: Language models (LMs) are commonly trained with Reinforcement Learning with Verifiable Rewards (RLVR) to enhance their reasoning capabilities. However, since RLVR does not explicitly account…

  86. Onboard Marine Anomaly Detection on $\Phi$sat-2: From Simulation-Based Development to In-Orbit Demonstration
    USarXiv cs.AI

    arXiv:2610.11735v1 Announce Type: new Abstract: Onboard Artificial Intelligence can improve responsiveness and bandwidth efficiency of Earth Observation systems by processing data directly on the satellite. This paper presents the…

  87. When Has a Bayesian Neural Network Sampled Enough? Adaptive Inference Time with Statistical Guarantees
    USarXiv cs.AI

    arXiv:2610.12212v1 Announce Type: new Abstract: Bayesian neural network predictions are commonly approximated using a fixed number of Monte Carlo samples per input, without controlling the resulting error that comes from this finite…

  88. Agent4RE: A Self-Refining Multi-agent Framework for End-to-End Software Requirements Engineering and Benchmarking
    USarXiv cs.AI

    arXiv:2610.10628v1 Announce Type: cross Abstract: Existing LLM-based approaches for software Requirements Engineering (RE) typically rely on basic prompting strategies or rudimentary agent collaboration, under-utilizing the full potential…

  89. Coverage-Aware Reasoning with Medical Tokens for Diagnosis Prediction
    USarXiv cs.AI

    arXiv:2610.10641v1 Announce Type: cross Abstract: Large language models (LLMs) offer promising potential for next-visit diagnosis prediction, owing to their ability to integrate longitudinal clinical evidence and reason over it in natural…

  90. Higher-Order Action Supervision Makes A Strong Policy Class
    USarXiv cs.AI

    arXiv:2610.11175v1 Announce Type: cross Abstract: Modern data-driven decision-making methods, such as imitation learning (IL) and reinforcement learning (RL), have achieved great success in solving many complex tasks. However, these…

  91. From Geometry to Generalization: Why Row Normalization Can Beat Adam and Muon
    USarXiv cs.AI

    arXiv:2610.11309v1 Announce Type: cross Abstract: Different optimizers can fit the same training data while selecting classifiers with substantially different geometries, but whether this difference provably affects population performance…

  92. RoboAware: Learning to Coordinate Embodied Skills from Counterfactual Outcomes
    USarXiv cs.AI

    arXiv:2610.11480v1 Announce Type: cross Abstract: Embodied coding agents can combine modular robot skills with frozen end-to-end policies, yet effective composition requires anticipating which policy family will succeed in the current…

  93. Adapting English Quality Classifiers for Multilingual LLM Pretraining Data Selection
    USarXiv cs.AI

    arXiv:2610.11585v1 Announce Type: cross Abstract: Recent advances in large language model (LLM) pretraining highlight the role of high-quality training data in improving performance. While model-based filtering has proven effective in…

  94. Same Outcome, Different Evidence: Intent Recovery in LLM Safety Evaluation
    USarXiv cs.AI

    arXiv:2610.11766v1 Announce Type: cross Abstract: Safety evaluations of large language models commonly summarize harmful-output behavior with attack success rate (ASR). Yet the same non-harmful outcome can arise for very different reasons…

  95. Evaluating Exact Output and Checkpoint-State Prediction in Real Programs
    USarXiv cs.AI

    arXiv:2610.11889v1 Announce Type: cross Abstract: We present a benchmark for predicting final output and checkpoint state from source and input alone. It extends CRUXEval-style output prediction with paired shorter- and longer-trace inputs…

  96. Humanoid World Action Model With Joint State--Action Generation
    USarXiv cs.AI

    arXiv:2610.12026v1 Announce Type: cross Abstract: Humanoid robots are a promising platform for general-purpose manipulation. Recent Vision-Language-Action (VLA) policies learn actions directly from multimodal observations, while World…

  97. EgoBench: An Interactive Egocentric Multimodal Benchmark for Tool-Using Agents
    USarXiv cs.AI

    arXiv:2605.27820v2 Announce Type: replace Abstract: As AI agents increasingly operate in open, real-world environments, they require a deep synergy of multimodal perception, tool invocation with multi-hop reasoning, and dynamic interaction…

  98. Multi2AV-Safety: Benchmarking Safety in Multimodal-to-Audio-Video Generation
    USarXiv cs.AI

    arXiv:2608.26535v2 Announce Type: replace Abstract: Recent audio-video generators increasingly support joint conditioning on text, images, audio, and video. These capabilities also enable attacks that exploit cross-modal interactions or…

  99. A Score Is Not a Policy: Measuring the Value of Adaptive Revision
    USarXiv cs.AI

    arXiv:2609.00874v2 Announce Type: replace Abstract: As agentic systems become compound systems, increasingly important decisions move above task execution itself: when should a higher-level controller preserve the strategy guiding another…

  100. Empty Commitments: When Agents Promise What They Cannot Deliver
    USarXiv cs.AI

    arXiv:2610.01045v2 Announce Type: replace Abstract: A chatbot that says "I will remind you tomorrow" will not run again until the user writes. We call such a promise an empty commitment: a promise of action after the current turn that…

  101. Multi-Task Evolution for Zero-Shot Cross-Problem Generalization using LLMs
    USarXiv cs.AI

    arXiv:2610.03316v2 Announce Type: replace Abstract: Designing effective heuristics for diverse combinatorial optimization problems requires substantial expertise and repeated search. Large language models (LLMs) automate heuristic…

  102. AgentTime: Can Agents Estimate and Control Their Own Runtime?
    USarXiv cs.AI

    arXiv:2610.09944v2 Announce Type: replace Abstract: An essential control of AI agents is their ability to manage runtime. This ability requires a sense of time-awareness, to predict and estimate wall-clock time and to control their own…

  103. BEAT: Balanced Frequency Adaptive Tuning for Long-Term Time-Series Forecasting
    USarXiv cs.AI

    arXiv:2501.19065v3 Announce Type: replace-cross Abstract: Long-term time-series forecasting supports a wide range of applications, including weather prediction and electricity demand planning. Frequency-domain methods address this task by…

  104. ExecVLA: Following Fine-Grained Execution Constraints in Vision-Language-Action Models with Bi-Level Action Representation
    USarXiv cs.AI

    arXiv:2603.17524v2 Announce Type: replace-cross Abstract: We study fine-grained execution-constraint following in vision-language-action (VLA) models. Given an invariant task goal, the policy must follow instruction-specified execution…

  105. AI Appeals Processor: A Deep Learning Approach to Automated Classification of Citizen Appeals in Government Services
    USarXiv cs.AI

    arXiv:2604.03672v2 Announce Type: replace-cross Abstract: Government agencies must register, classify and route every citizen appeal within statutory time limits, and much of this work is still done by hand. We describe AI Appeals…

  106. SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild
    USarXiv cs.AI

    arXiv:2605.07604v2 Announce Type: replace-cross Abstract: 3D animal reconstruction in the wild remains challenging due to large species variation, frequent occlusions, and the prevalence of multi-animal scenes, while existing methods…

  107. DCVD: Dual-Channel Cross-Modal Fusion for Joint Vulnerability Detection and Localization
    USarXiv cs.AI

    arXiv:2605.11015v2 Announce Type: replace-cross Abstract: Software vulnerability detection plays a critical role in ensuring system security, where real-world auditing requires not only determining whether a function is vulnerable but also…

  108. Beyond Accuracy: Robustness, Interpretability and Expressiveness of EEG Foundation Models
    USarXiv cs.AI

    arXiv:2605.17562v2 Announce Type: replace-cross Abstract: EEG foundation models (EEG-FMs) have been evaluated predominantly on clean, in-distribution accuracy, demonstrating modest gains over supervised baselines and weak frozen…

  109. VISTA: An End-to-End Benchmark for Visual Spec-to-Web-App Coding Agents
    USarXiv cs.AI

    arXiv:2605.26144v4 Announce Type: replace-cross Abstract: Coding agents can now build applications from design mockups, but a screen that looks right is not an application that works. Existing benchmarks either score visual reconstruction…

  110. Meddies-PII: A Multilingual Framework for Personally Identifiable Information Extraction in Clinical De-identification
    USarXiv cs.AI

    arXiv:2609.12544v2 Announce Type: replace-cross Abstract: Clinical de-identification relies on accurately identifying personally identifiable information (PII). However, manually annotated datasets are costly to construct, while existing…

  111. MA-JEPA: Joint-Embedding World Models for Multi-Agent Reinforcement Learning
    USarXiv cs.AI

    arXiv:2609.33563v2 Announce Type: replace-cross Abstract: World models improve sample efficiency by training policies on imagined trajectories, but their usefulness depends on learning representations that capture the information needed…

  112. How Do LLMs Change Predictions Under Negation?
    USarXiv cs.AI

    arXiv:2610.09571v2 Announce Type: replace-cross Abstract: Negation is an essential feature of human language, yet large language models (LLMs) remain unreliable in processing it. We evaluate recent open-source and closed-source LLMs on our…

  113. HRIL: Learning Multimodal Synergy via Higher-Order Tensor Modeling
    USarXiv cs.AI

    arXiv:2610.12393v1 Announce Type: new Abstract: Self-supervised multimodal representation learning has achieved remarkable success across diverse domains, yet capturing synergistic information remains challenging due to the complexity of…

  114. Prior or Feedback? What an LLM Uses When Adapting Neural Operators
    USarXiv cs.LG

    arXiv:2610.12325v1 Announce Type: cross Abstract: Do LLM scientific agents rely only on their initial task context, or do they adapt their decisions in response to experimental feedback? We study this question in neural operator…

  115. Plan-and-Patch: Diffusion Language Models for Agentic Planning
    USarXiv cs.AI

    arXiv:2610.10786v1 Announce Type: new Abstract: Planning is increasingly important for long-horizon agents, where successful execution requires coordinating subgoals, tool use, and intermediate outcomes over many steps. Yet assumptions…

  116. How Narrative Wrapping Affects LLM Refusal: A Cross-Language Benchmark and Defense
    USarXiv cs.AI

    arXiv:2610.11005v1 Announce Type: new Abstract: Safety-aligned language models often refuse a harmful request stated directly but answer the same request inside a role-play or narrative wrapper. We measure this vulnerability across…

  117. DuplexAgent-RSI: Recursive Harness Improvement for Full-Duplex Voice Agent Collaboration
    USarXiv cs.AI

    arXiv:2610.11299v1 Announce Type: new Abstract: Voice agents are converging on a collaboration pattern: a full-duplex interaction model stays on the live channel as the entry to the conversation, while search, reasoning, and coding are…

  118. When Should Agents Think? Adaptive Reasoning via Cross-Turn Estimation
    USarXiv cs.AI

    arXiv:2610.12061v1 Announce Type: new Abstract: Large language model (LLM)-based agents have demonstrated strong capabilities on complex tasks. They typically perform reasoning before each action throughout an interaction trajectory…

  119. OnTrack: Real-Time Monitoring and Intervention in LLM Agent Trajectories via Streaming Structure-Aware Optimal Transport
    USarXiv cs.CL

    arXiv:2610.12375v1 Announce Type: cross Abstract: Agents are deployed in applications from trip planners and stock trading to IT incident triage. In most cases, LLM agents work autonomously with minimal rule-based safeguarding, leading to…

  120. Internalizer: Portable Context-to-Parameter Mapping for Very Large Language Models
    USarXiv cs.AI

    arXiv:2610.11715v1 Announce Type: new Abstract: Hypernetworks that map a context directly to a LoRA adapter let a large language model carry that context in its weights, but prior work has demonstrated them only on base models of up to 14…

  121. DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation
    USarXiv cs.AI

    arXiv:2610.11659v1 Announce Type: cross Abstract: On-policy distillation (OPD) supervises student-generated trajectories with token-level teacher signals. Its sampled-token variant avoids the cost of full-vocabulary probabilities. Yet we…

  122. Bi-FORK: Generative Modeling of High-Dimensional Bifurcating Systems
    USarXiv cs.LG

    arXiv:2610.12449v1 Announce Type: new Abstract: Bifurcations are ubiquitous in physical systems, from structural buckling to fluid and climate dynamics, yet they remain largely unexplored in deep learning. At a symmetry-breaking…

  123. A Survey on Archetypal Analysis
    USarXiv cs.LG

    arXiv:2504.12392v3 Announce Type: replace-cross Abstract: Archetypal analysis (AA) was originally proposed in 1994 by Adele Cutler and Leo Breiman as a computational procedure for extracting distinct aspects, so-called archetypes, from…

  124. Evidence-Traceable Dynamic Interviewer Architecture for Expertise-Adaptive Qualitative Interviews Using Local LLMs
    USarXiv cs.AI

    arXiv:2610.11651v1 Announce Type: new Abstract: Automated interviewers and conversational agents are increasingly used in research, recruitment, customer service, and education. However, many existing systems rely on fixed question…

  125. Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized Deception
    USarXiv cs.AI

    arXiv:2610.12445v1 Announce Type: cross Abstract: Recent incidents have highlighted the challenge of monitoring LLM agents and the danger of models deceiving people. We show that white-box deception detection via probes can be scaled up to…

  126. Soft Tournament Equilibrium: Differentiable Set-Valued Inference for Non-Transitive Pairwise Comparisons
    USarXiv cs.AI

    arXiv:2604.04328v4 Announce Type: replace Abstract: Soft Tournament Equilibrium (STE) is a differentiable layer for Top-Cycle (TC) and Uncovered-Set (UC) inference from reciprocal pair probabilities. Normalized log-sum-exp reachability and…

  127. LTBD: Learnable Trust-Boundary Delimiters for Prompt Injection Defense
    USarXiv cs.AI

    arXiv:2610.11634v1 Announce Type: cross Abstract: Large language models (LLMs) perform remarkably well on complex tasks, yet remain highly vulnerable to prompt injection attacks, where malicious instructions embedded in external data can…

  128. Neural Network Verification for Deep Joint Source-Channel Coding
    USarXiv cs.AI

    arXiv:2610.11994v1 Announce Type: cross Abstract: Deep joint source-channel coding (DeepJSCC) transmits data end-to-end over wireless channels using a neural encoder-decoder, but reconstruction quality can degrade sharply under adversarial…

  129. Beyond Sample Copying: Structural Memorization in Diffusion Models
    USarXiv cs.LG

    arXiv:2603.13419v3 Announce Type: replace Abstract: Diffusion models generalize well in practice. Paradoxically, an optimal diffusion model fully memorizes the training data and therefore fails to generalize, raising the question of what…

  130. NavGPT-3: Harnessing Context in a Hierarchical Navigation Runtime
    USarXiv cs.AI

    arXiv:2610.10787v1 Announce Type: cross Abstract: Language models trained with long-horizon agentic reinforcement learning can generalize knowledge through reasoning, express precise actions, and pursue goals over many steps, raising the…

  131. When Citations Mislead? A Claim-Level Benchmark for Legal Hallucination Detection
    USarXiv cs.AI

    arXiv:2610.10971v1 Announce Type: cross Abstract: Large language models are increasingly used in legal research and drafting, but they can still produce claims that sound convincing without being supported by the cited source. We introduce…

  132. Compactness and Consistency: A Conjoint Framework for Deep Graph Clustering
    USarXiv cs.AI

    arXiv:2610.11506v1 Announce Type: cross Abstract: Graph clustering is a fundamental task in data analysis, aiming at grouping nodes with similar characteristics in the graph into clusters. This problem has been widely explored using graph…

  133. Prediction-Powered Data Fusion for Treatment Effect Estimation
    USarXiv cs.LG

    arXiv:2610.12332v1 Announce Type: cross Abstract: Randomized controlled trials (RCTs) identify treatment effects without confounding but are often small, whereas observational studies (OBS) are large but may be confounded. Many estimators…

  134. Efficient and Generalizable Archetypal Analysis for Discrete Data
    USarXiv cs.LG

    arXiv:2610.12035v1 Announce Type: cross Abstract: Archetypal Analysis (AA) represents observations as convex combinations of extremal data-driven profiles, yielding interpretable low-dimensional descriptions of complex datasets. Classical…

  135. One Block, Multiple Depths: Recurrent Vision Transformers with Depth-Programmed Experts
    USarXiv cs.LG

    arXiv:2610.12448v1 Announce Type: cross Abstract: In this work, we show that a single Transformer block, applied recurrently, can match the accuracy of a full-depth vision encoder at comparable inference FLOPs without intermediate feature…

  136. Type-Safe Is Not Error-Free: Typed Decision Models Follow the Option Name, Not the Definition Bound to It
    USarXiv cs.AI

    arXiv:2609.26758v3 Announce Type: replace Abstract: Typed decision models return structured results, but output-type correctness alone does not ensure that decisions follow explicit option definitions. Each option pairs a name with a…

  137. Distillation for Incrimination and Distillation for Capabilities
    USarXiv cs.AI

    arXiv:2610.11012v1 Announce Type: new Abstract: Powerful misaligned AI models might recognize alignment evaluations and strategically behave well on them, rendering direct audits uninformative. However, distilling such a model into a…

  138. Bi-temporal Image-driven Acute Stroke Evolution Analysis
    USarXiv cs.CV

    arXiv:2602.07535v2 Announce Type: replace Abstract: Acute ischemic stroke requires rapid treatment decisions that are strongly guided by emergency imaging. Admission computed tomography perfusion (CTP) is commonly used to estimate the…

  139. MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems
    USarXiv cs.CL

    arXiv:2605.28732v4 Announce Type: replace Abstract: Memory is essential for enabling large language models to support long-horizon reasoning, yet existing memory systems remain unreliable and difficult to debug. Tracing memory's dynamic…

  140. ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks
    USarXiv cs.CL

    arXiv:2609.29102v3 Announce Type: replace Abstract: Fully continuous diffusion language models (dLMs) denoise continuous representations without intermediate discretization, then decode all response tokens in parallel at the final step…

  141. Improving Image-Based Nutrition Estimation Through Multimodal Food-Item Verification and Recovery
    USarXiv cs.AI

    arXiv:2610.11144v1 Announce Type: cross Abstract: Single-image nutrition estimation can fail silently when visible foods are missed. Even when a food is correctly identified, its proposed region may not support portion estimation. We…

  142. Scalable Hierarchical Graph Generation via Soft Community Structure
    USarXiv cs.LG

    arXiv:2610.12163v1 Announce Type: new Abstract: Generating large attributed graphs requires reproducing the topology, generating attributes jointly with the structure, and remaining scalable. Many real-world graphs exist as a single large…

  143. Examining Social Attribution in LLM Reasoning: A Theory-Guided Probing Methodology
    USarXiv cs.CL

    arXiv:2610.12022v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed in sociotechnical systems where social attribution, the reasoning process attributing external events to the causes and reasons of…

  144. Predictive Inorganic Synthesis based on Machine Learning using Small Data sets: a case study of Hydrodynamic Diameter-controlled Cu Nanoparticles
    USarXiv cs.LG

    arXiv:2512.16545v3 Announce Type: replace-cross Abstract: Cu NPs have a broad applicability, yet their synthesis is sensitive to subtle changes in reaction parameters. This sensitivity, combined with the time- and resource-intensive nature…

  145. Can LLMs Fix It Without Code? Toward Automated Verification of No-Code Bug Fixes
    USarXiv cs.AI

    arXiv:2610.11963v1 Announce Type: cross Abstract: A no-code fix resolves an invalid bug report by directing the user to change a setting, update to a version where the problem is already fixed, or adjust their workflow. Manually verifying…

  146. Real-Time Motion Planning with Dynamic Hazards: Classical vs. Learning-Based Methods
    USarXiv cs.AI

    arXiv:2610.12249v1 Announce Type: cross Abstract: We study real-time motion planning in dynamic hazard fields through a controlled comparison between classical planning and learning-based methods. Rather than introducing a new planner, we…

  147. Simultaneous Computation with Multiple Prioritizations in Multi-Agent Motion Planning
    USarXiv cs.AI

    arXiv:2501.10781v2 Announce Type: replace-cross Abstract: Multi-agent path finding (MAPF) in large networks is computationally challenging. An approach for MAPF is prioritized planning (PP), in which agents plan sequentially according to…

  148. Test-Time Compute for Tabular Foundation Models: Mechanisms, Gains, and Limits
    USarXiv cs.LG

    arXiv:2610.12005v1 Announce Type: new Abstract: Which forms of test-time compute improve the predictions of strong pretrained tabular foundation models (TFMs)? We systematically study this along three axes: adaptation, aggregation, and…

  149. Beyond QAOA: A Review of AI and Quantum Computing for Adaptive Combinatorial Optimization
    USarXiv cs.LG

    arXiv:2610.11759v1 Announce Type: cross Abstract: Near-term quantum approaches to combinatorial optimization are limited by qubit counts, circuit fidelity, sampling cost, and the difficulty of encoding constraints, while machine learning…

  150. EvoAlloc: A Self-Evolving Resource Allocation Agent for Efficient Program Evolution
    USarXiv cs.AI

    arXiv:2610.12086v1 Announce Type: new Abstract: LLM-based program evolution relies on evaluation feedback to guide the iterative search for high-performing programs. However, evaluation is often computationally expensive, making it…