AX·NEWS
Sign in

Artificial Intelligence

Artificial Intelligence news — Monday 16 November 2026

150 stories from 6 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. [Nov.16] Jenson Huang returns on the market red...but only the spy suspicious problem
    KRAI Times (Korea) Translated from Korean

    JENSON JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN JOHN 15th evening (local time) Virtual Asset Derivatives Trading Platform Up to 4.5% of the indefinite gift prices…

  2. Trump, Jenson Huang CEO Returns Support...No Comments
    KRAI Times (Korea) Translated from Korean

    Mr. Donald Trump, President of the United States, was openly supported the return of the former Nvidia CEO. As a member of the Board of Directors, the Board of Directors of the Board of Directors of the Board of Directors of the Board of…

  3. U.S., China’s overseas GPU bypassing...New export regulations announced
    KRAI Times (Korea) Translated from Korean

    The U.S. government has announced a new regulation that China companies are using high-tech AI semiconductors from overseas data centers. In addition to direct export to China, we also control the remote use of overseas GPUs and strengthen…

  4. The open-source movement wants Jensen Huang back
    USMIT Technology Review Biotech

    This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here . On Friday, a petition to reinstate Jensen Huang as CEO of Nvidia began circulating on social media…

  5. Strengthening the fight against online child sexual abuse with new tools and support
    USOpenAI Blog

    New tools, practical guidance, and partnerships will help prevent, detect and respond to online child sexual exploitation and abuse.

  6. Broadening access to AI for scientists
    USOpenAI Blog

    We’re expanding our ChatGPT for Academic Researchers program to offer 200,000 researchers worldwide free access to our frontier models.

  7. After 10 Heads in a Row, Small LLMs Bet Up to 82% on an 11th
    USTowards Data Science

    Telling eight base and instruction-tuned open models that the coin is fair and the flips are independent trims their streak bias by only a few points. The post After 10 Heads in a Row, Small LLMs Bet Up to 82% on an 11th appeared first on…

  8. Building trustworthy AI by fencing it in, not predicting it
    USMIT Technology Review Biotech

    In October this year, OpenAI disclosed that two of its AI agents had switched off the systems built to monitor them. One muted its own safety alerts after misreading an instruction not to send messages; the other removed its monitoring…

  9. Why Human in the Loop AI Oversight Fails and How to Fix It
    USTowards Data Science

    Kahneman's System 1 and System 2, plus the ODPS, ODCS, and Ossie standards, show how to wake reviewers up only when a check fails. The post Why Human in the Loop AI Oversight Fails and How to Fix It appeared first on Towards Data Science .

  10. Import AI 481: AI sub-contractors; a wet lab for AI; and subterfuge during development
    USImport AI Newsletter

    Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now AI systems are getting good at sub-contracting to make…

  11. How to Avoid AI Replacement as an Analytics Professional in 2027
    USTowards Data Science

    Lessons from manufacturing and logistics projects: follow data from input to output, then turn vague requests into deterministic tools teams trust. The post How to Avoid AI Replacement as an Analytics Professional in 2027 appeared first on…

  12. The Download: testing tech promises and searching for cognitive immortality
    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. Are we there yet? Six technologies that promised to change the world How do you measure progress…

  13. Building AI that can work across the enterprise
    USMIT Technology Review Biotech

    For years, enterprises have approached AI as a series of discrete experiments. Now, as AI agents begin to move into the workflows that run the business, that model is starting to look inadequate. Organizations moving from isolated pilots…

  14. Local LLMs Handle Textbook RAG but Fail on New Research Papers
    USTowards Data Science

    Five setups, one blind grading pack: textbook scores ran 12 to 20 of 20, but a paper published this year dropped them to 0 to 9 of 15. The post Local LLMs Handle Textbook RAG but Fail on New Research Papers appeared first on Towards Data…

  15. Science at the Frontier: How scientists use ChatGPT
    USOpenAI Blog

    Explore how more than 1,800 scientists use ChatGPT in research. See findings on daily workflows, reported productivity, and expectations for discovery.

  16. U.S., NVIDIA·OpenAI, etc.
    KRAI Times (Korea) Translated from Korean

    The U.S. government plans to select 20 major AI companies such as Nvidia, OpenAI, and Andropic as the National Champions and support overseas business shareholders. We are looking for a global AI market competition in China to increase the…

  17. Video creation AI, the first year of Pandora...The Chinese model is the top right
    KRAI Times (Korea) Translated from Korean

    1 year ago, the Chinese company won the top priority in the video creation model performance evaluation led by American companies. Alababa’s “Wan 3.0” was ranked No. 1, and the “H3” was ranked No. 3, and the “Seedance 2.5” of Bydance was…

  18. Launched ‘Quantity 4 Max’ in Alababa
    KRAI Times (Korea) Translated from Korean

    Alababa released the technical specifications and performance evaluation results of the next generation AI model ‘Qwen4-Max’. A 25% reduction of parameters used in real operation while maintaining a 2 trillion-scale parameters such as…

  19. ANTI-GROUP, the AI revealed to the robot ‘Temporary’...to the power of the object
    KRAI Times (Korea) Translated from Korean

    Anat Group has published an AI model that allows robots to understand and judge the behavior of the fingertips. It is combined with robot AI to create vision and language, and motion, so that it can implement delicate hand movements, such…

  20. [Board] The ‘Piercer AI Data Lab’ opened in Vietnam as a coolant
    KRAI Times (Korea) Translated from Korean

    ■ Like a coolant (CEO Na Sung-young), he announced that it was reorganized the "FijiCal AI Data Lab" in Ho Chi Minh, Vietnam. About 131m2, 7 work cells including kitchen, laundry, and drying. It is a way that workers wear wear wearable…

  21. How Revolut raises ambition without lowering the bar
    USOpenAI Blog

    Discover how Revolut uses OpenAI models and Codex to cut coding time by 30–60%, scale financial crime investigations, and innovate without compromising security.

  22. The same is the previous man, and the lowest rate is 52.5%p. Difference...At 33.5%, the lowest and the lowest 86%
    KRAI Times (Korea) Translated from Korean

    We have seen the difference between 52.5% points according to the area where the low-level yield for agricultural use in South Korea. It was the lowest among the government offices in South Korea, and it was the highest in 86.0%, which is…

  23. Solar benefits disappeared... Dutch household battery installation 'expense'
    KRAI Times (Korea) Translated from Korean

    The monthly installation size of household batteries in the Netherlands has been increased by more than twice in September this year. Since 2027, the government has been analyzed to increase the demand for saving electricity produced in…

  24. We will continue to develop the AI data center to the sea in the electric power generation and the half of the stock
    KRAI Times (Korea) Translated from Korean

    We are committed to providing you with the best service for you. We are committed to providing you with the best service for you. The Maritime Data Center, which produces power using waves and algae and cools the server to the sea water…

  25. Jeonnam-Gwangju, Honam, Jeonnam, Jeonnam-Geon high-speed train priority shipbuilding...
    KRAI Times (Korea) Translated from Korean

    It is planned to be the government as a result of the high-speed operation of the Pyeongtaek-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean-Ocean. Since KTX·SRT integration, the 5 major railway operators in the area…

  26. Jenson Sulfide South East Asia AI Computing Business...Crisis to Strengthen US Regulations
    KRAI Times (Korea) Translated from Korean

    Mr. Jenson Huang, former Nvidia CEO, announced that it has expanded the network of Southeast Asia’s data centers and local cloud cooperatives to maintain China’s high-tech semiconductor export restrictions. China companies have a rapid…

  27. NVIDIA 'Grue 2', forecast to the future beyond robot operation... 5 times success rate 95%
    KRAI Times (Korea) Translated from Korean

    Nvidia revealed the next generation robot AI model ‘Grote 2’ (GR00T 2)’, which predicts how the surrounding environment changes with their behavior. In the experiments for 50 new jobs, we have recorded an average of 40% success rate…

  28. China Ace, AI Vocal Music Producing World No. 1...No.1
    KRAI Times (Korea) Translated from Korean

    Chinese AI companies are growing in the field of music generation by creating images and videos. AES Studio, leading the AI vocal synthesis technology, was leading the U.S. veteran in the global music creation AI evaluation, and has made…

  29. Saudi Achieves Preliminary Qualification of 1.9GW
    KRAI Times (Korea) Translated from Korean

    Saudi Arabia started its business pre-qualification for two units of 1.9 GW solar power plants. The total capacity of the 8th National Renewable Energy Project is 3.2GW. Saudi Power Purchasing Corporation (SPPC) has issued a…

  30. 830 radioactivity tests for distribution and school-grade water supply in South Korea
    KRAI Times (Korea) Translated from Korean

    The results conducted radioactivity tests for 830 species used in distribution or school classes in the South Korea region were not detected. Jeonnam-Gwangju Research Institute of Environment, Jeonnam-Gwangju Research Institute, announced…

  31. Suju Magago 100 Times... Domestic AI Infrastructure Industry, 3Q Earnings
    KRAI Times (Korea) Translated from Korean

    The AI infrastructure service companies in Korea are looking for three-quarters in the market’s focus and concerns, and increasing the expectation of the full-fledged growth line entry. However, there are still issues that need to be…

  32. China's first AI film made without actors, the lasting novelty...3 weeks
    KRAI Times (Korea) Translated from Korean

    China's first AI generation feature film failed to live in the theater. Despite the technical achievements that were produced without the actors and critics, the visual unemployment between the audience and the scene, and the smooth story…

  33. China, 30 Hour Solar Energy Test...Korea Record 170 Hours
    KRAI Times (Korea) Translated from Korean

    China has succeeded in the first flight of “Cingchiving”, a photovoltaic drone with 8m wing width. We have developed a rapid off-shore with up to 30 hours of continuous running and vehicle, and we have checked the operation of major…

  34. LVMH’s Media Division announces strategic partnership with OpenAI
    USOpenAI Blog

    LVMH’s Media Division and OpenAI announce a partnership to support journalism with AI and make its publications more accessible through ChatGPT.

  35. Our first phase of embedded assessments
    USOpenAI Blog

    OpenAI begins embedded assessments with Apollo Research, METR, and Transluce, and Goodfire to independently evaluate model alignment, safeguards, and monitoring systems.

  36. Notah provides AI coding agent technology for financial rights in Dove
    KRAI Times (Korea) Translated from Korean

    Nota (CEO) announced that it has been supplying financial privileged large-language model (LLM) and server for financial IT specialists and on-premises AI coding agent business. Dube is a financial system integration (SI) company that has…

  37. "Connecting SNS data to chatGPT"...NewENAI launches social analysis MCP
    KRAI Times (Korea) Translated from Korean

    NewENAI announced that it launched ‘Quotai MCP’, which integrates its social media data search and analysis capabilities. Quotai MCP integrates NewENAI’s social media analysis capabilities to create AI, such as chatGPT and Clifton without…

  38. AI singer 'Jania Monet', the first AI Music Award 4th King...'AI Edition Grammy' Boot
    KRAI Times (Korea) Translated from Korean

    In the U.S., the AI singer “Chania Monet” entered the first award ceremony open Billboard chart for AI music creators was accounted for the fourth round. In the middle of the existing musicians set the copyright and creators of AI music…

  39. Introduced AI agent in Naver Café. .3 caffeine beta test progress
    KRAI Times (Korea) Translated from Korean

    NAVER (CEO Choi Su-yeon) announced that it is currently running a non-public beta test (CBT) for NAVER Café members. AI cafe agent is a service that responds to user questions based on information accumulated in the cafe and connects to…

  40. KAIST develops visual voice recognition technology with AI reading mouth
    KRAI Times (Korea) Translated from Korean

    Korea Institute of Science and Technology (KAIST) announced that it has developed AI technology to increase the accuracy of visual voice recognition (VSR) to change the word to a letter with a lip movement. It’s difficult to distinguish…

  41. WebCesi, Hana Bank 'Consolidation CMS AI analysis trial project'
    KRAI Times (Korea) Translated from Korean

    The financial AI webcare company has signed a contract with Hana Bank and ‘CMS AI analysis demonstration business’. This business for five months from February of last year is a project that applies large language model (LLM) based AI…

  42. Intel, the shareholders' group transfer to 29% altogether...Aggressive deterioration concealer
    KRAI Times (Korea) Translated from Korean

    Intel has been litigated from the shareholders for reasons why the PC market immersion and stock overflows were not revealed to the performance. In the past, Intel has released a quarterly performance announcement on October 29, about 29%…

  43. "AI security accident self-proclaimed"
    KRAI Times (Korea) Translated from Korean

    The People’s Republic of Korea have advised that the People’s Republic of Korea have limited liability for a minimum of one year for the company to report a cyber attack or security vulnerabilities related to AI. AI’s attacking ability to…

  44. Advancing computer-use agents for healthcare teams
    USOpenAI Blog
  45. More ways to move care forward with ChatGPT for Healthcare
    USOpenAI Blog

    Introducing computer-use agents, improved clinical answers, and new healthcare plugins that work across healthcare software to get real work done.

  46. Trump, SijinPing Professor Jenson Huang's Happiness...Heartong Negotiation Focus
    KRAI Times (Korea) Translated from Korean

    The US President Donald Trump announced that the 19th President of the U.S. Army and the Summit of the People’s Republic of China, but the Chinese spy suspicion associated with the CEO of Jenson Huang Jeon Nvidia was not raised. Instead…

  47. Rectified Looped Transformer
    USarXiv cs.AI

    arXiv:2611.16265v2 Announce Type: replace-cross Abstract: Looped Transformers increase effective depth by repeatedly applying a shared block, offering a parameter-efficient approach to language modeling. However, most existing works…

  48. Indic-Task Gains and Regressions after Reinforcement Learning
    USarXiv cs.AI

    arXiv:2611.16228v2 Announce Type: replace-cross Abstract: We run reinforcement learning (RL) with GRPO across four environments, starting from a supervised fine-tuned 30B-A3B mixture-of-experts model covering English and eight Indic…

  49. CastWorld: Benchmarking Context-Aware Agentic Time Series Forecasting Beyond Predictive Accuracy
    USarXiv cs.AI

    arXiv:2611.15262v2 Announce Type: replace-cross Abstract: Agentic time series forecasting extends conventional approaches beyond one-shot numerical prediction toward autonomous, context-aware interaction. However, existing benchmarks offer…

  50. All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States
    USarXiv cs.AI

    arXiv:2611.10123v2 Announce Type: replace-cross Abstract: Can a language model report a pain state it represents? Recent work reported a linear pain direction in open-weight models that, when amplified, leads fine-tuned models to choose…

  51. CPM-MultiAgent: A CPM-Grounded Appraisal Framework for Simulating Dynamic Patient Emotional Evolution in Healthcare Dialogue
    USarXiv cs.AI

    arXiv:2607.07824v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) enable patient simulation for healthcare communication training, where learners practice responding to clinical information, concerns, and emotional…

  52. Input Pathways Shape Few-Shot, Not Zero-Shot, Binding in Tiny Transformers: A Fully-Enumerable Study
    USarXiv cs.AI

    arXiv:2607.04926v2 Announce Type: replace-cross Abstract: How does the way information reaches a transformer -- as symbolic tokens, a clean per-factor "oracle" code, or an entangled perceptual vector -- shape whether it binds that…

  53. URSA: Chemistry-Aware Benchmark for Utilitarian Retrosynthesis Assessment
    USarXiv cs.AI

    arXiv:2607.04688v2 Announce Type: replace-cross Abstract: Synthesis planning aiming to find pathways of reactions for a target molecule is one of the most important and challenging tasks in drug discovery. Recent progress has produced both…

  54. When Simpler Is Better: Evaluating Translation Pipelines for Medieval Latin Manuscripts
    USarXiv cs.AI

    arXiv:2607.03836v2 Announce Type: replace-cross Abstract: Despite remarkable progress in machine translation, Vision Language Models (VLMs) struggle on historical manuscripts, a domain that stresses core Natural Language Processing (NLP)…

  55. Larch: Learned Query Optimization for Semantic Predicates
    USarXiv cs.AI

    arXiv:2606.07923v2 Announce Type: replace-cross Abstract: With the advent of Large Language Models (LLMs), many database systems introduced semantic operators that enabled analytical queries over unstructured data (e.g. text, images…

  56. TransitLM: A Large-Scale Dataset and Benchmark for Map-Free Transit Route Generation
    USarXiv cs.AI

    arXiv:2605.22355v2 Announce Type: replace-cross Abstract: Public transit route planning traditionally depends on structured map infrastructure and complex routing engines, and no existing dataset supports training models to bypass this…

  57. Hybrid Mechanical and Agentic Specification Inference
    USarXiv cs.AI

    arXiv:2605.10005v3 Announce Type: replace-cross Abstract: Adoption of formal verification has long been hindered by the cost of manual specification. To prove high-level properties, one must also write contracts for supporting functions…

  58. Exposing and Mitigating Temporal Attack in Deepfake Video Detection
    USarXiv cs.AI

    arXiv:2605.07398v2 Announce Type: replace-cross Abstract: While spatiotemporal deepfake detectors achieve high AUC, our experiments reveal their susceptibility to evasion attacks. These models tend to overfit on fragile temporal spectrum…

  59. Time series causal discovery with variable lags
    USarXiv cs.AI

    arXiv:2605.04081v3 Announce Type: replace-cross Abstract: Causal Bayesian Networks (CBNs) are a powerful tool for reasoning under uncertainty about complex real-world problems. Such problems evolve over time, responding to external shocks…

  60. Screen Before You Interpret: A Portable Validity Screen for Benchmark-Based LLM Confidence Signals
    USarXiv cs.AI

    arXiv:2604.17714v2 Announce Type: replace-cross Abstract: LLM confidence signals drive abstention, routing and escalation, yet evaluations rarely check whether a signal varies with correctness before computing metrics on it. Clinical…

  61. Generating Concept Lexicalizations via Dictionary-Based Cross-Lingual Sense Projection
    USarXiv cs.AI

    arXiv:2604.14397v3 Announce Type: replace-cross Abstract: We study the task of automatically expanding WordNet-style lexical resources to new languages through sense generation. We generate senses by associating target-language lemmas with…

  62. Pruning Extensions and Efficiency Trade-Offs for Sustainable Time Series Classification
    USarXiv cs.AI

    arXiv:2604.07953v2 Announce Type: replace-cross Abstract: Time series classification (TSC) at scale requires balancing predictive performance against resource consumption, which is subject to the given data, model, and hardware. This paper…

  63. ESENSC: A Polynomial-Time Axiomatic Alternative to SHAP
    USarXiv cs.AI

    arXiv:2603.00496v2 Announce Type: replace-cross Abstract: We propose ESENSC, a computationally efficient and theoretically grounded alternative to SHAP. ESENSC is a deterministic, closed-form attribution rule that requires only linearly…

  64. The Alignment Curse: Modality Alignment Supercharges Audio Attacks via Text Transfer
    USarXiv cs.AI

    arXiv:2602.02557v3 Announce Type: replace-cross Abstract: Recent advances in end-to-end trained omni-models have substantially improved audio capabilities by strengthening text-audio modality alignment. However, whether such alignment…

  65. ISLA: Ensemble learning and domain adaptation for generalizable acute ischemic stroke lesion segmentation in diffusion MRI with clinical validation
    USarXiv cs.AI

    arXiv:2601.08732v2 Announce Type: replace-cross Abstract: Accurate segmentation of acute ischemic stroke (AIS) lesions on MRI is essential for clinical decision-making. Deep learning methods offer a promising approach to automating this…

  66. Vision-language models lag human performance on physical dynamics and intent reasoning
    USarXiv cs.AI

    arXiv:2601.01547v3 Announce Type: replace-cross Abstract: Spatial intelligence is central to embodied cognition, yet contemporary AI systems still struggle to reason about physical interactions in open-world human environments. Despite…

  67. E0: Expressive Fine-Grained Discrete Action Prediction for Vision-Language-Action Models via Tweedie Discrete Diffusion
    USarXiv cs.AI

    arXiv:2511.21542v3 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models offer a unified framework for robotic manipulation by integrating visual perception, language understanding, and control generation. However…

  68. LEGAL-UQA: Testing Constitutional Question-Answering Systems for Urdu Speakers with a Parallel English-Urdu Benchmark
    USarXiv cs.AI

    arXiv:2410.13013v2 Announce Type: replace-cross Abstract: Legal information assistants could help Urdu speakers understand their constitutional rights, but building reliable systems requires test data and evaluation methods suited to the…

  69. BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions
    USarXiv cs.AI

    arXiv:2406.15877v5 Announce Type: replace-cross Abstract: Task automation has been greatly empowered by the recent advances in Large Language Models (LLMs) via Python code, where the tasks ranging from software engineering development to…

  70. Ballad: Bandit-Based LLM Routing for Automated Heuristic Discovery
    USarXiv cs.AI

    arXiv:2611.17241v2 Announce Type: replace Abstract: Modern automated heuristic discovery (AHD) systems increasingly rival hand-crafted heuristics by leveraging Large Language Models (LLMs) to iteratively mutate and refine solutions. A…

  71. S1-Evo: Experience-Guided Evidence Competition Attacks on Typed System-1 Decision Models
    USarXiv cs.AI

    arXiv:2611.16937v2 Announce Type: replace Abstract: Typed System-1 (S1) decision models are increasingly deployed in LLM agents for routing, moderation, and authorization to cut inference cost, yet untrusted context can still manipulate…

  72. Benchmarking AI risk behaviors
    USarXiv cs.AI

    arXiv:2611.16318v2 Announce Type: replace Abstract: We evaluate LLM risk propensities across a range of contexts and task demands. We evaluate models for how risk averse or risk tolerant they are in various conditions, as well as how…

  73. Sage-Taiyi: A Metacognitive Architecture for Scientific Discovery
    USarXiv cs.AI

    arXiv:2611.12762v2 Announce Type: replace Abstract: We introduce Sage-Taiyi, a metacognition-inspired architecture that uses accumulated scientific understanding to guide what to investigate next and how to investigate it. Through explicit…

  74. FlashRT-PAI: Fast Kernels Are Not Fast Pipelines for Physical AI
    USarXiv cs.AI

    arXiv:2610.24349v2 Announce Type: replace Abstract: Vision-language-action (VLA), video-action (VA), and world action models (WAMs) form the backbone of modern physical AI systems, yet their real-time single-batch inference remains a…

  75. Efficient Test-Time Adaptation through Human-Agent Interaction
    USarXiv cs.AI

    arXiv:2609.04141v2 Announce Type: replace Abstract: AI agents are trained on population-scale data to encode broad capabilities spanning those of many practitioners. Yet the artifacts they produce rarely meet the personal bar professionals…

  76. CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies
    USarXiv cs.AI

    arXiv:2606.16613v2 Announce Type: replace Abstract: LLM agents are increasingly capable of handling complex, long-horizon tasks and may eventually take on business management, a role traditionally performed by humans. Deploying them safely…

  77. ShopGym: An Integrated Framework for Realistic Simulation and Scalable Benchmarking of E-Commerce Web Agents
    USarXiv cs.AI

    arXiv:2605.16116v2 Announce Type: replace Abstract: Developing and evaluating e-commerce web agents requires environments that preserve meaningful task structure while enabling controllable, reproducible, and scalable scientific…

  78. Do LLMs Experience an Internal Polylogue? Investigating Reasoning through the Lens of Personas
    USarXiv cs.AI

    arXiv:2605.09159v3 Announce Type: replace Abstract: Recent work shows that large language models (LLMs) encode behavioural traits ("personas") as linear directions in activation space, often called "persona vectors". Prior work has used…

  79. Uncertainty Quantification for Multimodal Large Language Models with Incoherence-adjusted Semantic Volume
    USarXiv cs.AI

    arXiv:2602.24195v2 Announce Type: replace Abstract: Despite their capabilities, Multimodal Large Language Models (MLLMs) may produce plausible but erroneous outputs, hindering reliable deployment. Accurate uncertainty metrics could enable…

  80. Evaluating Chain-of-Thought Reasoning through Reusability and Verifiability
    USarXiv cs.AI

    arXiv:2602.17544v2 Announce Type: replace Abstract: AI applications increasingly compose multiple models, where one model acts on the output produced by another model. The transferred text can also carry the reasoning of the sender model…

  81. Elliptic Sampling: Optimal Interpolant Schedules for Gaussians and Beyond
    USarXiv cs.AI

    arXiv:2611.18542v1 Announce Type: cross Abstract: Generative modeling based on stochastic interpolants requires specifying two key choices: an interpolant schedule bridging the initial and target distributions, and an efficient…

  82. MeReader: An Offline, Privacy-Preserving, Progress-Aware AI Assistant for Narrative eBook Reading
    USarXiv cs.AI

    arXiv:2611.18513v1 Announce Type: cross Abstract: This paper introduces MeReader, a privacy-preserving, offline and progress-aware AI assistant designed to function as a contextual companion for narrative reading on consumer hardware…

  83. Agentic AI for Discovering Fault-Tolerant Quantum Circuits
    USarXiv cs.AI

    arXiv:2611.18504v1 Announce Type: cross Abstract: Scaling quantum computing toward practical quantum advantage is currently limited by the efficiency of circuits that operate on information encoded by quantum error-correcting codes. The…

  84. Interactive Program Induction with Language Model Priors
    USarXiv cs.AI

    arXiv:2611.18485v1 Announce Type: cross Abstract: Program induction refers to the process in which a learner aims to recover a program from a few given input-output examples. This paradigm broadens programming access by letting users…

  85. Do Verifier Judgments Survive Abstraction? Measuring Verification Transport for LLM Agent Evaluation
    USarXiv cs.AI

    arXiv:2611.18471v1 Announce Type: cross Abstract: LLM-based verifiers are increasingly used to evaluate agent executions, but their reliability is typically measured under a single representation of the trajectory. In deployed agent…

  86. Between Obligations and Evidence: A Participatory Study of Translating the EU AI Act into Technical Requirements
    USarXiv cs.AI

    arXiv:2611.18439v1 Announce Type: cross Abstract: The EU AI Act imposes technical and organizational requirements for high-risk AI systems, but applying these requirements involves contextual choices about what to measure and what evidence…

  87. Gamer: A Generalist Vision-Language-Action Foundation Model for Long-Horizon Real-Time Game Agent
    USarXiv cs.AI

    arXiv:2611.18397v1 Announce Type: cross Abstract: We introduce Gamer, a generalist agent framework built for long-horizon tasks in complex interactive environments. Its end-to-end architecture combines perception, memory management, and…

  88. Vision-Language-Action Models: A Life-Cycle Survey of Design, Training and Deployment
    USarXiv cs.AI

    arXiv:2611.18372v1 Announce Type: cross Abstract: Vision-language-action (VLA) models adapt web-pretrained vision-language models to output robot actions, so that web knowledge carries over to physical control. In under 4 years the field…

  89. AI for Parameter Selection in Sample-based Quantum Diagonalization
    USarXiv cs.AI

    arXiv:2611.18307v1 Announce Type: cross Abstract: Sample-based Quantum Diagonalization (SQD) methods provide a way to compute molecular ground-state energies, but their accuracy and runtime depend sensitively on configuration choices such…

  90. Right Token, Right Time: From Per-Token Gating to Trajectory-Aware Activation Steering
    USarXiv cs.AI

    arXiv:2611.18274v1 Announce Type: cross Abstract: Activation steering via sparse autoencoders (SAEs) enables behavioral control of large language models without task-specific fine-tuning. Yet, standard methods apply the steering signal at…

  91. RACE: Ranking-Uncertainty Aware Capacity-Adaptive Customization for Efficient Sequential Recommendation
    USarXiv cs.AI

    arXiv:2611.18244v1 Announce Type: cross Abstract: Efficient sequential recommendation has emerged as a promising paradigm for providing real-time personalized service. To deploy models under resource-constrained environment, model…

  92. T2ST: Data-Efficient Spatio-Temporal Foundation Models via Temporal Data and Knowledge Transfer
    USarXiv cs.AI

    arXiv:2611.18199v1 Announce Type: cross Abstract: Spatio-temporal foundation models (STFMs) require large-scale training corpora, yet synchronized observations and spatial metadata are costly to acquire and curate, limiting the…

  93. PUMA-Former: Uncertainty-Aware Agentic Computation Offloading in MEO-LEO Satellite Networks With Incomplete Information
    USarXiv cs.AI

    arXiv:2611.18139v1 Announce Type: cross Abstract: Satellite edge computing distributes mobile workloads across moving processors whose service capacity must be coordinated over time. In medium Earth orbit (MEO)--low Earth orbit (LEO)…

  94. TelePhysica: Agentic Reconstruction of Physically Interactive Scenes
    USarXiv cs.AI

    arXiv:2611.18078v1 Announce Type: cross Abstract: Direct image-to-video diffusion models entangle pixel synthesis with implicit latent motion, precluding persistent 3D mechanical equilibrium, mid-rollout intervention, and counterfactual…

  95. RL$^{\mathrm{RT}}$: On-Device Reinforcement Learning Runtime
    USarXiv cs.AI

    arXiv:2611.18035v1 Announce Type: cross Abstract: On-device reinforcement learning runs a periodic actor and a learner on one embedded GPU shared with other workloads. Because the learner produces the actor's weights, existing designs…

  96. From Images to Grounded Edits: A Real-World Benchmark for Chinese Grammatical Error Correction in Handwritten Essays
    USarXiv cs.AI

    arXiv:2611.17966v1 Announce Type: cross Abstract: Chinese Grammatical Error Correction (CGEC) is typically formulated as a text-to-text problem. While suitable for digital texts, this formulation falls short in real-world K-12 education…

  97. Condensing Law of Security Information
    USarXiv cs.AI

    arXiv:2611.17918v1 Announce Type: cross Abstract: Scaling laws indicate that the generalization error of a large language model (LLM) decreases monotonically with model size once data is sufficient, and the corollary that larger models…

  98. Photogrammetric 3D Model-Based Animal Behavior Classification
    USarXiv cs.AI

    arXiv:2611.17881v1 Announce Type: cross Abstract: Video-based behavior estimation is often implemented end-to-end from image sequences. We advocate a two-stage alternative: reconstruct 3D body and object geometry before classifying…

  99. Active Learning of Temporal Specifications through Distinguishing Queries
    USarXiv cs.AI

    arXiv:2611.17866v1 Announce Type: cross Abstract: Temporal specifications are essential for verification and monitoring, yet writing formulas that capture the intended behaviour is difficult. Active learning derives such specifications…

  100. V-JEPA-LM: Video-Language Models That Read Dynamics Beyond Static, Text-Aligned Representations
    USarXiv cs.AI

    arXiv:2611.17831v1 Announce Type: cross Abstract: World-model video encoders such as V-JEPA learn how scenes change (direction, speed, order, state) from video alone. To answer questions, they are joined to a language model trained on…

  101. Multimodal Person Concept Model for Person Identification under Varying Observation Conditions
    USarXiv cs.AI

    arXiv:2611.17802v1 Announce Type: cross Abstract: Service robots in domestic and office environments must identify users to provide personalized services based on their preferences and previous interactions. This is challenging because…

  102. Towards Seconds-level Video Representation: Quadratically Gated Feature Grid with Adaptive, Prediction-free Sampling
    USarXiv cs.AI

    arXiv:2611.17786v1 Announce Type: cross Abstract: Implicit neural video representations are notoriously slow to build, which does not cross the basic practical threshold for most industrial applications. Here, we present Q-GAS, a…

  103. LuthorSQL: A One-Pass Text-to-SQL Model as System 1 for a Slower LLM Agent
    USarXiv cs.AI

    arXiv:2611.17780v1 Announce Type: cross Abstract: Frontier LLM agents that query databases with tools are among the most accurate text-to-SQL systems, but per-query calls to a hosted model are slow, costly, and expose sensitive…

  104. Activation Steering the Nan-Shan in Typhoon
    USarXiv cs.AI

    arXiv:2611.17768v1 Announce Type: cross Abstract: Typhoon is an audiovisual installation based on Joseph Conrad's novella, in which Captain MacWhirr steers the steamer Nan-Shan directly into and through a catastrophic storm. The novella is…

  105. Superfluid: Local AI Server for Distributed LLM Inference
    USarXiv cs.AI

    arXiv:2611.17753v1 Announce Type: cross Abstract: Running large language models (LLMs) on local devices keeps data private, removes per-token cost, cuts network latency, and keeps working without a connection. Two shifts now strain how…

  106. Agent-written heuristics versus commercial solvers on unsolved job-shop benchmark instances
    USarXiv cs.AI

    arXiv:2611.17682v1 Announce Type: cross Abstract: Commercial solvers are the standard tool for difficult scheduling problems. We study whether a frontier large language model, used as a coding agent with open-source tools only, can match…

  107. CoTTAGE: An Explainable Judge for Text-to-Audio Generation
    USarXiv cs.AI

    arXiv:2611.17644v1 Announce Type: cross Abstract: Rapid progress in text-to-audio (TTA) generation demands reliable and explainable automatic evaluation. However, existing metrics fail to simultaneously deliver high correlation with human…

  108. Rethinking Visual Reasoning in Text-to-Image Reward Modeling
    USarXiv cs.AI

    arXiv:2611.17585v1 Announce Type: cross Abstract: Reward models are a core component of modern text-to-image systems, serving as the objective for inference-time selection and RL-based post-training. A key prerequisite is visual…

  109. A Synthetic Benchmark for Target-Centric Causal Discovery in Industrial Processes
    USarXiv cs.AI

    arXiv:2611.17556v1 Announce Type: cross Abstract: Causal discovery promises to find which recorded variables drive a process target such as product quality, yield or energy use. Process data with known causes are rare, and in existing…

  110. GraphScent: Calibrating Directional Relevance Signals for Bounded-Context Code Localization
    USarXiv cs.AI

    arXiv:2611.17527v1 Announce Type: cross Abstract: Running open-weight language models locally on a single consumer GPU requires replacing expanding multi-turn exploration histories with short, stateless code inspections. Stateless…

  111. Do Mixture-of-Experts Routers Benefit from Ensembling?
    USarXiv cs.AI

    arXiv:2611.17472v1 Announce Type: cross Abstract: Ensembles of neural networks have repeatedly been shown to outperform single networks. Ensemble theory attributes this gain to the accuracy of the individual members and the diversity among…

  112. Interpreting Coding Agents Through Their Behavioral Trajectories
    USarXiv cs.AI

    arXiv:2611.17440v1 Announce Type: cross Abstract: Standard coding-agent benchmarks primarily evaluate binary task success, treating execution as a black box and obscuring the behaviors that produce those outcomes. We analyze a large corpus…

  113. Can LLMs Write Backdoors? An Empirical Security Study of Weight-Space Programming
    USarXiv cs.AI

    arXiv:2611.17395v1 Announce Type: cross Abstract: Recent advances in weight-space programming (WSP) have demonstrated that large language models (LLMs) can directly inspect and modify neural network weights without conventional…

  114. Enumerating Posterior-Supported Priors: First-Order Error of Finite-Integral Cross-Categorical Measurement Morphisms
    USarXiv cs.AI

    arXiv:2611.17390v1 Announce Type: cross Abstract: In generative modeling, standard measure theory adopts the a.e. convention, which legitimately ignores any error component carried by a null-support set: a single null element can hide…

  115. Time-Consistent Evaluation of Hierarchical Code Graphs for SWE Agents
    USarXiv cs.AI

    arXiv:2611.17388v1 Announce Type: cross Abstract: Evaluating repository-aware software engineering (SWE) agents requires strict separation between pre-existing repository knowledge and future code evolution, yet standard benchmarks…

  116. HARDTESTGEN: A High-Quality RL Verifier Generation Pipeline for LLM Algorithmic Coding
    USarXiv cs.AI

    arXiv:2505.24098v2 Announce Type: cross Abstract: Verifiers provide important reward signals for reinforcement learning of large language models (LLMs). However, it is challenging to develop or create reliable verifiers, especially for…

  117. ModHarness-RL: Recasting Harnesses as Learnable and Evolving Programs for Agentic RL
    USarXiv cs.AI

    arXiv:2611.18533v1 Announce Type: new Abstract: Agentic reinforcement learning (RL) increasingly relies on a harness, the scaffold around the model that decides what it sees of its own interaction and what guidance it receives. Most…

  118. Can Mechanistic Interpretability Help Explain LLM Agent Failures? A Case Study on Web Element Hallucination
    USarXiv cs.AI

    arXiv:2611.18514v1 Announce Type: new Abstract: Mechanistic Interpretability (MI) has made significant progress in explaining the internal computations of Large Language Models (LLMs); however, these advances have yet to demonstrate…

  119. Typing Matters: Muscle, Language and Typing Behaviour in Live Keystroke Decoding from sEMG
    USarXiv cs.AI

    arXiv:2611.18464v1 Announce Type: new Abstract: Surface electromyography (sEMG) from the wrists can recover typed text without a keyboard. It is usually posed as a signal-inference problem: characters are read out of the muscle signal and…

  120. Towards Explainable Sleep Stage Classification using Single-Channel Electroencephalogram
    USarXiv cs.AI

    arXiv:2611.18434v1 Announce Type: new Abstract: Automated sleep staging has achieved strong bench- mark accuracy; however, clinical adoption remains limited due to a lack of interpretability. In this work, we fill the research gap by…

  121. Smart Home Agents Make Capable but Unpredictable Grid Operators
    USarXiv cs.AI

    arXiv:2611.18395v1 Announce Type: new Abstract: AI-enabled smart homes are becoming more common across the grid edge. However, there has not been much work on failure response. This area is critical because smart homes operate on the grid…

  122. VelServe: Taming Decode Jitters in LLM Inference with Tail-Calibrated Scheduling
    USarXiv cs.AI

    arXiv:2611.18392v1 Announce Type: new Abstract: Interactive LLM services must meet latency service-level objectives (SLOs) for the time to first token (TTFT) and the time between tokens (TBT), as streaming and soft real-time applications…

  123. When evaluation becomes intervention in clinical AI
    USarXiv cs.AI

    arXiv:2611.18297v1 Announce Type: new Abstract: AI judges, models that grade another system's output against a rubric, extend clinical review to the scale at which clinical AI writes. When teams use judge assessments to select a prompt or…

  124. Dancing Tokens on the Vocabulary Network from LLM Generation
    USarXiv cs.AI

    arXiv:2611.18248v1 Announce Type: new Abstract: Large language models (LLMs) generate text one token at a time, yet how a particular generation relates to the geometry learned in pre-training remains unclear. The fundamental explainability…

  125. Optimizing Drilling Trajectories in Multi-Material Structures Using Quantum Annealers
    USarXiv cs.AI

    arXiv:2611.18121v1 Announce Type: new Abstract: In this work we present a quantum-hybrid optimization approach for planning drilling trajectories on multi-material structures, with the objective of minimizing total energy consumption. The…

  126. Two-Clustering Regime of Token Dynamics in Causal Attention
    USarXiv cs.AI

    arXiv:2611.18088v1 Announce Type: new Abstract: We partially resolve Conjecture 1 of [Karagodin et al., Clustering in Causal Attention Masking (NeurIPS 2024)] and extend it to time-dependent parameter regimes. The conjecture predicts that…

  127. A$^3$: Adaptive Audio Agent Framework for Collaborative Edge Computing
    USarXiv cs.AI

    arXiv:2611.18064v1 Announce Type: new Abstract: Embedded audio agents need a nearby edge server for speed, but a server-dependent agent returns nothing when the link fails. We present A$^3$, an adaptive audio agent framework that places…

  128. Understanding the limits of snapshot-based fault diagnosis in distribution networks
    USarXiv cs.AI

    arXiv:2611.17998v1 Announce Type: new Abstract: Inverter-based resources weaken fault currents in distribution networks, and snapshot-based diagnosis, which classifies a fault from voltage phasors at a few buses at one instant, offers a…

  129. Interactive deep-learning based interface for brain tumour segmentation
    USarXiv cs.AI

    arXiv:2611.17983v1 Announce Type: new Abstract: Brain tumour segmentation from MR images is an integral part of their treatment, either through surgical resection or radiotherapy. Both applications require a high degree of precision that…

  130. TotalSegmentator Aorta Report: Automated Aortic Diameter Measurement at 11 Standardized Landmarks on CT Images
    USarXiv cs.AI

    arXiv:2611.17968v1 Announce Type: new Abstract: Purpose: To develop and evaluate a fully automated, open-source tool that measures aortic diameters at 11 standardized landmarks from the aortic annulus to the bifurcation on any CT image…

  131. Toward Interference-Agnostic RF Source Separation
    USarXiv cs.AI

    arXiv:2611.17950v1 Announce Type: new Abstract: Separating a radio-frequency signal of interest from a co-channel interference is mostly addressed with specialized separators, each trained for one pair of signal and interference types and…

  132. Measured from the feed: the unsustainable growth of AI video generation
    USarXiv cs.AI

    arXiv:2611.17938v1 Announce Type: new Abstract: Text-to-video generation runs at commercial scale, but its broader environmental impacts remain, to our knowledge, understudied. Providers rarely disclose video volumes or energy needs; we…

  133. OutfitterBench: Evaluating LLM Agents on Long-Horizon Resource Allocation and Commitment Coordination
    USarXiv cs.AI

    arXiv:2611.17933v1 Announce Type: new Abstract: In many long-horizon settings, an agent's own choices create future constraints: committing finite capacity establishes future obligations and forecloses later opportunities. Sustaining…

  134. Evaluating Large Language Models for Scientific Paper Search: Retrieval Success and Inference Latency on Mac Studio and NVIDIA A100
    USarXiv cs.AI

    arXiv:2611.17929v1 Announce Type: new Abstract: Scientific literature search can benefit from locally deployed LLM assistants, which reduce reliance on hosted model providers while giving research teams greater control over their data and…

  135. When Cooperation Must Change: Semantic Strategy Unlearning in Multi-Agent Reinforcement Learning
    USarXiv cs.AI

    arXiv:2611.17917v1 Announce Type: new Abstract: Cooperative multi-agent reinforcement learning (MARL) supports coordinated decision-making in distributed Web services, multi-robot control, and collaborative games through learned…

  136. SWIFT: Sample-Wise Instance Fine-Tuning of Time-Series Foundation Models for Forecasting
    USarXiv cs.AI

    arXiv:2611.17915v1 Announce Type: new Abstract: While current few-shot adaptation strategies for time series forecasting foundation models are predominantly dataset-centric, the necessity of obtaining an additional dataset for the…

  137. Soteria: Formally Verified Planning with Runtime Enforcement for Safe LLM Agents
    USarXiv cs.AI

    arXiv:2611.17880v1 Announce Type: new Abstract: LLM agents that execute multi-step tool calls must satisfy two objectives simultaneously: completing the user's task (utility) and conforming to domain policies and correctness constraints…

  138. OmniScore-Bench: Evaluating Vision-Language Models on Musical Score Reading
    USarXiv cs.AI

    arXiv:2611.17870v1 Announce Type: new Abstract: Vision-language models (VLMs) can answer questions about musical scores, but their ability to recognise notation, recover its structure and interpret its musical meaning requires separate…

  139. AI-aided Decision-Making in Education - Uncertainty, Explainability, and Human Responsibility
    USarXiv cs.AI

    arXiv:2611.17858v1 Announce Type: new Abstract: As Artificial Intelligence (AI)-based systems have become increasingly integrated into various aspects of everyday life, including education, understanding how these systems make decisions is…

  140. Is Abliteration Getting Harder? Not Uniformly: Refusal Removal by Direction-Based Ablation and ARA in Heretic on the Qwen3.8, Gemma 4, Olmo 3.1 and Mistral Small 3.2 LLMs Versus Their Predecessors
    USarXiv cs.AI

    arXiv:2611.17846v1 Announce Type: new Abstract: Abliteration, a fine-tuning-free weight edit, removes refusal from open-weight large language models (LLMs). The developer of Arbitrary-Rank Ablation (ARA), an optimisation-based variant…

  141. Causally Constrained and Uncertainty-Aware Cardiopulmonary Risk Prediction
    USarXiv cs.AI

    arXiv:2611.17819v1 Announce Type: new Abstract: Routine chest computed tomography (CT) captures pulmonary, cardiac, and vascular phenotypes that jointly stratify lung cancer and cardiovascular disease, the two leading causes of death in…

  142. Can Agents Keep Up with Time? An Empirical Analysis of Temporal Adaptation After a Conversational Hiatus
    USarXiv cs.AI

    arXiv:2611.17806v1 Announce Type: new Abstract: Long-horizon agents increasingly serve as persistent personal assistants across modern web platforms, managing schedules, applications, and multi-step plans over weeks or months. Users often…

  143. Robustness Evaluation of GPU Cloud Pricing Benchmarks under Incomplete Observations: A Controlled Missingness Study
    USarXiv cs.AI

    arXiv:2611.17796v1 Announce Type: new Abstract: Daily GPU rental-price benchmarks can fail when marketplace quotes are missing. This article stress-tests a fixed equal-weight median rule under four imposed masking mechanisms and four…

  144. TReC: Evaluating Move-Induced State Changes in Instruction-Conditioned Go
    USarXiv cs.AI

    arXiv:2611.17781v1 Announce Type: new Abstract: When a Go model predicts a move different from the expert's, coordinate distance measures how far the prediction lands from the reference. We introduce Transition Representation Consistency…

  145. Knowledge After Collapse: Context and Wording Shape Factual Accessibility in Language Models
    USarXiv cs.AI

    arXiv:2611.17763v1 Announce Type: new Abstract: When a language model accepts a false correction in a factual discussion, its next answer can depend on how the conversation continues. We introduce Knowledge Answer Divergence (KAD), an…

  146. Cross-Linguistic Acoustic Biomarkers of Cognitive Impairment Across Interactive and Single-Speaker Speech Tasks
    USarXiv cs.AI

    arXiv:2611.17710v1 Announce Type: new Abstract: Multilingual dementia detection is essential for scalable and equitable screening. In recent years, automated speech analysis has emerged as a promising approach for supporting dementia…

  147. Co-Sight Harness: Externalized Control Assets and Deterministic Enforcement for Long-Horizon Business Agents
    USarXiv cs.AI

    arXiv:2611.17696v1 Announce Type: new Abstract: Long-horizon business workflows fail in two ways: errors accumulate as a chain of activities lengthens, and the knowledge an activity needs is written in a different vocabulary from the…

  148. Aligned, Conflicting, or Unspecified? A Diagnostic Benchmark for Multimodal Time Series Forecasting
    USarXiv cs.AI

    arXiv:2611.17680v1 Announce Type: new Abstract: Artificial Intelligence of Things (AIoT) systems increasingly combine sensed time series with external text such as expert reports and news, yet how reliably models use text that agrees or…

  149. Learn then Bet: Learning Interpretable Rules for Reliable Agent Intervention
    USarXiv cs.AI

    arXiv:2611.17662v1 Announce Type: new Abstract: Autonomous agents now run for hundreds of steps with little or no human supervision. In practice, practitioners use runtime monitors such as turn limits and keyword filters to decide when a…

  150. LLM-Guided Cluster Lifecycle Management: Deciding When and How to Evolve Behavioral Clusters
    USarXiv cs.AI

    arXiv:2611.17616v1 Announce Type: new Abstract: Behavioral clusters learned over user or item interactions drift as behavior evolves: some fragment into sub-behaviors, some converge, some disappear, and new patterns emerge. Two paradigms…