AX·NEWS
Sign in

Artificial Intelligence

Artificial Intelligence news — Thursday 19 November 2026

150 stories from 5 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. Amoday "AI development speed adjustment...No relapse on business and correct pressure"
    KRAI Times (Korea) Translated from Korean

    Amorepacific’s CEO of Leo Amoday and Tropic recognized the entry to adjust the model development speed for AI safety immediately after securing funds from the company (IPO). After the listed company has been listed, it will prioritize AI…

  2. [Nov 19] Antropicy IPO Front OpenAI Founder Jin Jin Jin
    KRAI Times (Korea) Translated from Korean

    OpenAI’s Sam Altman CEO and Greg Brockman’s president attended ‘New York Times De North Summit’ held on 18th (local time). In recent years, Zenson Huang’s Nvidia retirement and the company’s work (IPO) and AI safety issues in Dagaon and…

  3. U.S., China Semiconductor Export Authority...Samsung·TSMC suspended some shipment
    KRAI Times (Korea) Translated from Korean

    Samsung Electronics and TSMC suspend chip shipments for Chinese customers while the US closed the existing export permit exception for some high-tech semiconductors supplied in China. As the product that could be supplied without…

  4. Acquired USD 2 trillion for the first day of listing on Nasdae...spaceX
    KRAI Times (Korea) Translated from Korean

    Antropic began trading at 34% higher than the first day of the Nassau listed on the U.S. Nassau’s Nassau’s Nassau’s Nassau’s Nassau. In the first day of the listing, the company’s company value was beyond the spaceX listed last June…

  5. The 0.95 Expectile Is Not the 95th Percentile
    USTowards Data Science

    Exact uniform math and 100 synthetic Python runs show when to target 95% coverage and when asymmetric squared-error costs call for the expectile. The post The 0.95 Expectile Is Not the 95th Percentile appeared first on Towards Data Science…

  6. How to Run Claude Code and Other Coding Agents in the Cloud 24/7
    USTowards Data Science

    Two setups: flip Claude Code's cloud environment on for the quick path, or rent a server for full control. The post How to Run Claude Code and Other Coding Agents in the Cloud 24/7 appeared first on Towards Data Science .

  7. How to Oversee AI Agents at Scale With a Three-Tier Architecture
    USTowards Data Science

    Separate runtime oversight from human review: an autonomous supervisor handles reversible fixes, while humans approve high-impact changes. The post How to Oversee AI Agents at Scale With a Three-Tier Architecture appeared first on Towards…

  8. The Download: a living liver, a billion-dollar battery deal, and AI hype
    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. A human liver has been kept alive outside the body for a record seven weeks Researchers in Australia…

  9. A human liver has been kept alive outside the body for a record seven weeks
    USMIT Technology Review Biotech

    Researchers in Australia claim they’ve kept a human liver alive, on life support, outside the body for seven weeks, opening up speculative new possibilities in biomedical engineering. The feat, whose details are pending publication, was…

  10. Analytics Engineering's Missing Layer Is Business Communication
    USTowards Data Science

    Why grains, joins, filters, and tests need business translations, and how a four-step framework keeps stakeholders from misreading metric changes. The post Analytics Engineering's Missing Layer Is Business Communication appeared first on…

  11. What a billion-dollar deal says about the US battery market
    USMIT Technology Review Biotech

    After a tough few years littered with some high-profile failures, there’s some positive news for next-generation batteries in the US. Sila just signed an agreement with Panasonic Energy, a global battery manufacturer, to supply…

  12. OpenAI improves AMD GPU design with AI... “1 year progress in 5 months”
    KRAI Times (Korea) Translated from Korean

    OpenAI is an AI model that optimizes the circuit design of AMD flagship GPUs, which has confirmed the improvement of up to 14% in power, performance and area (PPA) indicators. AMD engineers explained that there are already five months in…

  13. Google search Adds ‘Common conversation’ in AI mode... Save Records·Excluding Personalization
    KRAI Times (Korea) Translated from Korean

    Google has added the ‘emergency conversation’ function that does not reflect in chat history and personalization in AI mode of search services. You can choose whether the content you searched for is in record or not affect the…

  14. MS, openAI model supply expansion propelled to byte dance...Mi Congress description required
    KRAI Times (Korea) Translated from Korean

    Microsoft (MS) has been a report that is close to the agreement to add an AI cloud service of $500 billion (approx. 6650 billion) to the Chinese byte dance. Among the openAI to limit the model access of Chinese enterprises, the U.S…

  15. Meta announces new shares of shareholder litigation...AI Investment Fund Procurement Plan
    KRAI Times (Korea) Translated from Korean

    Meta has suspended the ‘Class C’ stock issue plan without the right to the lawsuit of investors. In the case of an investor in AI, we have been deploying the capital structure to maintain the governance of Markburg CEO. The Financial Times…

  16. AI Times·Personal order, "AI times, faith and reading books." . AI insights
    KRAI Times (Korea) Translated from Korean

    We are committed to the cooperation of the trustworthy "God and Wisdom" in AI-time and AI-information. The AI Times and Spiritualities have signed a partnership agreement for mutual cooperation at the Seoul Metropolitan Government’s…

  17. The company is located in the center of Seoul, Korea. The company is located in the center of Seoul, Korea
    KRAI Times (Korea) Translated from Korean

    Jeonnam-Gwangju has formulated AI data center. As a result, the company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s company’s…

  18. LG Group 2027 Executive Director, AIDC·Robot·Pickling
    KRAI Times (Korea) Translated from Korean

    LG Electronics has strengthened the AI-related sector in the 2027 executives in the 2027. LG Group opened the 19th affiliate board and held the 2027 executives. In this executive company, AI-related talents are invisible. LG Electronics…

  19. Developing chatGPT voice... Real-time text conversion·10 minutes recording limited waste paper
    KRAI Times (Korea) Translated from Korean

    OpenAI was able to check the converted sentences on the screen while improving the voice acceptance of ‘ChatGPT’. It is easy to enter a long question or description of the existing 10 minutes recording limit. OpenAI announced that ‘Live…

  20. HeggingFace, Coding Agent Custom Adapter Revealed...'Crew' Henando Rating 39%→65%
    KRAI Times (Korea) Translated from Korean

    HeggingFace revealed the LoRA adapter that has been added to a specific coding agent. Explanation of success in the 51 challenges assessments of ‘SWE-Venture Pro’ explained that the success rate was increased from 39.2% to 64.7% and the…

  21. Intel secures the first external customer in the advanced '18A-P' process...
    KRAI Times (Korea) Translated from Korean

    Intel announced that it has secured the first external customers at the ‘18A-P’, the performance improvement of the advanced semiconductor manufacturing process ‘18A’. Although it was accomplished in the Foundry business that wants to…

  22. U.S. Secretary-General “Prohibiting Envia High-Tech Chips to Strengthen AI Military Power”
    KRAI Times (Korea) Translated from Korean

    The report of the U.S. Congress Advisory Organization that promotes high-tech semiconductor securing in NVIDIA to strengthen AI-based military capabilities. China’s performance and manufacturing and software technology has remained…

  23. Gwangju MedicalAI Platform 45M Higher...Accommodation·Accommodation 465 medical records backup support
    KRAI Times (Korea) Translated from Korean

    We are committed to protecting and protecting your personal information. We are committed to protecting your personal information and protecting your personal information. 465 clinicians and hospitals are key to building a backup and…

  24. D-Information “Environmental GPU rental route connection to NVIDIA, Alibaba”
    KRAI Times (Korea) Translated from Korean

    Nvidia’s sales team connects Aliba to Alice Huang’s chip rental business, and after export control of the U.S., he appeared on how to borrow AI chips from Southeast Asia. This is the content that has been helped to secure the customer of…

  25. Google “AI Nose”... New York, LA, Sikago Educational District, Replacing Search Review
    KRAI Times (Korea) Translated from Korean

    The American New York City and Los Angeles, Chicago Education are reviewing the implementation of changing Google search to another search engine. In the search for a student, a half of the Google’s policy is made with a real service…

  26. Solar Panel·ESS Equipment Detergently Reduced...Ecuador New Regulations
    KRAI Times (Korea) Translated from Korean

    Ecuador began to apply additional value tax (IVA) and 0% exemption tax (ISD) to designated power facilities such as solar panels and energy storage devices (ESS). The new regulations specified in the total number of items for 55 items for…

  27. In Romania, the roof photovoltaic in-house electric fuel is applied to other households
    KRAI Times (Korea) Translated from Korean

    In Romania, it is possible to use the electricity bills produced by solar power generation in other homes or business areas. The new system, which distributes settlements to multiple addresses with the same power production and…

  28. Launched 'Balconic Solar' system for the Netherlands and outlet type...Hiyuan 111 39
    KRAI Times (Korea) Translated from Korean

    The Netherlands has a small solar system installed on the balcony and required the Dutch government to secure connection to the home electrical circuit. In the 2027, the agreement that urges to establish the previous related standards was…

  29. How OpenAI’s Performance Factory keeps ChatGPT fast
    USOpenAI Blog

    Learn how OpenAI built the Performance Factory, a system of Codex-powered workflows that finds and fixes performance issues to help keep ChatGPT fast.

  30. Nvidia Launches ‘RTX Spark’ Small Desktop Global...Starting a Pre-book in Korea
    KRAI Times (Korea) Translated from Korean

    Nvidia has released a small desktop for the “RTX Spark” that allows local AI agents to run 24 hours. AsuS·Del·MSI, etc., we will introduce our products, and we will be ready to make a reservation from 19th to 19th in Korea. The RTX Spark…

  31. Trump Executive, China-based AI-enabled fact public duty review in the US industry
    KRAI Times (Korea) Translated from Korean

    The Trump administration is reviewing the implementation of an obligation to disclose the use of Chinese AI model in the US industry. It is a remedy to reduce the dependence on Chinese model through reputable burdens due to publicity than…

  32. “I’ve been investing in the industry”
    KRAI Times (Korea) Translated from Korean

    The Chairman of the Soft Bank of the Hands-on has become a mistake that the AI company has not invested in an antique. For the reason that we are focusing on openAI, we have abandoned investments against the competitors, but we’re going…

  33. AI infra of the year 3200.. .' 'Membration' accelerates expenditure
    KRAI Times (Korea) Translated from Korean

    In 2027, the world’s “AI infrastructure spending” has been viewed that it will take $200 trillion (approximately 3200 trillion KRW) to account for 59% of total AI spending. Among AI infrastructure expenditures, the “AI Optimization Server”…

  34. North Korea, AI Chatbot ‘Gyeong Hye-Sam’ released...Notification of questions banned·Overseas model
    KRAI Times (Korea) Translated from Korean

    North Korea has published AI chatbot ‘Sam Hye’ to answer questions in the field of science and technology. We have a similar interactive interface with ChatGPT, but we are prohibiting political and social issues, and we are likely to take…

  35. OpenAI 'GPT-6.2' released after some suspicious...Altman "Active improvement required"
    KRAI Times (Korea) Translated from Korean

    'GPT-6.2 Astra' and 'Sol', which has been released by OpenAI for high benchmark performance. The ability to perform complex tasks is outstanding, but it doesn’t matter if you’re looking for a real-time experience, such as not to finish a…

  36. Added "code mode" to the closure code... Handle multiple tool calls once
    KRAI Times (Korea) Translated from Korean

    Antropic has added the code mode to connect multiple development tools to JavaScript. I had to call one tool in the existing, but now I can handle multiple tasks at once and manage them inside the code until the middle result, reducing…

  37. Kim Ki-bang Kim Huigyeong's entertainment room, appeared hello AI Era Human Value Rejoice
    KRAI Times (Korea) Translated from Korean

    Kim Ki-bang Kim Huigyeong's entertainment room, appeared hello AI means that it is interested in human characteristics and values that are distinguished by the machine while entering the human unique area, such as writing and creation…

  38. Gwangju Biennale AI Commentary... How to Use the Museum of Art
    KRAI Times (Korea) Translated from Korean

    In Gwangju Biennale, more than 20 million questions were received for AI work. Singapore, the United States, and the UK’s major art galleries are also introduced or tested by AI-enabled communications services, from the description of the…

  39. Google releases an agent "Stitch Loop" which improves software after launch
    KRAI Times (Korea) Translated from Korean

    Google has released the AI agent ‘Stitch Loop’ which continuously finds and improves security vulnerabilities and performance degradation, even after the software launch. It is characterized by analyzing real operating data and repeating…

  40. Datamaker, AI solution ‘Snaps’ businessization performance, the Minister’s Award
    KRAI Times (Korea) Translated from Korean

    AI Professional Datamaker (CEO, Lee-kuk) announced that it was awarded the Minister of Small and Medium-sized Venture Business Administration by the Ministry of SMEs R&D Excellence and 50th Business Prize. Datamaker has been producted with…

  41. ReactBench-AD: Do Driving VLAs and World-Action Models React to Imminent Hazards in Their Path?
    USarXiv cs.AI

    arXiv:2611.20011v2 Announce Type: replace-cross Abstract: Vision-Language-Action models (VLAs) and World-Action Models (WAMs) have emerged in autonomous driving with the promise of strong generalization to long-tail scenarios. Although…

  42. Typed Knowledge over Constructive Institutions: A Proof-Relevant Type Theory for Verified Science
    USarXiv cs.AI

    arXiv:2611.15108v3 Announce Type: replace-cross Abstract: Mathematical proof assistants such as Coq, Agda, and Lean 4 build proof irrelevance into definitional conversion ($p \equiv q$ for $p, q : P$ with $P : Prop$), erasing which…

  43. How Path Integral Diffusion Decides a Finite Spin Glass
    USarXiv cs.AI

    arXiv:2611.11826v2 Announce Type: replace-cross Abstract: When an AI diffusion-type generator samples a frustrated spin glass, which macroscopic alternatives does it decide first, and when? We study this for bridge-diffusion models of the…

  44. The Digital-Energy Recomposition of the Kitchin Cycle: Exploratory Evidence from U.S. IT and Irish Data-Centre Series (1995-2025)
    USarXiv cs.AI

    arXiv:2610.33122v2 Announce Type: replace-cross Abstract: Using quarterly FRED and CSO Ireland data over 1995Q1-2025Q4, we examine whether U.S.\ IT investment and Irish data-centre electricity are recomposing the Kitchin cycle around…

  45. Hearsay: Who Can Audit the Record a Deployed Agent Harness Writes?
    USarXiv cs.AI

    arXiv:2609.32495v2 Announce Type: replace-cross Abstract: The record of an agent's run is written by the harness that drives the agent. When the run is later disputed, the auditor was not there and has only this record. We ask whether…

  46. OracleZoom: Reference-Constrained Recursive Image Super-Resolution Beyond Ground Truth
    USarXiv cs.AI

    arXiv:2609.06490v2 Announce Type: replace-cross Abstract: Recursive image Super-Resolution (SR) reaches extreme magnification by feeding each prediction back into the same model, like zooming into an image again and again. Real ground…

  47. Autonomous discovery of new structure-plausibility laws for explainable and rapid crystal diagnosis and screening
    USarXiv cs.AI

    arXiv:2609.01209v2 Announce Type: replace-cross Abstract: Generative models and autonomous agents now propose candidate inorganic crystals by the million, yet most screens test little beyond a minimum-distance cutoff and give no chemical…

  48. SciFigQual-Bench: Scientific Figure Quality Assessment with Manuscript-Linked Context
    USarXiv cs.AI

    arXiv:2607.27084v4 Announce Type: replace-cross Abstract: Scientific figure quality is bound to the manuscript: a crop can be visually clear and still contradict its caption or the paragraph that cites it. Most image quality assessment and…

  49. Operational Proto-Introspection in Looped Language Models: Reading Correctness Before the Answer
    USarXiv cs.AI

    arXiv:2607.18553v4 Announce Type: replace-cross Abstract: Can one tell, while a language model reasons, whether its answer will be right? Looped models are a good place to ask: they run the same layers several times per token, leaving one…

  50. A Geometric Coding Theorem
    USarXiv cs.AI

    arXiv:2607.13796v3 Announce Type: replace-cross Abstract: This paper establishes a direct analogue of the classical Coding Theorem in the setting of symmetry groups. We consider computable bijections on the set of binary strings and define…

  51. SharQ: Bridging Activation Sparsity and FP4 Quantization for LLM Inference
    USarXiv cs.AI

    arXiv:2606.26587v2 Announce Type: replace-cross Abstract: Low-bit floating-point formats and semi-structured sparsity are increasingly supported by modern accelerators, yet combining them for LLM activation compression remains challenging…

  52. Robustness of Text-Only Adapted Speech LLMs for Spoken Semantic Parsing
    USarXiv cs.AI

    arXiv:2606.21453v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly used to generate structured outputs from natural language instructions, including semantic frames and function calls. When the input is…

  53. Rational Sparse Autoencoder
    USarXiv cs.AI

    arXiv:2606.14990v3 Announce Type: replace-cross Abstract: Sparse autoencoders (SAEs) are standard tools for mechanistic interpretability, but current SAE families are constrained by fixed encoder nonlinearities such as ReLU, JumpReLU, and…

  54. What competing constraints reveal about representation learning
    USarXiv cs.AI

    arXiv:2605.15995v5 Announce Type: replace-cross Abstract: Learning latent states from temporal, multimodal, and partially observed data requires decisions about what information representations should retain, discard, and organize. I argue…

  55. LLM-Assisted Reranking to Operationalize Nuanced Objectives in Recommender Systems
    USarXiv cs.AI

    arXiv:2606.02883v2 Announce Type: replace-cross Abstract: Recommender systems shape what information we see, raising concerns about filter bubbles, radicalization, polarization, and social inequality. Large language models (LLMs) enable…

  56. Measuring What Matters: Synthetic Benchmarks for Concept Bottleneck Models
    USarXiv cs.AI

    arXiv:2606.04326v2 Announce Type: replace-cross Abstract: Concept bottleneck models predict using high-level concepts detected in inputs. Although these models provide a simple way to reap benefits from interpretability, they are hard to…

  57. MedExAgent: Training LLM Agents to Ask, Examine, and Diagnose in Noisy Clinical Environments
    USarXiv cs.AI

    arXiv:2605.07058v2 Announce Type: replace-cross Abstract: Real-world clinical diagnosis is a complex process in which the doctor is required to obtain information from both interaction with the patient and conducting medical exams…

  58. HorizonBench: Long-Horizon Personalization with Evolving Preferences
    USarXiv cs.AI

    arXiv:2604.17283v2 Announce Type: replace-cross Abstract: Personal assistants now talk with the same user for months, and users' preferences change as their lives change; an assistant that keeps serving a preference the user has outgrown…

  59. Reading Human Preference from Looped Transformer Internal States
    USarXiv cs.AI

    arXiv:2604.09870v3 Announce Type: replace-cross Abstract: Show a frozen language model two replies to the same request, one of which a person preferred. Can a small probe on the model's hidden states tell which? On Ouro-2.6B-Thinking, a…

  60. HYPER: Hypothesis-Guided Planning for Evidence-Seeking Retrieval
    USarXiv cs.AI

    arXiv:2603.19008v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) improves Large Language Models (LLMs) by grounding generation in external, non-parametric knowledge. However, when a task requires choosing…

  61. SHAMISA: SHAped Modeling of Implicit Structural Associations for Self-supervised No-Reference Image Quality Assessment
    USarXiv cs.AI

    arXiv:2603.13669v3 Announce Type: replace-cross Abstract: No-Reference Image Quality Assessment (NR-IQA) aims to estimate perceptual quality without access to a pristine reference. Learning an NR-IQA model faces a fundamental bottleneck…

  62. Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems
    USarXiv cs.AI

    arXiv:2602.15198v3 Announce Type: replace-cross Abstract: Multi-agent systems, where LLM agents communicate through free-form language, enable sophisticated coordination for solving complex cooperative tasks. This surfaces a unique safety…

  63. Generating Adversarial Events: A Motion-Aware Point Cloud Framework
    USarXiv cs.AI

    arXiv:2602.08230v2 Announce Type: replace-cross Abstract: Event cameras have been widely adopted in safety-critical domains such as autonomous driving, robotics, and human-computer interaction. A pressing challenge arises from the…

  64. CoreQ: Mismatch-Corrected Successive Rounding for LLM Quantization
    USarXiv cs.AI

    arXiv:2602.05902v3 Announce Type: replace-cross Abstract: Post-training quantization (PTQ) enables efficient deployment of large language models by mapping pretrained weights to low-bit formats without retraining, typically using a small…

  65. MonitorVLM:A Vision Language Framework for Safety Violation Detection in Mining Operations
    USarXiv cs.AI

    arXiv:2510.03666v3 Announce Type: replace-cross Abstract: Industrial accidents, particularly in high-risk domains such as surface and underground mining, are frequently caused by unsafe worker behaviors. Traditional manual inspection…

  66. Improving Outdoor Multi-cell Fingerprinting-based Positioning via Mobile Data Augmentation
    USarXiv cs.AI

    arXiv:2509.19405v3 Announce Type: replace-cross Abstract: Accurate outdoor positioning in cellular networks is hindered by sparse, heterogeneous measurement collections and the high cost of exhaustive site surveys. This paper introduces a…

  67. FedS2R: One-Shot Federated Domain Generalization for Synthetic-to-Real Semantic Segmentation in Autonomous Driving
    USarXiv cs.AI

    arXiv:2507.19881v4 Announce Type: replace-cross Abstract: Federated domain generalization has shown promising progress in image classification by enabling collaborative training across multiple clients without sharing raw data. However…

  68. Reinforcement Learning with Reference-Based Rewards for Language Model Reasoning and Alignment
    USarXiv cs.AI

    arXiv:2504.09895v4 Announce Type: replace-cross Abstract: Reinforcement learning (RL) powers two major lines of large language model (LLM) post-training: reasoning, which elicits deliberate, verifiable problem solving, and alignment, which…

  69. Mixtures of In-Context Learners
    USarXiv cs.AI

    arXiv:2411.02830v2 Announce Type: replace-cross Abstract: In-context learning (ICL) adapts LLMs by providing demonstrations without fine-tuning the model parameters; however, it does not differentiate between demonstrations and…

  70. Design Optimization of Nuclear Fusion Reactor through Deep Reinforcement Learning
    USarXiv cs.AI

    arXiv:2409.08231v2 Announce Type: replace-cross Abstract: In this research, we investigate deep reinforcement learning (DRL) as a practical tool for constrained, multi-objective optimization for tokamak fusion reactor designs. The task is…

  71. Co-Pilot for Health: Personalized Algorithmic AI Nudging to Improve Health Outcomes
    USarXiv cs.AI

    arXiv:2401.10816v3 Announce Type: replace-cross Abstract: The ability to shape health behaviors of large populations automatically, across wearable types and disease conditions at scale has tremendous potential to improve global health…

  72. Agentic On-Policy Distillation Implicitly Reshapes LLMs' ReAct Patterns
    USarXiv cs.AI

    arXiv:2611.19244v2 Announce Type: replace Abstract: On-policy distillation (OPD) transfers specialized capabilities to language-model agents, but its cross-domain effects on interaction remain unclear. We investigate how OPD reshapes ReAct…

  73. Query-Driven Learning for Lifted Reasoning
    USarXiv cs.AI

    arXiv:2610.34747v2 Announce Type: replace Abstract: Learning and reasoning are both essential to intelligent agents, but have traditionally been treated as two separate stages: first learning a knowledge base from data (or taking one as…

  74. Linearized 2-Simplicial Attention
    USarXiv cs.AI

    arXiv:2608.09307v2 Announce Type: replace Abstract: We present a linearized form of 2-simplicial attention by rewriting the trilinear score as an inner product between a composite query and a key, so that the sum over keys takes the same…

  75. Beyond Routing Weights: Faithful Response-Level Interpretation of Mixture-of-Experts Reward Models via Contribution Contrast
    USarXiv cs.AI

    arXiv:2608.06400v2 Announce Type: replace Abstract: Reward models are central to learning from human preferences, yet identifying what drives their predictions remains challenging. Recent sparse Mixture-of-Experts (MoE) reward models seek…

  76. JUDOR: CTDE-Based Multi-Agent Reinforcement Learning for the Co-Optimization of Joint UAV Motion and Decentralized Opportunistic Routing in Delay-Tolerant Network
    USarXiv cs.AI

    arXiv:2608.04590v2 Announce Type: replace Abstract: The growing deployment of delay-tolerant networks (DTNs) has made store-carry-forward (SCF) communication indispensable for operating under sparse intermittent connectivity. However…

  77. CoNav-UAV: Cooperative Dual-Altitude Aerial Navigation via Stackelberg Learning
    USarXiv cs.AI

    arXiv:2608.01802v2 Announce Type: replace Abstract: Aerial search missions often require an unmanned aerial vehicle (UAV) to locate a target from only a brief description of its appearance and surroundings, a task known as target-oriented…

  78. Making Mathematical Knowledge Explainable, Accessible and Interoperable Through Large Language Model Integration
    USarXiv cs.AI

    arXiv:2607.24512v2 Announce Type: replace Abstract: Mathematical models are central to formalizing research problems, yet their documentation often falls short of FAIR principles. Knowledge bases such as the Mathematical Model Database…

  79. Disagreement-Guided Agentic Data Curation for Multimodal Safety
    USarXiv cs.AI

    arXiv:2607.14256v2 Announce Type: replace Abstract: Written safety policies in multimodal settings define categories whose visual boundaries span many contexts. We study disagreement as a signal for identifying useful synthetic boundary…

  80. AgentAbstain: Do LLM Agents Know When Not to Act?
    USarXiv cs.AI

    arXiv:2607.10059v2 Announce Type: replace Abstract: Agent systems based on large language models (LLMs) are increasingly deployed for autonomous tasks, yet existing evaluations mostly focus on task success rather than whether agents know…

  81. ElephantAgent: Contextual State Continuity in Agentic Systems
    USarXiv cs.AI

    arXiv:2607.01919v2 Announce Type: replace Abstract: Agentic systems enhance their capabilities by invoking external tools and maintaining persistent memory. However, recent tool and memory poisoning attacks can covertly bias agent…

  82. When Does Learning to Stop Help? Configuring Risk-Controlled Early Exit for Reasoning-Model Services
    USarXiv cs.AI

    arXiv:2606.30852v3 Announce Type: replace Abstract: Early exit lowers the cost of serving a reasoning model by stopping its reasoning once a provisional answer looks final. An operator must choose between a learned stopping rule and one…

  83. GA-VINO: A Geometry-Aware Variational Physics-informed Neural Operator for Mindlin-Reissner Plates
    USarXiv cs.AI

    arXiv:2606.16624v3 Announce Type: replace Abstract: Plate and shell structures are widely used in engineering fields. Rapid response prediction for such structures under complex geometries, heterogeneous materials, and varying loads is…

  84. CaMBRAIN: Real-time, Continuous EEG Inference with Causal State Space Models
    USarXiv cs.AI

    arXiv:2605.28792v3 Announce Type: replace Abstract: Electroencephalography (EEG) is a critical, non-invasive method to monitor electrical brain activity. EEGs can span anywhere from a couple seconds to multiple hours, posing a major hurdle…

  85. Prior Knowledge or Search? A Study of LLM Agents in Hardware-Aware Code Optimization
    USarXiv cs.AI

    arXiv:2605.19782v2 Announce Type: replace Abstract: LLM discovery and optimization systems are increasingly applied across domains, implementing a common propose-evaluate-revise loop. Such optimization or discovery progresses via context…

  86. CXReasonAgent: Evidence-Grounded Diagnostic Reasoning Agent for Chest X-rays
    USarXiv cs.AI

    arXiv:2602.23276v3 Announce Type: replace Abstract: Chest X-ray plays a central role in thoracic diagnosis, and its interpretation inherently requires multi-step, evidence-grounded reasoning. However, large vision-language models (LVLMs)…

  87. Games played by Exponential Weights Algorithms
    USarXiv cs.AI

    arXiv:2407.06676v2 Announce Type: replace Abstract: This paper studies the last-iterate convergence properties of the exponential weights algorithm with constant learning rates. We consider a repeated interaction in discrete time, where…

  88. Human-AI Interactions and Societal Pitfalls
    USarXiv cs.AI

    arXiv:2309.10448v5 Announce Type: replace Abstract: When working with generative artificial intelligence (AI), users may see productivity gains, but the AI-generated content may not match their preferences exactly. To study this effect, we…

  89. Mapping the AI Supply Chain: Dependencies, Vulnerabilities, and Market Concentration in the AI Ecosystem
    USarXiv cs.AI

    arXiv:2611.24927v1 Announce Type: cross Abstract: AI systems are built across many organizations that contribute models, data, compute, capital, talent, and other assets. While the layers of the AI stack are relatively well understood, the…

  90. Who are you? The Personal Imitation Game
    USarXiv cs.AI

    arXiv:2611.24888v1 Announce Type: cross Abstract: Large Language Models (LLMs) now perform strongly on Alan Turing's 1950 Imitation Game, so we propose a harder challenge, the Personal Imitation Game, which asks whether a machine can…

  91. Auditing black-box clinical prediction models with a large language model through queries alone: a blinded test with planted flaws
    USarXiv cs.AI

    arXiv:2611.24877v1 Announce Type: cross Abstract: Large language models (LLMs) are used to explain clinical prediction models, on the promise that they can say what a model computes, find its faults and even reveal new knowledge in it. We…

  92. Does the Action Matter? Probing Action-Conditioned Semantic Forecasting as a World Model for Autonomous Driving
    USarXiv cs.AI

    arXiv:2611.24840v1 Announce Type: cross Abstract: World models are increasingly used in autonomous driving to predict how a scene evolves under the ego vehicle's actions, supporting counterfactual reasoning and model-based planning…

  93. LAaDT-DSS: decision support for non-experts on rooftop PV, incentive and grid simulations under Turkish tariff rules
    USarXiv cs.AI

    arXiv:2611.24810v1 Announce Type: cross Abstract: Rooftop photovoltaics raise three questions: should a household invest, which incentive costs the state least, and does the feeder stay within limits? Those who ask are rarely modellers. We…

  94. Bias-Only Training Dynamics in ReLU Networks
    USarXiv cs.AI

    arXiv:2611.24797v1 Announce Type: cross Abstract: Neural networks with randomly initialized, fixed weights and trainable biases are universal approximators, yet the dynamics of bias-only training remain largely unexplored. We study these…

  95. Isabelle Pool Server: a pooled verification service for prover training and multi-agent proving
    USarXiv cs.AI

    arXiv:2611.24788v1 Announce Type: cross Abstract: Isabelle Pool Server (IPS), a verification service with a pool of warm headless Isabelle sessions behind an HTTP API, with session leasing and reclamation, memory-aware admission control…

  96. LLM-Guided Cyber-Physical Attack Path Identification in Smart Grids
    USarXiv cs.AI

    arXiv:2611.24765v1 Announce Type: cross Abstract: Identifying high-impact attack paths in cyber-physical power systems is critical for improving grid resilience. Existing methods typically rely on either structural topology analysis or…

  97. Learning Trajectories Leave Decodable Traces in Neural Network Weight Space and Internal Representations
    USarXiv cs.AI

    arXiv:2611.24782v1 Announce Type: cross Abstract: Neural networks trained on the same task develop related weight-space structures, yet remain distinguishable after learning. Do these differences merely reflect alternative solutions, or do…

  98. How Generative Models Generate: A Concise Guide to the Generative AI Landscape
    USarXiv cs.AI

    arXiv:2611.24728v1 Announce Type: cross Abstract: The rapidly expanding generative AI landscape brings together methods that differ in architecture, learning objective, mathematical representation and sampling procedure. A concise…

  99. Openhagen Worlds: Calibrated Environments for Measuring and Training Scientific Discovery Agents
    USarXiv cs.AI

    arXiv:2611.24711v1 Announce Type: cross Abstract: Agents that do science must choose informative experiments and reason from their results, but most discovery benchmarks cannot separate that skill from recall, unstated scoring rules or…

  100. When Does Continual Pre-Training Help? A Controlled Study of Dialectal Adaptation for Qu\'ebec French
    USarXiv cs.AI

    arXiv:2611.24699v1 Announce Type: cross Abstract: Continual pre-training (CPT) is widely used to adapt pretrained language models to dialectal and non-standard language varieties, yet it remains unclear when this additional training is…

  101. Choosing the Right Representations: Task-Aware Fusion for Multimodal Materials Language Models
    USarXiv cs.AI

    arXiv:2611.24641v1 Announce Type: cross Abstract: Multimodal language models for materials science rely on effective mechanisms for incorporating three-dimensional structural representations. MatterChat addresses this challenge using a…

  102. Hypnos: A Dual-Phase End-to-End Platform for Scalable Continual Learning at the Edge
    USarXiv cs.AI

    arXiv:2611.24637v1 Announce Type: cross Abstract: Continual learning (CL) enables neural networks to adapt to non-stationary data streams by mitigating catastrophic forgetting, offering a path to robust, private, and adaptable edge…

  103. Clover-35B-A3B-Preview Technical Report
    USarXiv cs.AI

    arXiv:2611.24636v1 Announce Type: cross Abstract: We introduce Clover-35B-A3B-Preview (Clover), a mixture-of-experts diffusion language model with 35 billion total parameters and 3 billion activated parameters, supporting thinking-mode…

  104. Negotiating Control and Accountability among Stakeholders in High-Risk AI Use Cases: Development and Evaluation of the Method ALIGN
    USarXiv cs.AI

    arXiv:2611.24609v1 Announce Type: cross Abstract: Characteristics of Artificial Intelligence (AI) severely restrict human control. This creates new challenges for socio-technical systems design especially in safety-critical domains, where…

  105. Chain-of-Thought Hidden States Reveal Task-Solving Directions for Direct Answers
    USarXiv cs.AI

    arXiv:2611.24598v1 Announce Type: cross Abstract: Chain-of-Thought (CoT) prompting improves the performance of Large Language Models by eliciting intermediate reasoning steps. However, CoT is far from perfect: on some tasks, models perform…

  106. Do Frontier Model Ingredients Add Up? A Replicated Factorial Study of Small Language Models
    USarXiv cs.AI

    arXiv:2611.24540v1 Announce Type: cross Abstract: Building a small language model requires choosing which published improvements to combine. We compare five changes in the same model, using the same data and training budget. We test…

  107. Harvesting Expert-Parallel Bubbles for Fine-Tuning
    USarXiv cs.AI

    arXiv:2611.24496v1 Announce Type: cross Abstract: GPUs are expensive, yet latency-critical large language model (LLM) inference leaves part of them idle. Colocation reclaims this capacity by running best-effort work, such as fine-tuning…

  108. Probing agent overeagerness in realistic settings with continuation audits
    USarXiv cs.AI

    arXiv:2611.24491v1 Announce Type: cross Abstract: Frontier coding agents increasingly exhibit overeagerness, taking surprising, sometimes reckless actions beyond a user's mandate --- reported in recent system cards from frontier model…

  109. Mitigating fix reversal: Counterfactual constraints for function-level vulnerability detection
    USarXiv cs.AI

    arXiv:2611.24489v1 Announce Type: cross Abstract: Function-level vulnerability detectors often exhibit fix reversal, assigning a higher vulnerability probability to a fixed version than to its vulnerable counterpart. Fixes usually lengthen…

  110. Harmonic Circuit: Chord Progression Lies in a Sparse Principal Information Flow in Symbolic Music Models
    USarXiv cs.AI

    arXiv:2611.24453v1 Announce Type: cross Abstract: Mechanistic Interpretability (MI) seeks to decode neural network computations into human-understandable logic. A core methodology in MI is circuit discovery: identifying the information…

  111. Keeping Shielded Data-Driven Controllers Safe Under System Evolution via Model Consistency
    USarXiv cs.AI

    arXiv:2611.24428v1 Announce Type: cross Abstract: Cyber-physical systems (CPS) increasingly contain data-driven controllers, which raise safety concerns given their black-box nature. A common mitigation is to enforce system safety via a…

  112. Temporal Rank in Recurrent Neural Networks
    USarXiv cs.AI

    arXiv:2611.24394v1 Announce Type: cross Abstract: Recurrent Neural Networks (RNNs) are a natural framework for temporal modelling, yet are often dismissed due to their sequential, non-parallelisable computation and difficulty in capturing…

  113. LLM-Assisted Multi-Fidelity Bayesian Optimization: Fusing Expert Knowledge with Numerical Data
    USarXiv cs.AI

    arXiv:2611.24377v1 Announce Type: cross Abstract: Multi-fidelity Bayesian optimization (MFBO) reduces engineering design cost by combining inexpensive low-fidelity (LF) models with a few accurate high-fidelity (HF) evaluations and…

  114. Towards Pose-Robust Few-Shot Vision-Language-Action Models
    USarXiv cs.AI

    arXiv:2611.24360v1 Announce Type: cross Abstract: Fine-tuning a vision-language-action (VLA) model on a few dozen demonstrations is the standard way to put a manipulation policy on a robot, yet the resulting policy silently assumes that…

  115. OCTR2A: Real-Time RGB-to-BEV Trajectory Projection and Forecasting at Urban Intersections
    USarXiv cs.AI

    arXiv:2611.24352v1 Announce Type: cross Abstract: Trajectory forecasting at urban intersections is often studied assuming that reliable bird's-eye-view (BEV) motion histories are available, which in practice requires costly resources such…

  116. Can Weight-Space Programming Alter Learned Ethical Priorities?
    USarXiv cs.AI

    arXiv:2611.24330v1 Announce Type: cross Abstract: We investigate whether a weight edit can alter how a model trades off honoring a commitment against benefiting more people. In a concept-supervised text model, we alter fitted choice…

  117. What Survives the Shift from Text to Pixels? Conditional Protection against Prompt Injection in Image-Rendered Documents
    USarXiv cs.AI

    arXiv:2611.24231v1 Announce Type: cross Abstract: Multimodal language models increasingly answer questions over external documents, exposing them to indirect prompt injection: commands planted in untrusted content to hijack the response…

  118. Towards Reliable Scaling Laws for Tabular Foundation Models
    USarXiv cs.AI

    arXiv:2611.24214v1 Announce Type: cross Abstract: Scaling laws are widely used to predict foundation model performance and guide compute allocation. However, tabular foundation models (TFMs) differ from conventional foundation models in…

  119. Conditional What-If Generation for Radio Access Network Energy under Rare and Unseen Contexts
    USarXiv cs.AI

    arXiv:2611.24208v1 Announce Type: cross Abstract: Energy-saving mechanisms in RANs must balance energy reduction and network performance under varying traffic conditions. Conditional generative models support what-if analysis by generating…

  120. Worth Another Look:Constrained World-Model Planning for Visual Inspection
    USarXiv cs.AI

    arXiv:2611.24187v1 Announce Type: cross Abstract: Visual inspection in field robotics and environmental monitoring must localize multiple targets under strict compute and sensing caps. Exhaustive sweeps waste calls on absent classes, while…

  121. CounselTraj: Synthesizing Process-Aware Psychological Counseling Dialogues via Dialogue-Act Trajectory Planning
    USarXiv cs.AI

    arXiv:2611.24168v1 Announce Type: cross Abstract: The growing demand for psychological counseling has motivated the development of LLM-based systems that support human counselors, but training such systems requires large-scale…

  122. Once Upon a Bias: Gender Stereotyping in LLM-Generated Children's Stories
    USarXiv cs.AI

    arXiv:2611.24156v1 Announce Type: cross Abstract: Large language models are increasingly used to create stories for children. Since bedtime stories can shape children's perceptions of social roles, this raises concerns about whether LLMs…

  123. Inherited Adapter Experts for Replay-Free Domain-Incremental Learning
    USarXiv cs.AI

    arXiv:2611.24145v1 Announce Type: cross Abstract: Replay-free domain-incremental learning trains a model on the same classes in successive domains without keeping earlier images. Per-domain learners freeze a pre-trained ViT and train an…

  124. How Well Do Generalist CT Foundation Models Differentiate Liver Tumors?
    USarXiv cs.AI

    arXiv:2611.24122v1 Announce Type: cross Abstract: Generalist CT foundation models detect many findings with high performance, but their ability to differentiate detected lesions remains largely unevaluated. This retrospective study…

  125. The 7th International Verification of Neural Networks Competition (VNN-COMP 2026): Summary and Results
    USarXiv cs.AI

    arXiv:2611.24114v1 Announce Type: cross Abstract: This report summarizes the seventh International Verification of Neural Networks Com- petition (VNN-COMP 2026), held as part of the Symposium on AI Verification (SAIV) co-located with the…

  126. APS: Adaptive Partitioning-Based Frame Selection for Long Video Understanding
    USarXiv cs.AI

    arXiv:2611.24079v1 Announce Type: cross Abstract: Recent advances in large vision-language models (LVLMs) have significantly improved video understanding capabilities, yet processing long videos remains challenging due to the substantial…

  127. Replay, Don't Resample: Off-Policy Consolidation of Long-Horizon RL Experts
    USarXiv cs.AI

    arXiv:2611.24059v1 Announce Type: cross Abstract: Frontier post-training follows a common recipe: train one expert per domain with reinforcement learning (RL), then consolidate the experts into one model via multi-teacher on-policy…

  128. From Data to Valves: A Self-Learning, Production-Grade Edge Stack for Sustainable HVAC Control
    USarXiv cs.AI

    arXiv:2611.24052v1 Announce Type: cross Abstract: Learned HVAC controllers rarely leave the simulator, and those that do are typically trained offline, frozen at deployment and retrained from the cloud. We present a production stack in…

  129. FormalFixer: Intent-Aligned Automated Hardware Repair via Joint Refinement of Specifications, Assertions, and RTL Implementations
    USarXiv cs.AI

    arXiv:2611.24037v1 Announce Type: cross Abstract: Functional errors in hardware designs can compromise correctness and security, and must be fixed before tape-out. Formal property verification (FPV) provides strong guarantees by proving or…

  130. Intent Contracts: Taking Tools Away from Agents
    USarXiv cs.AI

    arXiv:2611.24025v1 Announce Type: cross Abstract: Intent contracts are an authorization interface for tool-using LLM agents in which the planning model holds no tools. For each step it submits a short logged contract: a goal, a verbatim…

  131. ProtoBridge-CLIP: Observation-Masked Finding Prototypes for Medical Image--Report Representation Learning
    USarXiv cs.AI

    arXiv:2611.24002v1 Announce Type: cross Abstract: Radiology reports distinguish affirmed findings, explicit negations, and unmentioned concepts, but conventional global image--report contrastive learning often compresses these different…

  132. A Mechanistic Analysis of In-Context Learning and Out-of-Context Hallucination
    USarXiv cs.AI

    arXiv:2611.24000v1 Announce Type: cross Abstract: Large language models can infer and follow specific rules through in-context learning (ICL), yet may fail to remain faithful to provided evidence, giving rise to out-of-context (OOC)…

  133. TOMAS: Topology-Aware Operation Modeling and Adaptive Staging for Egocentric Hand Forecasting
    USarXiv cs.AI

    arXiv:2611.23970v1 Announce Type: cross Abstract: Forecasting future hand movements from egocentric videos plays an important role in understanding human intentions and hand-object interactions. However, different manipulation intentions…

  134. SEE: Unlocking the Hidden Perception of Frozen VLMs
    USarXiv cs.AI

    arXiv:2611.23962v1 Announce Type: cross Abstract: Vision-language models (VLMs) grasp the overall content of an image but often miss the fine details a question hinges on, such as a small object in a large photo or a number in a dense…

  135. Angular Conditional Distillation: Overcoming Classifier Geometry Mismatch
    USarXiv cs.AI

    arXiv:2611.23941v1 Announce Type: cross Abstract: Classifier outputs provide a fundamental and univer- sally available knowledge interface for knowledge distil- lation (KD). However, standard logit objectives entangle transferable semantic…

  136. Kernel Ridge Attention: Nonlinear Memory with Selective Forgetting
    USarXiv cs.AI

    arXiv:2611.23896v1 Announce Type: cross Abstract: Kernel Ridge Attention (KRA) is a recurrent sequence mixer that combines learned nonlinear features with covariance-guided memory updates. It stores associations through recursive ridge…

  137. NullTok: empirical-null token selection for training-free long video understanding
    USarXiv cs.AI

    arXiv:2611.23881v1 Announce Type: cross Abstract: Long video understanding remains a daunting challenge for existing multimodal large language models (MLLMs), mainly due to the large number of visual tokens accumulated in long videos…

  138. In Situ Training of Optical Neural Networks by Coding Agents
    USarXiv cs.AI

    arXiv:2611.23875v1 Announce Type: cross Abstract: Optical neural networks promise fast, energy-efficient computing, but every device departs from its design and must be trained with the hardware. In situ training relies on an expert-built…

  139. LAGR: Language-Aware Grasp Re-Ranking for Task Oriented Grasping
    USarXiv cs.AI

    arXiv:2611.23858v1 Announce Type: cross Abstract: Grasping is a fundamental capability for robotic manipulation, with the appropriate grasp depending on the downstream task. For example, a hammer should be held by its handle for hammering…

  140. RePlan-VLA: Progress-Aware Replanning for Long-Horizon Vision-Language-Action Models
    USarXiv cs.AI

    arXiv:2611.23843v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models have emerged as a promising paradigm for general-purpose robotic manipulation, yet their performance often degrades on long-horizon tasks that require…

  141. Learning Martial Arts Motion Skills for Animating Opponents in VR Sparring Games
    USarXiv cs.AI

    arXiv:2611.23825v1 Announce Type: cross Abstract: Animating virtual opponents is an important yet challenging task in game development. In martial arts games, opponent motions must not only be physically plausible but also exhibit…

  142. Beyond Replacement: Levenshtein Editing for Diffusion Language Models
    USarXiv cs.AI

    arXiv:2611.23750v1 Announce Type: cross Abstract: Diffusion language models (DLMs) offer a flexible alternative to autoregressive models, enabling any-order infilling. While most existing decoding methods treat decoded tokens as fixed…

  143. Toward Out-of-Distribution Generalization in Neural AC Power Flow
    USarXiv cs.AI

    arXiv:2611.23729v1 Announce Type: cross Abstract: Growing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reliability of traditional power flow solvers at scale…

  144. Learning through Discernment: when a small language model learns from user feedback on the job, and when it decays
    USarXiv cs.AI

    arXiv:2611.23716v1 Announce Type: cross Abstract: Small language models on devices could keep improving by learning, overnight, from their own daily work. Without a true outcome, however, something must decide which attempts are worth…

  145. Rational Ensemble Logic for Cohort Discovery with Real-World Electrophysiology Data
    USarXiv cs.AI

    arXiv:2611.23708v1 Announce Type: cross Abstract: Biomedical knowledge representation should make the passage from observations to conclusions transparent, rigorous, understandable, explainable, and reproducible (TRUER). We introduce…

  146. MEM-STRESS: Longitudinal Field Study of a Wearable-Informed AI Chatbot for Stress Management
    USarXiv cs.AI

    arXiv:2611.23691v1 Announce Type: cross Abstract: Stress is an increasingly prevalent concern, creating a need for accessible and personalized support. Stress-management chatbots primarily rely on what users report, which may not capture…

  147. Structured Extraction of Diffraction Data from Covalent Organic Frameworks
    USarXiv cs.AI

    arXiv:2611.23680v1 Announce Type: cross Abstract: Powder X-ray diffraction (PXRD) is the primary structural evidence for covalent organic frameworks (COFs) and many other materials. Yet decades of measured patterns exist only as images in…

  148. Lodestone: Automated Semantic Discovery of Datacenter Hardware Specialization
    USarXiv cs.AI

    arXiv:2611.23663v1 Announce Type: cross Abstract: Hardware specialization can improve datacenter efficiency. Unfortunately, identifying computations that justify dedicated hardware is difficult. Existing approaches analyze hot functions…

  149. Safe Personal RSI: Evaluating and Repairing Personal-Agent Updates
    USarXiv cs.AI

    arXiv:2611.23651v1 Announce Type: cross Abstract: Kicker is a personal health assistant. Changes to an assistant's reusable instructions can improve task handling while changing what it discloses, which requests it follows, and how it…

  150. Toaster: Nonlinear Memory from Cascaded Linear Memories and Target Propagation
    USarXiv cs.AI

    arXiv:2611.23648v1 Announce Type: cross Abstract: Recurrent memories let language models process long sequences with a fixed-size state. Linear memories such as DeltaNet and KDA store one matrix and take one delta-rule step per token. The…