AX·NEWS
Sign in

Artificial Intelligence

Artificial Intelligence news — Thursday 05 November 2026

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

Open in the live feedRSS feedOn a TV

  1. [Nov. 5] Ministry of Land, Infrastructure, Transport, Infrastructure, Transport, Transport, Transport and Tourism
    KRAI Times (Korea) Translated from Korean

    39th 1.3th (local time) is the result of a resident vote related to AI data center conducted with the middle of the United States. In 39 cases of the results were identified in the 4th aggregated 41 of the 13 states of the United States…

  2. OpenAI Launches ‘GPT-6 Cyber’... Improves 51% Attacking Performance
    KRAI Times (Korea) Translated from Korean

    OpenAI launches ‘GPT-6 Cyber’, a Cyber Security Specialization AI model that has 51% high attack success rate for security vulnerabilities than its existing models. In Linux operating systems, we have developed 948 attack codes that work…

  3. Cult.fit builds operational muscle with Codex
    USOpenAI Blog

    Cult.fit is using Codex to help teams across functions build tools, move faster, and change how work gets done.

  4. New NVIDIA Cosmos 3 Open Models for Interactive World Simulation
    USHugging Face Blog
  5. Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips
    GBGoogle DeepMind
  6. Australian universities are creating AI-native graduates
    USOpenAI Blog

    Australian universities are expanding access to ChatGPT and bringing AI into teaching and learning to prepare students for an AI-enabled workforce.

  7. Federated Authentication to CoreWeave Kubernetes with an External OIDC Provider
    USWeights & Biases Blog

    How to configure CoreWeave Kubernetes Service for external OIDC authentication, then use kubelogin so kubectl users can sign in through Google OAuth or another identity provider.

  8. How to Report for 100+ Clients With One Semantic Model
    USTowards Data Science

    Configuration-driven reporting: one semantic model, one metric catalog, and a few lines of configuration per audience The post How to Report for 100+ Clients With One Semantic Model appeared first on Towards Data Science .

  9. How We Built NASA's Zero-Shot Detector for Gas Leaks from Space
    USTowards Data Science

    We trained one model on methane. Without retraining, it also maps carbon monoxide, ammonia and nitrogen dioxide plumes, including the first automated CO detections from NASA’s EMIT. The post How We Built NASA's Zero-Shot Detector for Gas…

  10. Hamed Karimi wins ECML PKDD Test-of-Time Award
    USAmazon Science Blog

    In Naples last month, at the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD), Amazon senior applied scientist Hamed Karimi and colleagues won the 2026 ECML PKDD…

  11. How GEO, WebMCP, and Bot Paywalls Are Building the AI-Native Internet
    USTowards Data Science

    The internet of the future will look much different than the one today. Businesses are already transitioning. The post How GEO, WebMCP, and Bot Paywalls Are Building the AI-Native Internet appeared first on Towards Data Science .

  12. The Download: Texas’s fragile AI boom and energy policy shifts
    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. Texas was the new epicenter of the AI data center buildout. That might be kaput. Texas is the…

  13. Why a Tiny GPT Can Know Two Facts and Still Not Connect Them
    USTowards Data Science

    A tiny GPT memorized every fact I gave it, then couldn’t answer a single question that needed two of them. I tested the one-line fix from a recent paper, and it only half worked. The post Why a Tiny GPT Can Know Two Facts and Still Not…

  14. Antarpic, AI model ‘Check identity’ technology release.. . Clarification Samsung Hospital, Park, Soo - Jin Bae Yong-ICU visitation privileges
    KRAI Times (Korea) Translated from Korean

    Antarpic has released an open source technology that mathally validates the AI model that promised to receive the user. We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will…

  15. A new bill could change how the US builds energy infrastructure 
    USMIT Technology Review Biotech

    A sweeping bit of legislation designed to cut red tape could speed up the buildout of energy infrastructure in the US. With the dust barely settled on this week’s momentous midterms, the Senate is set to vote on the Bipartisan American…

  16. Doorshot AI, ‘Kimi K3.1 Pro’ Weighting Public...“Deep Thyso-Attension Combo”
    KRAI Times (Korea) Translated from Korean

    Moonshot AI revealed the weight of the ‘Kimi K3.1 Pro’, so developers could run and modify their model in their own infrastructure. The model card has combined deep-shik blissation with internal structures and low-density operation…

  17. Developing facility safety management platform directly using public resources, AI
    KRAI Times (Korea) Translated from Korean

    Jeonnam Optical Co., Ltd. is a facility safety management platform that uses AI as a development tool. The period of writing in the design of safety inspection and diagnosis was shortened within one day of the existing 30 days. Jeonnam…

  18. Save 26 trillion won for 18 months with Spain and Solar
    KRAI Times (Korea) Translated from Korean

    In Spain, the solar power and self-storage equipment has been analyzed to lower the purchase cost of the power wholesale market for 18 months. If there was no photovoltaic, the average wholesale power price was 86% higher than actual…

  19. Gangjin clinic clinic 1 person per year clinic fee 427won... 1.7 times
    KRAI Times (Korea) Translated from Korean

    The annual clinic cost per person is the highest in Gangjin County and the lowest in Gwangyang city. The treatment cost of the two areas was about 1.7 times, and it was a big difference in the use of the medical institution. The 5th…

  20. [Bulletin] LG, Safety and Health Conference, 15 cases using AI
    KRAI Times (Korea) Translated from Korean

    ■ LG announced that the Seoul Martial Arts LG Science Park held a group safety reporting environment conference ‘LG Cheek(SHEC)’. More than 350 employees of the affiliate safety environment shared 21 innovation cases, and 15 of them were…

  21. BLOCKCHAIN “The first half of the coding market classic...”
    KRAI Times (Korea) Translated from Korean

    OpenAI was classic in the coding market of the first half of the year, but it was successful in the company’s market in the last three minutes. Currently, corporate customers account for more than half of the total sales, and the entry…

  22. NVIDIA is supplying only 1% of the order of "RTX" chips in accordance with the Chinese client
    KRAI Times (Korea) Translated from Korean

    Chinese major technology companies such as Alibaba, Tencent, and Bybyds, are expected to be supplyed by an export regulator of $700 billion (about $94 trillion) in Nvidia, but it is expected to be supplying about 1% of total water for 4…

  23. GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents
    USOpenAI Blog

    Introducing GPT‑6 Cyber, OpenAI’s cybersecurity model for finding exploitable vulnerabilities and developing fixes, trained to stay within authorized scope.

  24. OpenAI acquires code storage startup Pierre.. .Country of codecs
    KRAI Times (Korea) Translated from Korean

    OpenAI acquires Pierre Computer Company with code storage and management technology. AI coding agent ‘code’ is a strategy that integrates code storage and change history management capabilities to handle the entire process of creating…

  25. Canba, growing in 10 years in AI.. . Venues
    KRAI Times (Korea) Translated from Korean

    Australia’s representative design platform, Canva announced a massive construction aim for 1271 people, equivalent to 22% of the entire staff. For the first time in 10 years, Dada AI introduced the cost and the cost of the competition, and…

  26. SKT is a global leader in the world’s best-in-class products.
    KRAI Times (Korea) Translated from Korean

    SK Telecom’s 21st Busan Fireworks Festival will be held on the 7th floor of SK Telecom and will strengthen the quality management by adding ‘Quality Satisfaction Rate’ indicator to the AI-based communication network operation system. SKT…

  27. Google expands hope for 5 countries in AI investment... Deep Mind
    KRAI Times (Korea) Translated from Korean

    We are expanding our hope to reduce costs by expanding AI investment. While overall workforce is growing, it is the idea of reconciliation of manpower and cost based on the priority of business. According to the 4th (local time) Rotor…

  28. China’s largest NAND memory YMTC, IPO pending with proficiency rapids and regulatory
    KRAI Times (Korea) Translated from Korean

    China’s largest NANDFlash memory company, YMTC, held the company’s opening (IPO) of $49 billion (approx. 6 trillion KRW). The Duda Financial Group is full of influences that the Chinese prosecutor has strengthened the new listing screening…

  29. 'Change on the Korean Peninsula' and '2026 Deep Tech Startup Challenge'
    KRAI Times (Korea) Translated from Korean

    The top prize of DeepTech Startup Challenge was held in June last year for early excavation and growth support of the government’s deeptech startups. The GS Tower was selected as the 6 public IR screenings and awards selected among the 182…

  30. Meta acquires AI coding startup contactor. .Establishment of the disease agent Tuning technology
    KRAI Times (Korea) Translated from Korean

    Meta acquired a startup contractor that supports running and managing multiple AI coding agents at the same time. Beyond AI model competition, developers will be pooled with strategies to secure the coding environment that can efficiently…

  31. Solar Panel Overheating cooling technology...Secure up to 32°C
    KRAI Times (Korea) Translated from Korean

    In the experiment that reproduces the heating of the solar module, combining the reputable heat pipe with water cooling system, the surface top temperature is reduced to 32°C. A study that reduces the “Hot Spot”, the state-of-the-art…

  32. Apple, employee AI use budget up to 3/frequency.. . Skip to content
    KRAI Times (Korea) Translated from Korean

    As Apple encourages AI usage, it promotes the implementation of the use of AI models for employees to reduce their budget by up to 2/3. It is a movement that will maximize AI adoption in the in-house business and software development, and…

  33. Japan, U.S.A. AI Enterprise, Inc. vulnerability information sharing request
    KRAI Times (Korea) Translated from Korean

    Japan government plans to request software vulnerability information that is not yet disclosed to the US AI company to respond to cyber attacks. In the future, we ask for the expansion of the access rights of high-performance AI models to…

  34. Quality, SpaceX and Bandage Up to 5G contract...Start link building
    KRAI Times (Korea) Translated from Korean

    Qualcomm has signed a contract for supplying 5G infrastructure for years with SpaceX. SpaceX intends to enter the U.S. mobile communication market based on the satellite communication service star link, which utilizes the semiconductor…

  35. Merge builds product and business tools with OpenAI technology
    USOpenAI Blog

    Merge uses GPT-6 Astra, ChatGPT, and Codex to review product code and support marketing. Its teams also use AI to plan executive dinners and prepare sales updates.

  36. Altman changed its name to ‘Lam Altman’ in the RAM field
    KRAI Times (Korea) Translated from Korean

    Sam Altman OpenAI CEO was a succession to change his X account name to ‘ram Altman’ while he was criticized as an expression about the use of personal PC’s memory. Altman, the CEO of Ilron Musk, has been deemed to use as much RAM as people…

  37. If you have more than 4 governments, you will be able to earn more than 3 billion SMB
    KRAI Times (Korea) Translated from Korean

    The 12 countries of Jeonnam-Gwangju are applied to all the government officials of the agricultural village in the next year, and the more locals selected for the government’s business, the increased financial burden of the special city…

  38. "Be beyond the bar code "Vision AI" as an unmanned asset"...Finderseiai, bakery store
    KRAI Times (Korea) Translated from Korean

    Saturday 11am. In front of the large bakery calculated table, the guest will be placed on the top of the Calculation table and the Calculation table. After a while, the above camera will separate the bread and quantity within 1 to 2…

  39. This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies
    KRAI Times (Korea) Translated from Korean

    This year, about 95% of the reorganization area of autumn leaves were shown to be equivalent to an increase in the special city of Gyeongju. The demand for Kimjang cabbage is expected to decrease 9.1% more than last year, and the price of…

  40. The mining sector prepared for the semiconductor hills...Japan rises the land value, the United States go for traffic and environment
    KRAI Times (Korea) Translated from Korean

    The land of Gwangju County Airport was designated as a candidate for the National Industrial Complex of Honam National Semiconductor, and the mining industry began to discuss the issues of residents and traffic, share, education and…

  41. KT·Seoul University, AI fusion security talent training... Establishment of convergence security contract department
    KRAI Times (Korea) Translated from Korean

    KT (CEO Park Yoon-young) announced that it has signed a contract for the establishment of ‘Convergence Security Contracts’ for the procurement of fusion security experts in Seoul National University and AI era. This agreement was promoted…

  42. China GLM-5.3, use 7.4 M DASH for each answer...Call load is almost not
    KRAI Times (Korea) Translated from Korean

    When creating an answer every AI model, it appears that there is a distinct difference in the sentence. China Z.ai's 'GLM-5.3' has been investigated to use an average 7.4 long dash symbol for a response, but the latest closure model of…

  43. Kakao, 3Q sales and sales profit increase... Venues
    KRAI Times (Korea) Translated from Korean

    Kakao (CEO Myeong-ah) announced that it recorded KRW 2,867 trillion in sales and KRW 2403 billion in sales as of Q3 this year. 6% sales profit increased by 7% The sales of the platform sector was aggregated by KRW 1 trillion in 17% growth…

  44. AMD, 4Q Sales 23Q... In the first time after the creator, Intel Chughan
    KRAI Times (Korea) Translated from Korean

    AMD has increased AI data center demand and recorded the largest quarterly sales in the world. The fourth quarter sales was expected to be approximately $170 billion (about 23 trillion KRW), and the possibility of the competitors’ Intel…

  45. PointRAFT: 3D deep learning for high-throughput prediction of potato tuber weight from partial point clouds
    USarXiv cs.AI

    arXiv:2512.24193v2 Announce Type: replace-cross Abstract: Key phenotypic traits, such as crop size, volume, and weight, can be estimated non-destructively from point clouds. In practice, however, point clouds of crops are typically…

  46. OpInf-LLM: Parametric PDE Solving with LLMs via Operator Inference
    USarXiv cs.AI

    arXiv:2602.01493v3 Announce Type: replace-cross Abstract: Solving diverse partial differential equations (PDEs) is fundamental in science and engineering. Large language models (LLMs) have demonstrated strong capabilities in code…

  47. Phantoms and Disclosures: A Statistical Framework for Auditing Privacy in Synthetic Data
    USarXiv cs.AI

    arXiv:2606.16952v3 Announce Type: replace-cross Abstract: The rapid adoption of generative AI and Large Language Models (LLMs) has spurred interest in synthetic data as a privacy-preserving alternative to sensitive real-world datasets…

  48. A Lightweight Multi-View Approach to Short-Term Load Forecasting
    USarXiv cs.AI

    arXiv:2602.09220v2 Announce Type: replace-cross Abstract: Time series forecasting is a critical task across domains such as energy, finance, and meteorology, where accurate predictions enable informed decision-making. While…

  49. Same Words, Different Judgments: How Preferences Vary Across Modalities
    USarXiv cs.AI

    arXiv:2602.22710v4 Announce Type: replace-cross Abstract: Preference-based reinforcement learning (PbRL) is the dominant framework for aligning AI systems to human preferences. However, evaluation protocols for such data were designed for…

  50. Beyond the Prompt: Jailbreaking Function-Calling LLMs via Simulated Moderation Traces
    USarXiv cs.AI

    arXiv:2607.00481v2 Announce Type: replace-cross Abstract: Jailbreak attacks remain a critical threat to the safe deployment of large language models (LLMs). While prior work has primarily studied attacks and defenses at the prompt level…

  51. Aligning Training and Rollout Policies for End-to-End FP4 RL on Mixture-of-Experts LLMs
    USarXiv cs.AI

    arXiv:2611.04376v2 Announce Type: replace-cross Abstract: In this work, we study end-to-end FP4 reinforcement learning (RL) post-training for Mixture-of-Experts (MoE) large language models. Recent work on FP4 RL has introduced effective…

  52. Dexterous Robotic Hands in the Foundation-Model Era: A Full-Stack Survey of Hardware, Data, Cross-Embodiment Transfer, Learning, and Evaluation
    USarXiv cs.AI

    arXiv:2611.06724v1 Announce Type: cross Abstract: Dexterous robotic hands are increasingly coupled to large-scale human data and foundation models, yet progress in hardware, representation learning and evaluation is often assessed…

  53. Smartphone-Aided VR Exergames for HIIT: Gamifying Fitness with Real-Time Pose Estimation
    USarXiv cs.AI

    arXiv:2611.06734v1 Announce Type: cross Abstract: This paper introduces an innovative virtual reality (VR) exergame system designed to gamify High-Intensity Interval Training (HIIT) workouts through 11 predefined combo moves, including…

  54. PREBA: Surgical Duration Prediction via PCA-Weighted Retrieval-Augmented LLMs and Bayesian Averaging Aggregation
    USarXiv cs.AI

    arXiv:2603.13275v4 Announce Type: replace-cross Abstract: Accurate prediction of surgical duration is pivotal for hospital resource management. Although recent supervised learning approaches, from machine learning (ML) to fine-tuned large…

  55. Reasoning over Non-Parametric Memory in Multilingual LLMs: A Cognitive-Based Analysis of Relational Knowledge
    USarXiv cs.AI

    arXiv:2611.06715v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) assumes that language models can reason over relations stated in their context rather than fall back on associations memorized during pretraining. We…

  56. Don't Replace, Complement: A Detector-Agnostic Subspace Signal for Near-OOD Detection
    USarXiv cs.AI

    arXiv:2611.06688v1 Announce Type: cross Abstract: Post-hoc out-of-distribution (OOD) detection is mature on far-OOD yet remains weak on near-OOD. These two settings differ in their use of class-characteristic feature directions: far-OOD…

  57. Human Motion Prediction via Continual Prior Compensation
    USarXiv cs.AI

    arXiv:2611.06637v1 Announce Type: cross Abstract: Human Motion Prediction (HMP) aims to predict future human poses at different moments according to observed past motion sequences. Previous approaches mainly treated the prediction of…

  58. Trick-or-Treat: Adversarial Evaluation of Multimodal Medical Reasoning with Synthetic Images
    USarXiv cs.AI

    arXiv:2611.06631v1 Announce Type: cross Abstract: We used medical image synthesis to create a challenging multimodal evaluation framework that exposes hallucination behaviors in frontier models. Evaluating multimodal reasoning has…

  59. From Reconstruction to Imputation: Test-Time Optimization for Missing Time-Series Channels
    USarXiv cs.AI

    arXiv:2611.06612v1 Announce Type: cross Abstract: Complete sensor-channel failures create a difficult form of time-series missingness: local interpolation is unavailable, and deploying a separately trained imputer for every failure pattern…

  60. Signed, Sealed, Leaked: Trust Beyond Signature for MCP Toolchains
    USarXiv cs.AI

    arXiv:2611.06585v1 Announce Type: cross Abstract: AI agents rely on third-party tools, such as Model Context Protocol (MCP) servers, whose publishers can be authenticated and whose execution can be restricted by runtime security policies…

  61. IDSISA: Influence-Distilled Sharded Training for Efficient Machine Unlearning
    USarXiv cs.AI

    arXiv:2611.06519v1 Announce Type: cross Abstract: Machine unlearning removes the influence of specific training examples to comply with privacy regulations. Full retraining gives strong guarantees but is computationally prohibitive at…

  62. NaijaFacts: A Framework for Disentangling Language from Knowledge in Facts about Nigeria
    USarXiv cs.AI

    arXiv:2611.06504v1 Announce Type: cross Abstract: Multilingual benchmarks for African languages mostly translate generic questions into local languages, measuring linguistic competence but not whether models know facts about Africa. We…

  63. TailSteer: Trajectory-Tail Steering for Drift-Resistant Multi-turn Image Editing
    USarXiv cs.AI

    arXiv:2611.06741v1 Announce Type: cross Abstract: Strong single-turn editing performance does not guarantee reliable multi-turn interaction: as edited outputs are repeatedly reused, small unintended changes can accumulate into severe…

  64. EvoWorld: Towards Out-of-Sight World Evolution in Generative Video Models
    USarXiv cs.AI

    arXiv:2611.06748v1 Announce Type: cross Abstract: Generative video models are increasingly used as visual world simulators, yet their primary strength lies in rendering plausible pixels rather than reasoning about how a world should…

  65. ReMoA: Recursive Modular Adaptation via Sparse Latent Refinement in Multimodal Diffusion Models
    USarXiv cs.AI

    arXiv:2604.25299v2 Announce Type: replace-cross Abstract: Diffusion models have achieved success in high-fidelity data synthesis, yet their capacity for more complex, structured reasoning such as text following remains constrained. While…

  66. PixCon: Clean-Positive Contrastive Learning for Foundation-Model Semi-Supervised Segmentation
    USarXiv cs.AI

    arXiv:2607.03068v2 Announce Type: replace-cross Abstract: Can a simple labeled-pixel contrastive branch improve semi-supervised segmentation with a strong pretrained encoder? We study this question with PixCon, which adds a class-wise…

  67. Message-passing graph persistence
    USarXiv cs.AI

    arXiv:2611.06801v1 Announce Type: cross Abstract: Message-passing graph neural networks (GNNs) are limited in expressive power by the one-dimensional Weisfeiler-Leman test. Persistent homology (PH) provides topological descriptors that can…

  68. Enabling AWS Cloud FPGA Development Using High-Level Vitis HLS
    USarXiv cs.AI

    arXiv:2611.06829v1 Announce Type: cross Abstract: We present a toolchain for development of ML models on AWS EC2 F2 cloud FPGA infrastructure from high-level language, implementing a complete Vitis HLS to HDK bridge to AFI workflow that…

  69. Ordinal Credit Assignment: Multi-Step Reinforcement Learning from Comparisons Alone
    USarXiv cs.AI

    arXiv:2605.06375v2 Announce Type: replace-cross Abstract: Policy-gradient methods need a scalar reward, yet for many tasks no formula captures what we want, while judging which of two outcomes is better is easy. We study reinforcement…

  70. CounterFactual Retrieval for Evidence-Aware Individual Animal Identification
    USarXiv cs.AI

    arXiv:2611.06864v1 Announce Type: cross Abstract: Automated identification of individual animals is moving from small curated studies toward shared benchmarks and reusable foundation models. This progress is useful, but current evaluation…

  71. Computational Phenomenology of Borderline Personality Disorder: A Comparative Evaluation of LLM-Simulated Expert Personas and Human Clinical Experts
    USarXiv cs.AI

    arXiv:2508.19008v4 Announce Type: replace Abstract: Building on a human-led thematic analysis of clinical life-story interviews (> 150,000 words) with inpatients with Borderline Personality Disorder, this study examines the capacity of…

  72. Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves
    USarXiv cs.AI

    arXiv:2605.06395v3 Announce Type: replace-cross Abstract: Modern deep learning architectures increasingly contend with sophisticated signals that are natively infinite-dimensional, such as time series, probability distributions, or…

  73. Token Time Continuous Diffusion for Language Modeling
    USarXiv cs.AI

    arXiv:2607.14106v2 Announce Type: replace-cross Abstract: In this paper we introduce token time continuous diffusion (TTCD), a new diffusion language model which (a) operates in continuous space, deterministically mapping Gaussian noise to…

  74. Supersampling Stable Diffusion and Beyond: Training-Free Neural Network Scaling via Signal Interpolation
    USarXiv cs.LG

    arXiv:2605.08698v3 Announce Type: replace-cross Abstract: This work presents a training-free approach for interpolating neural networks using signal interpolation methods (i.e., linear, bilinear, bicubic). We discovered this phenomenon…

  75. Ethical Fairness in Ubiquitous Health Sensing without Known Attributes
    USarXiv cs.LG

    arXiv:2603.13373v5 Announce Type: replace-cross Abstract: Wearable and mobile sensing models infer health and well-being in everyday life, yet strong average performance can conceal failures for particular people, contexts, or devices…

  76. Evolutionary System Prompt Learning for Reinforcement Learning in LLMs
    USarXiv cs.AI

    arXiv:2602.14697v4 Announce Type: replace Abstract: Building agentic systems that can autonomously self-improve from experience is a longstanding goal of AI. Large language models (LLMs) today primarily self-improve via two mechanisms…

  77. AdaMamba: Adaptive Frequency-Gated Mamba for Long-Term Time Series Forecasting
    USarXiv cs.AI

    arXiv:2604.23239v2 Announce Type: replace Abstract: Time-series forecasting is foundational to a wide spectrum of web-scale services, spanning personalized recommendations, energy scheduling, and infrastructure capacity planning. Accurate…

  78. Structural Rationale Distillation via Reasoning Space Compression
    USarXiv cs.AI

    arXiv:2605.07139v2 Announce Type: replace-cross Abstract: When distilling reasoning from large language models (LLMs) into smaller ones, teacher rationales for similar problems often vary wildly in structure and strategy. Like a chef who…

  79. OTAP: Structure-Aware Optimal Transport for Evaluating Planning and Execution in Agent Trajectories
    USarXiv cs.AI

    arXiv:2607.17082v3 Announce Type: replace Abstract: Large language model agents solve tasks by generating trajectories that interleave planning, tool calls, and intermediate results. Current evaluation metrics reduce such a trajectory to a…

  80. CivBench: A Long-Horizon Benchmark for Tool-Mediated Agents in Civilization VI
    USarXiv cs.AI

    arXiv:2609.02459v2 Announce Type: replace Abstract: We present CivBench, an open-source benchmark for evaluating language model agents in long-horizon, tool-mediated environments through the Model Context Protocol (MCP). A single episode…

  81. Lattice Deduction Transformers
    USarXiv cs.AI

    arXiv:2605.08605v2 Announce Type: replace-cross Abstract: We introduce the Lattice Deduction Transformer (LDT), a recurrent transformer that approximates logically sound deduction by projecting its latent state through a lattice between…

  82. Revisiting data-driven dynamic security assessment with a tabular foundation model
    USarXiv cs.AI

    arXiv:2607.16031v2 Announce Type: replace-cross Abstract: Data-driven pre-fault dynamic security assessment (DSA) rapidly evaluates the dynamic risk of credible contingencies on a power system using machine learning (ML). Existing…

  83. When to Ask, When to Answer: Strategic Clarification for Ambiguous Reference Resolution in Dialogue
    USarXiv cs.AI

    arXiv:2611.05517v1 Announce Type: cross Abstract: For humans, referential disambiguation is a collaborative process involving clarification questions asked when they will meaningfully change the outcome. However, large language models…

  84. How Finely Can Vision-Language Models See Biology? Bio-VQA: A Benchmark for Taxonomic, Morphological, Spatial and Individual Reasoning
    USarXiv cs.AI

    arXiv:2611.05559v1 Announce Type: cross Abstract: In biodiversity monitoring, a single image raises several questions: which taxon, how many, which trait, and whether the individual has been seen before. Vision-language models (VLMs) could…

  85. Isharah-Selfie: Continuous Sign Language Recognition Dataset for One-handed Signing
    USarXiv cs.AI

    arXiv:2610.33200v2 Announce Type: replace Abstract: Current sign language recognition and translation benchmarks are commonly collected in controlled settings using fixed or studio-style cameras, particularly limiting their ability to…

  86. Discarded or Unreachable? Diagnosing and Repairing Spatial Relation Failure in Vision--Language Similarity
    USarXiv cs.AI

    arXiv:2611.05565v1 Announce Type: cross Abstract: Contrastive vision-language models score left/right relations near chance, but the symptom has two causes that call for opposite repairs: the relation may survive in the pooled embedding…

  87. Jev with Foresight: Safe Language-Guided Autonomous Driving via a Language World Model
    USarXiv cs.AI

    arXiv:2611.05448v1 Announce Type: cross Abstract: Large language models understand passengers' driving commands, but as driving agents they reason in generated text for seconds and must infer each action's consequences implicitly from the…

  88. SurgTransfer: Multi-Step Latent Action Modeling for Action Transfer in Surgical World Model
    USarXiv cs.AI

    arXiv:2611.05443v1 Announce Type: cross Abstract: Surgical world models have the potential to provide intra-operative guidance, accelerate risk-free training and facilitate transfer of expert surgical maneuvers. However, the use of world…

  89. Constraint-Preserving Flow Matching for Spectral Graph Generation
    USarXiv cs.AI

    arXiv:2611.05436v1 Announce Type: cross Abstract: Spectral graph generation models graphs through the eigenvalues and eigenvectors of their Laplacians, providing a compact and continuous alternative to direct graph construction. However…

  90. LLM-based NPCs in Video Games: A Systematic Mapping Study
    USarXiv cs.AI

    arXiv:2611.05427v1 Announce Type: cross Abstract: For decades, non-player characters (NPCs) have been driven by rule-based systems whose personalities and interactions are scripted before release. Large language models (LLMs) promise NPCs…

  91. CAST-Bench: Tracing Every Sound to Its Source in Multi-Shot Audio-Video Generation
    USarXiv cs.AI

    arXiv:2611.05422v1 Announce Type: cross Abstract: Existing benchmarks for joint text-to-audio-video (T2AV) generation evaluate perceptual quality and audiovisual alignment, but balanced sound-source sam?pling and separate reporting by shot…

  92. AI-Driven SaaS Platform Integration: AI Express for Unified Generative AI Services with Pay-As-You-Go Credit Systems and Scalable Microservice Architecture
    USarXiv cs.AI

    arXiv:2611.05389v1 Announce Type: cross Abstract: Generative AI is currently fragmented, with separate subscriptions required for text generation, image synthesis, and code assistance, leading to high costs and workflow disruptions. AI…

  93. Behavior Graphs: Unsupervised Graph Structures for Interpreting Robot Deployment Data
    USarXiv cs.AI

    arXiv:2611.00405v1 Announce Type: cross Abstract: Data flywheels promise continuous improvement via insights from deployment data. Understanding how the deployed model succeeds and fails can steer subsequent data collection towards…

  94. Investigating Anthropometric Fidelity in SAM 3D Body
    USarXiv cs.AI

    arXiv:2601.06035v2 Announce Type: cross Abstract: The release of SAM 3D Body is a recent development in human mesh recovery, demonstrating improved performance in producing clean, topologically coherent meshes from single images. By…

  95. FrontierPhysics: Evaluating Agents for End-to-End Physics Research
    USarXiv cs.AI

    arXiv:2611.06910v1 Announce Type: new Abstract: AI agents are playing an increasingly important role in physics research. However, existing physics benchmarks only evaluate fixed questions, isolated sub-tasks, or paper reproduction with…

  96. Dull, Dirty, Dangerous? Working Conditions and the Limits of Occupational AI Exposure
    USarXiv cs.AI

    arXiv:2611.06888v1 Announce Type: new Abstract: Automation is often justified as relief from dull, dirty, and dangerous work. I examine how these conditions relate to occupational AI exposure using 21 O*NET items across 754 US occupations…

  97. On Turn-Level Best-of-N Sampling for Scaling LLM Agents
    USarXiv cs.AI

    arXiv:2611.06868v1 Announce Type: new Abstract: In this work, we evaluate a simple inference-time scaling approach based on \textit{turn-level best-of-N sampling} (\method) for large language model (LLM) agents on complex, long-horizon…

  98. ServerBench: How can your coding agents design specialized LLM serving systems?
    USarXiv cs.AI

    arXiv:2611.06841v1 Announce Type: new Abstract: Large language model (LLM) coding agents develop software through experimentation, offering an opportunity to design the serving systems supporting their own operation. Effective serving…

  99. Rules, Traces and Routing Statistics: Training and Calibration Data for Open Models on ARC-AGI-2
    USarXiv cs.AI

    arXiv:2611.06827v1 Announce Type: new Abstract: We study which training and calibration data help three open models solve ARC-AGI-2 tasks when the solver runs offline on four L4 GPUs for at most 12 hours. We compare data offline on public…

  100. ConceptARC at Scale: Studying Conceptual Understanding in Reasoning Models
    USarXiv cs.AI

    arXiv:2611.06732v1 Announce Type: new Abstract: Large reasoning models (LRMs) perform well on challenging benchmarks, yet it remains unclear whether they form generalizable abstractions across conceptually related reasoning tasks. The…

  101. Where Search Agents Go Off the Rails: Graph-Grounded Step-Level Verification
    USarXiv cs.AI

    arXiv:2611.06662v1 Announce Type: new Abstract: Search-augmented reasoning systems can reach correct answers while making errors in intermediate steps. Reliable auditing must check whether evidence $D$ covers the question $Q$, reasoning…

  102. Using Experience Twice: Dynamic Process Rubrics for Privileged Token-Level Credit Assignment
    USarXiv cs.AI

    arXiv:2611.06543v1 Announce Type: new Abstract: Agent trajectories often contain both successful decisions and failed attempts, yet optimization with sparse trajectory-level rewards struggles to identify which critical decisions should be…

  103. ElecLoad: Adaptive Electricity Load Forecasting Guided by Multiple Events
    USarXiv cs.AI

    arXiv:2611.06508v1 Announce Type: new Abstract: Accurate electricity load forecasting provides critical decision support for energy procurement, generation scheduling, and supply-demand balancing, contributing to the economic efficiency…

  104. FPTR: Feasibility-Preserving Transactional Refinement for Deadline-Constrained Joint Beam and Resource Allocation
    USarXiv cs.AI

    arXiv:2611.06491v1 Announce Type: new Abstract: Short-interval wireless schedulers must return a complete, feasible beam-and-resource allocation before each execution deadline, and a late or invalid output gives no utility in the current…

  105. D3D: Decoupled Diffusion Drafting for Asynchronous Speculative Decoding
    USarXiv cs.AI

    arXiv:2611.06425v1 Announce Type: new Abstract: Target-conditioned diffusion drafters predict candidate tokens in parallel, yet each new block must wait for updated target-model hidden states. This forces drafting and verification to run…

  106. GIST: Gibbs-Targeted Hierarchical Data Selection for Offline RLVR
    USarXiv cs.AI

    arXiv:2611.06395v1 Announce Type: new Abstract: Offline reinforcement learning with verifiable rewards (RLVR) trains on a fixed pool of sampled responses and typically supervises every token of every response, although responses and tokens…

  107. LegalEvolver: Agentic Legal Judgment Prediction via Rubric-Guided Self-Evolution
    USarXiv cs.AI

    arXiv:2611.06381v1 Announce Type: new Abstract: As LLM-based agents develop, they have been increasingly applied in the legal domain. Legal Judgment Prediction~(LJP) requires models to conduct multi-step reasoning and external legal…

  108. Corralling AI Risks with Grassroots Contracts
    USarXiv cs.AI

    arXiv:2611.06374v1 Announce Type: new Abstract: AI's inflection point is a fountain of alarming risks. Digital platforms in particular are vulnerable to three major AI risks, as those in control of global platforms cede control to AI, of…

  109. EnzymeRSI: A dual-loop recursive self-improvement system for enzyme design
    USarXiv cs.AI

    arXiv:2611.06321v1 Announce Type: new Abstract: Improvement in enzyme engineering is usually pursued by optimising which candidates to assay within a fixed model and workflow; when the bottleneck arises from a mismatched method or a…

  110. STEK: Structural-Terrain-Enhanced Koopman-Inspired Cross-Channel Regularization for Retrospective Battery Multi-Target Estimation
    USarXiv cs.AI

    arXiv:2611.06299v1 Announce Type: new Abstract: Retrospective battery characterization across heterogeneous cells is challenging because cycling records vary substantially with chemistry, protocol, cell format, capacity, and operating…

  111. DyPSI: Dynamic Physics Sensing via Joint Field and Sensor-Trajectory Generation
    USarXiv cs.AI

    arXiv:2611.06280v1 Announce Type: new Abstract: Physics sensing in evolving environments requires reconstructing dense spatiotemporal fields from sparse observations while adapting where future observations should be collected. Existing…

  112. A Definition of Catastrophic Performance
    USarXiv cs.AI

    arXiv:2611.06219v1 Announce Type: new Abstract: A central challenge in AI is specifying objectives that align with human preferences. While existing theoretical results elucidate how misspecification impacts performance, they typically do…

  113. CARM: Conformal Adaptive Risk Masking for Reliable Scientific Super-Resolution
    USarXiv cs.AI

    arXiv:2611.06151v1 Announce Type: new Abstract: Deep learning-based super-resolution (SR) can reconstruct fine-scale Earth-observation fields from coarse observations, but reconstruction accuracy alone does not provide a calibrated rule…

  114. SafeStart: Mitigating the Overlooked Safety Cost of Active Expert Reduction in MoE LLM Serving
    USarXiv cs.AI

    arXiv:2611.06147v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) language models send each token through a few of many expert subnetworks. Serving methods make these models faster by running even fewer experts per token, and they…

  115. Beyond Risk Cues: A Mechanism-Level Benchmark and Verifiable Debate for Scam Video Detection on the Web
    USarXiv cs.AI

    arXiv:2611.06110v1 Announce Type: new Abstract: Online scam videos pose significant threats to individuals and society. Existing studies primarily focus on binary classification, offering limited insights into how scams work and the…

  116. Reglos: A Self-Evolving Personal Memory Agent with Recursive Self-Improvement
    USarXiv cs.AI

    arXiv:2611.06049v1 Announce Type: new Abstract: Personal agents draw on past conversations to tailor their responses to each user. Their memory systems store and update what each user says, and coding agents now also rewrite the harness…

  117. From Waste to Discovery: Recycling High-Scored Unexposed Candidates in Taobao Feed Recommendation
    USarXiv cs.AI

    arXiv:2611.05997v1 Announce Type: new Abstract: Helping users discover new interests is vital to the long-term health of industrial recommender systems, yet the usual remedy, boosting exposure of under-explored (discovery) content, often…

  118. Split-KV: Efficient Context Parallelism for Training Diffusion Drafters
    USarXiv cs.AI

    arXiv:2611.05926v1 Announce Type: new Abstract: Diffusion drafters accelerate speculative decoding by denoising multiple draft tokens in parallel. Their training uses anchor sampling to construct compact draft queries and performs…

  119. Routes of Reasoning: Reading and Reshaping Test-Time Compute in MoE Models
    USarXiv cs.AI

    arXiv:2611.05919v1 Announce Type: new Abstract: Reasoning models allocate test-time compute across distinct behaviors such as planning, deriving, and verifying. However, models frequently "overthink," continuing to verify and re-derive…

  120. CHARM+: Cross-Hardware Attention with Re-Merge Multistream Mechanism
    USarXiv cs.AI

    arXiv:2611.05906v1 Announce Type: new Abstract: As sequence lengths continue to grow, distributed attention based on sequence parallel (e.g., Ring-Attention) has become a mainstream solution for large language models (LLMs). Nevertheless…

  121. EchoKV: Self-Corrective KV Cache Eviction via Iterative Query Feedback
    USarXiv cs.AI

    arXiv:2611.05903v1 Announce Type: new Abstract: KV cache eviction is not merely a token-ranking problem: the workspace retained for generating eviction queries also determines the quality of those queries. Consequently, an early workspace…

  122. RECAP: Context-Aware Prediction Recycling with Request-Local and Persistent Memories for Multimodal Speculative Decoding
    USarXiv cs.AI

    arXiv:2611.05894v1 Announce Type: new Abstract: Recycling a target model's own predictions offers a training-free source of proposals for speculative decoding. In vision-language models, however, existing recycling indexes predictions by…

  123. Risk Control Certifies Rates, Not Magnitudes: An Audit of Selective Fetal Biometry
    USarXiv cs.AI

    arXiv:2611.05851v1 Announce Type: new Abstract: Prenatal ultrasound biometry must return a head circumference (HC) and refuse when the image cannot support one. Selective prediction trades coverage against error, but its finite-sample…

  124. Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers
    USarXiv cs.AI

    arXiv:2611.05798v1 Announce Type: new Abstract: Recognizing visual content from neural-network weights requires separating functional information from arbitrary parameter coordinates. We audit this separation with hidden-neuron…

  125. FlyHash for LLM Agent Memory: An Empirical Validation and a New Application
    USarXiv cs.AI

    arXiv:2611.05791v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly rely on external memory stores to retain information across conversations, typically via dense vector embeddings searched with approximate…

  126. Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex
    USarXiv cs.AI

    arXiv:2611.05753v1 Announce Type: new Abstract: Organizations accumulate evidence in conversations, documents, decisions and relationships, but they act on interpretations: judgements assembled from several observations, resting on…

  127. Spatio-Temporal Multi-Order Interaction for Video Recognition
    USarXiv cs.AI

    arXiv:2611.05633v1 Announce Type: cross Abstract: Many temporal approaches for video recognition rely on explicit cross-channel aggregation projections, which have quadratic complexity in the number of channels, introducing significant…

  128. AutoEnergyDesign: An LLM-Driven Agent for Automatic Energy System Design Optimization
    USarXiv cs.AI

    arXiv:2610.35160v2 Announce Type: replace Abstract: The design optimization of building energy systems represents a complex, high-dimensional search problem involving discrete technology selections, continuous sizing parameters, and…

  129. Every finite group admits a just finite presentation
    USarXiv cs.AI

    arXiv:2605.10402v2 Announce Type: replace-cross Abstract: A finite presentation < X | R > of a finite group is called `just finite' if removing any relation from R results in a presentation for an infinite group. It has been an open…

  130. Text-Only Domain Adaptation for Low-Resource Medical ASR: Shallow Fusion and Density Ratio Rescoring for Hausa
    USarXiv cs.AI

    arXiv:2611.05744v1 Announce Type: cross Abstract: Automatic speech recognition (ASR) systems often struggle with specialised medical vocabulary, especially in low-resource languages where domain-specific speech data are scarce. We…

  131. BobFlow: Towards Human-Agent Collaborative Data Science Using Dataflows
    USarXiv cs.AI

    arXiv:2611.05761v1 Announce Type: cross Abstract: Recent advances in large language models (LLMs) have established human-agent collaboration as a promising paradigm for data science. Effective collaboration requires a shared computational…

  132. Signals: Continuous Diagnostics of AI Agents
    USarXiv cs.AI

    arXiv:2611.04507v2 Announce Type: replace Abstract: AI agents can make incorrect decisions or waste resources even when their tools and infrastructure work normally. Their nondeterministic behavior makes these problems difficult to…

  133. DARE: Difficulty Analysis for Repair Enhancement
    USarXiv cs.AI

    arXiv:2611.05797v1 Announce Type: cross Abstract: LLM-based coding agents are primarily evaluated by their solve rates. However, this metric does not reveal how difficult the solved tasks are or account for differences intask difficulty…

  134. When File Ingestion Misleads Language Models: Measuring and Mitigating Divergences in Office Documents
    USarXiv cs.AI

    arXiv:2611.05806v1 Announce Type: cross Abstract: Office documents are now routinely uploaded to large language models (LLMs) for processing. However, an LLM does not always see what a human sees, because it typically reads the text that…

  135. Learning to Incentivize Perfect-Knowledge Agents with Unobserved Rewards
    USarXiv cs.AI

    arXiv:2304.07407v3 Announce Type: replace-cross Abstract: Motivated by a number of real-world applications from domains like healthcare and sustainable transportation, in this paper we study a scenario of repeated principal-agent games…

  136. GraphIP-Bench: How Hard Is It to Steal a Graph Neural Network, and Can We Stop It?
    USarXiv cs.AI

    arXiv:2605.12827v3 Announce Type: replace-cross Abstract: Graph neural networks (GNNs) deployed as cloud services can be *stolen* through *model-extraction attacks*, which train a surrogate from query responses to reproduce the target's…

  137. From Representations to Behaviors: Exploring the Person-Situation-Behavior Triad in LLMs
    USarXiv cs.AI

    arXiv:2607.26853v2 Announce Type: replace-cross Abstract: Human personality theories characterize traits not as isolated attributes captured by a single score, but as stable individual tendencies expressed through the interplay among…

  138. Edge AI and TinyML for Resource-Constrained Agriculture: Hardware Platforms, Model Optimization, and Deployment Strategies
    USarXiv cs.LG

    arXiv:2603.15085v2 Announce Type: replace-cross Abstract: Deploying artificial intelligence at the edge is increasingly important for systems that must operate under strict limits on memory, computation, energy, connectivity, and hardware…

  139. Learning When Human Decisions Matter: Executor-Relative Consultation Value for LLM Agents
    USarXiv cs.AI

    arXiv:2611.05814v1 Announce Type: cross Abstract: Large language models increasingly automate complex tasks, but human decision capacity does not scale with their capabilities. In many applications, humans remain essential for providing…

  140. Autonomous ETL Pipelines for Salesforce M&A Consolidation: A Reference Framework, Jitterbit Mapping, and Reproducible Simulation Study
    USarXiv cs.AI

    arXiv:2611.05918v1 Announce Type: cross Abstract: Merging companies that each run Salesforce must consolidate their customer data into a single organization while both source systems remain in use. This requires reconciling different…

  141. When the Agent Learns Too: Learning to Incentivize with Unobserved Rewards
    USarXiv cs.AI

    arXiv:2308.06717v2 Announce Type: replace-cross Abstract: In practice, incentive providers (i.e., principals) often cannot observe the reward realizations of incentivized agents, which is in contrast to many principal-agent models that…

  142. Ray-Tracing-Based Dynamic Channel Sounding With Autoregressive Wi-Fi CSI Prediction
    USarXiv cs.AI

    arXiv:2611.05970v1 Announce Type: cross Abstract: Channel sounding in Wi-Fi multiple-input multiple-output (MIMO) systems must balance channel state information (CSI) accuracy against CSI acquisition overhead. Device mobility causes…

  143. How Unbalanced Must a Latin Rectangle Be?
    USarXiv cs.AI

    arXiv:2611.06003v1 Announce Type: cross Abstract: We view a $k\times n$ Latin rectangle as $n$ points in the grid $[n]^k$. Each symbol gives one point, whose coordinates are the columns it occupies. The rectangle is balanced when these…

  144. ENACT-Heart -- ENsemble-based Assessment Using CNN and Transformer on Heart Sounds
    USarXiv cs.AI

    arXiv:2502.16914v2 Announce Type: replace-cross Abstract: ENACT-Heart is a two-branch ensemble for heart-sound classification. A Vision Transformer (ViT) processes spectrogram images, while a convolutional neural network (CNN) processes…

  145. Prediction-Intervention Games and Invariant Sets
    USarXiv cs.AI

    arXiv:2605.16828v2 Announce Type: replace-cross Abstract: We consider the following two-player game: using observational data, the leader chooses a prediction function for a response variable $Y$ from given covariates. The follower then…

  146. CVE-Conditioned Exploit Generation with LLMs: A Data-Centric Study of Benchmarking and Fine-Tuning
    USarXiv cs.LG

    arXiv:2606.15123v2 Announce Type: replace-cross Abstract: We study the task of CVE-conditioned exploit generation, where a model drafts proof-of-concept (PoC) exploits given software vulnerability context. We adopt a data-centric approach…

  147. Touch\'{e}World tech report : Towards Online Unified Camera--Hand Exploration in Generative Worlds
    USarXiv cs.AI

    arXiv:2611.06061v1 Announce Type: cross Abstract: Recent video world models allow users to explore generated environments through camera navigation. Continuous hand motion provides a fine-grained action channel for interacting with the…

  148. Sparse stochastic digital twins can outperform exact ones
    USarXiv cs.LG

    arXiv:2606.16219v2 Announce Type: replace-cross Abstract: Digital twins are often built under the assumption that greater model fidelity leads to better downstream performance. We show that this is not always the case: retaining weakly…

  149. Agent trajectories as programs: fingerprinting and programming coding-agent behavior
    USarXiv cs.LG

    arXiv:2606.16988v2 Announce Type: replace-cross Abstract: Benchmark scores tell you what an agent got right; they do not tell you how it got there. In this work, we introduce methods for comparing agents procedurally in different contexts…

  150. The Harness Is the Model: Agent Scaffolding Changes a Fixed LLM's Compliance with Banking Conduct Rules, Unevenly Across Models
    USarXiv cs.AI

    arXiv:2611.06116v1 Announce Type: cross Abstract: Banks validate language models, but they deploy systems: a model inside a harness that decides what it sees, which tools it can call and who checks its actions. We test whether that harness…