AX·NEWS
Sign in

Artificial Intelligence

Artificial Intelligence news — Wednesday 04 November 2026

150 stories from 7 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. [Nov. 4] "Grow beyond the code"...LM evolves into a video production tool
    KRAI Times (Korea) Translated from Korean

    Intro to the 3 days of the claim that the American AI model "Fable 5.5" made the 2D animation of the Japanese theater board level. Developers claimed that they were made only 100% code without using images or video creation AI, but they…

  2. "Musk, the largest paradigm of the mid-line"...More than $500 trillion in business in a shield
    KRAI Times (Korea) Translated from Korean

    ILON MSK CEO has become one of the largest defeaters in the US. In order to win the publicity party, we put about $500 billion (about $600 billion), but the democratic party will be encouraging and win the core business base in Texas, and…

  3. BigTech Corp is a global leader in democratization...
    KRAI Times (Korea) Translated from Korean

    In the middle of the U.S., the Democratic Party was overwhelmingly likely to be a braking in the construction of the AI data center, but it was rather urgent to be a big tech company such as Amazon and Microsoft (MS). As the new data…

  4. What AI's growing share of tech spending means for enterprises in 2027
    GBAI Business

    AI spending is reshaping enterprise IT budgets. How CIOs can manage rising costs, prioritize investments and measure returns beyond productivity gains.

  5. AWS goes inside the enterprise to find the real value of AI
    GBAI Business

    The tech giant works closely with its customers on putting its AI systems into action.

  6. Texas was the new epicenter of the AI data center buildout. That might be kaput.
    USMIT Technology Review Biotech

    Texas is the fastest-growing data center market in the US and ground zero for Big Tech’s largest AI buildouts. But the blue wave that swept the Lone Star State may jeopardize that position. Democrat James Talarico won the US Senate seat…

  7. Roundtables: Inside the Quest for Cognitive Immortality
    USMIT Technology Review Biotech

    Tuesday, November 17, 2026 In an effort to reach an elusive continuity of the self, some far-thinking technologists are working on ways to keep your brain going—even beyond death. Join science editor Mary Beth Griggs and senior editor for…

  8. Dharma Field Extractor: Any Document In, Structured Fields in JSON Out
    USHugging Face Blog
  9. What’s next for US energy policy after the midterm elections?
    USMIT Technology Review Biotech

    I’ve been locked in to my various screens since results from midterm elections started to roll in last night. And it didn’t take long to see the beginning of the blue wave: Democrats took control of both the US House of Representatives and…

  10. Gwangyang Gyeongsangjae Office Completed Hanhwa Airros Face Plant
    KRAI Times (Korea) Translated from Korean

    The Gwangyang Manufacture Economic Zone Administration (Cheongchung-gon) announced the completion ceremony of the Hanhwa Airlos Face Netcheon Plant in the 1st mountain on the 4th. On the day of completion, 200 people attended major…

  11. Baidu’s new world model highlights value, challenge of physical AI
    GBAI Business

    The vendor trained the model on millions of miles of data from its Apollo Go robotaxi fleet.

  12. A Computational Analysis of Women’s Suffrage in Historical Newspapers
    USTowards Data Science

    Using 600,546 digitized newspaper pages to measure how suffrage coverage changed from 1890 to 1920 The post A Computational Analysis of Women’s Suffrage in Historical Newspapers appeared first on Towards Data Science .

  13. OpenAI to acquire Pierre Computer Company, Inc.
    USOpenAI Blog

    OpenAI will acquire Pierre Computer Company, expanding Codex’s tools for building, understanding, and reviewing code.

  14. How to Combine Contradictory Filters in DAX
    USTowards Data Science

    Sometimes we need to calculate results based on contradictory filters. Without any action, these filters will lead to an empty result. Let's see how we can improve the data model to make it work. The post How to Combine Contradictory…

  15. The Download: blended meat, restored sight, and the agentic web
    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. These startups hope “blended” food will cut our appetite for red meat In the fall of 2025, Jeff…

  16. Securing the agentic enterprise with data lineage
    USMIT Technology Review Biotech

    A sales representative asks an AI agent for customer pricing, and the agent pulls data from multiple sources: restricted pricing systems, internal communications, and archived quotes. The output looks accurate to the agent, so a second…

  17. Let Decision Models Guide Your Voice Assistant
    USTowards Data Science

    GPT-Live does the talking. A decision model tracks what’s confirmed and picks what to ask next. The post Let Decision Models Guide Your Voice Assistant appeared first on Towards Data Science .

  18. Beyond Classical: A Quantum Approach to Time-Series Prediction
    USTowards Data Science

    Introducing CoQS-LSTM, a Coherent Quantum Long Short-Term Memory Architecture The post Beyond Classical: A Quantum Approach to Time-Series Prediction appeared first on Towards Data Science .

  19. Development of Chinese Brainco, Brain Stimulation 'Electronic Obesity Therapy'
    KRAI Times (Korea) Translated from Korean

    China Brain-computer interface (BCI) company BrainCo develops wearable devices that suppress appetite by stimulating the brain. It is noticed in a new way that can replace the existing GLP-1 family obesity therapy, with the incision of the…

  20. "The more you work on AI, the more supervisory ability to retire"...Andropic engineer warning
    KRAI Times (Korea) Translated from Korean

    The AI agent has a warning that it can be weak in human judgment and professionalism while increasing coding and work. In particular, we need human expertise to validate the results made by AI, but we’ve thought that “Paradox of…

  21. Antropicy "A AI Safety Countermeasures Recurring"...Intensive Review
    KRAI Times (Korea) Translated from Korean

    Andropic has alerted that research and development using AI is rapidly sucking, which can speed up security and security measures. In particular, we acknowledge that the development rate was greatly reduced, and we have raised the need for…

  22. OpenAI, AI Hardware Design Completed In A Week...
    KRAI Times (Korea) Translated from Korean

    OpenAI announced that it accelerates the development of hardware from mechanical, electrical engineering to system design and prototype production. Design drawings and simulation settings, as well as work required for manufacturing…

  23. "AI ultra-shorting speed up to 35 times higher"...Jerfra, new entry technology release
    KRAI Times (Korea) Translated from Korean

    The American AI startup Zyphra revealed a new rare action technology that can significantly increase the speed of the draft context. We suggest predictions that we can provide up to 35 times faster processing speed than existing general…

  24. "AI On-site Engineers Up to KRW 1 billion"...Insisting FDE
    KRAI Times (Korea) Translated from Korean

    As a result of the development of AI’s real-world work, the company has been in the competition of “FDE”, and the body value of them is inclined. The annual total number of top-level personnel has been increased to $14 million (about $19…

  25. "Nvidia Heart Gaze"... OpenAI also promotes export control of China chips
    KRAI Times (Korea) Translated from Korean

    With the overwhelming dominance of AI chip market, NVIDIA has been a press release that has a strong influence on client’s business decisions, as well as the US government’s export policy. OpenAI has been deteriorating the relationship…

  26. EU Supreme Court Legal Affairs "AI Article unauthorized summary in copyright infringement". .Gallery of the GGoogle Minion
    KRAI Times (Korea) Translated from Korean

    The law of the European Court of Justice (ECJ), the Supreme Court of the European Union, announced that Google’s AI chatbot ‘Jeminai’ may infringe copyrights during the process of searching the press releases and creating an answer. We use…

  27. Jeonnam-Gwangju, Kwaksa 1362Farm Fire Check...Electric Factor 48.8%
    KRAI Times (Korea) Translated from Korean

    AI-produced video conjugate special inspection for the prevention of fire extinguishing of the winter season. In the last five years, the electric factor was 48.8%. On December 10, Jeonnam-gwang has announced that the pig farming farm is…

  28. “Design a health”... A new vitality to the public health space by the Gyungang Boguns, stone art walls
    KRAI Times (Korea) Translated from Korean

    In addition, we have a wide range of products and services, and we have a wide range of products and services. This project focuses on providing a friendly and comfortable environment for residents who visit the health department and…

  29. The 61th anniversary of the Sustainable World Island Expo...It’s 250 million on the day before the fall, leaving the island tourism possibilities
    KRAI Times (Korea) Translated from Korean

    The 2026 Yeosu World Island Expo will finish the schedule for 61 days. After reinforcement of the second half of the opening and improving the viewing environment, the late visitor increased rapidly, but the cumulative performance until…

  30. "China AI threat exaggeration"... benchmark reliability debate in performance gaps
    KRAI Times (Korea) Translated from Korean

    In the new performance evaluation of ATIFICAL ANTI-ASIS (AA), the top-notch model in the U.S. and the gaps in the Chinese open model have been determined that China AI’s competitiveness is not oversized. In particular, after increasing the…

  31. "OpenAI, Ignore the internal security vulnerability warning for years"... Developing speed constant confidential leak risk
    KRAI Times (Korea) Translated from Korean

    OpenAI has been pointed out that it has not been actively responding to warning about internal security vulnerabilities from employees for years. In order to prioritize the competition and technology development competitions such as Google…

  32. Hyundai Steel, In-house Producing AI Integrated Platform ‘Steel-X’ Operation... Warrior AX Pedaling
    KRAI Times (Korea) Translated from Korean

    Hyundai Steel launches the next generation AI integration platform and accelerates warrior AX innovation. Hyundai Steel has recently announced that it was on the operation of the platform ‘Steel-X’, which provides integrated…

  33. ATIUM INVESTMENT COMPANY ORIGINAL STORY
    KRAI Times (Korea) Translated from Korean

    The company’s AI expertise has been listed in Japan. ATIUM INVESTMENT announced that it has been listed on the Gross Market in Tokyo Stock Exchange in Japan, one of its investment portfolios. I am starting from California in the United…

  34. [Bulletin] Noz-Ai, 20 Billion Free Seated 2nd Investment
    KRAI Times (Korea) Translated from Korean

    ■ KOSDAQ announced that KOSDAQ has attracted a second investment of KRW 20 billion from the future asset venture investment and Korea Investment Partners. The cumulative investment amount is KRW 60 billion and it is also selected as TIPS…

  35. SELZ launches the global expansion of the Silicon Valley VC conference
    KRAI Times (Korea) Translated from Korean

    Eselz (CEO, K.S.M., Jeong Jae-Hyun) announced that it won the final winner at the ‘2026 Startup World Cup Korea Finals’ hosted by the Silicon Valley Venture Capital (VC). As a result, Eselz was ranked as a global connection stage in San…

  36. Jeonnam National University was established in 36 years as a pilgrimage...Capital 60 people
    KRAI Times (Korea) Translated from Korean

    Jeonnam National University of Medicine was established in 36 years of the first time in Korea. After the consolidation negotiations of the ranch and the ranch, the ranch was in front of 1.30, and the government was finalized by the law…

  37. A thousand fewer paths to root
    USOpenAI Blog

    OpenAI built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel.

  38. Nvidia opens the "Oto-Ontology" to understand corporate data...AI Accuracy
    KRAI Times (Korea) Translated from Korean

    Nvidia automatically analyzes the meaning and relationship of the internal data of the company and reveals the open source technology to help AI agents utilize. AI is a technology that helps to understand the exact meaning and calculation…

  39. OpenAI opens the "life balance" of daily work AI evaluation.. .Private secretary performance 1 year new 49%→81%
    KRAI Times (Korea) Translated from Korean

    OpenAI has released a performance assessment system for developing chatGPT as a private AI assistant that handles real daily tasks in a simple chat tool. It is a provocation that measures how well it performs life tightening tasks such as…

  40. Google "You need to give up search to AI in school"... Educational AI Limitation and Controversy
    KRAI Times (Korea) Translated from Korean

    Google has basically blocked AI functions such as "Jeminai" and "Google Class Room" in the U.S. elementary, middle and high school (K-12) student accounts. However, if the school permits Google search, the AI function in the search should…

  41. Antropic, a private public announcement of IPO in Australia, Japan...Global Investor Relations
    KRAI Times (Korea) Translated from Korean

    Andropic will lead the sale of shares for private investors in Australia, following Japan in front of the U.S. Corporation (IPO). It is a movement to expand the global investor base in the way the stocks listed in the U.S. are sold…

  42. OpenAI, former US Army Minister.. .Enhancing national defense business in expanding the adoption of the American AI
    KRAI Times (Korea) Translated from Korean

    OpenAI is the former President of the United States Army Dan Drift Co., Ltd. and has been strengthening AI business in the field of national security and defense. We appreciate that the introduction of AI from the United States and the…

  43. LG AI Compliant Agent 'Ex-One Nexus', Awarded National Governor'
    KRAI Times (Korea) Translated from Korean

    LG AI Research Institute has recognized the technical merits of the ‘2026 International Copyright Technology Conference (ICOTEC)’, hosted by the Ministry of Culture, Sports and Tourism, as well as the National Intellectual Property Office…

  44. C&M contracts supply of open package robot and robot drive parts from the development team of ‘Hubbo’
    KRAI Times (Korea) Translated from Korean

    C&M (CEO Kim Hyung) announced that the robot has signed a supply contract for the core drive parts of the robot such as open-pass robot and actuator. Yangjae establishes the mid-term supply system from 2027 to 2030, and jointly responds to…

  45. Daybreak Red is now available in AWS GovCloud (US-East)
    USOpenAI Blog

    Daybreak Red brings GPT-5.6 Cyber to approved customers in AWS GovCloud (US-East) through Amazon Bedrock for authorized cybersecurity work.

  46. Home > Products > Air Conditioner > Air Conditioner
    KRAI Times (Korea) Translated from Korean
  47. SKT, 3Q communications·AI 2 axis growth...
    KRAI Times (Korea) Translated from Korean

    SK Telecom announced that it recorded KRW 41.1 billion in sales of KRW 4973 billion in Q3 of the year and sales profit of KRW 55.3 billion in sales profit and KRW 4118 billion in revenue. In addition, sales was KRW 3156 trillion, operating…

  48. AMD, Furelios and Promise Boosts... Establishment of the next spring heterogeneous AI demonstration infrastructure
    KRAI Times (Korea) Translated from Korean

    AMD builds a new AI cluster that combines local AI semiconductors to its CPU·GPU in spring Seoul. AMD announced that the ‘AMD ADVANCED AI World Tour Seoul 2026’ at the ‘AMD ADVANCED AI World Tour Seoul 2026’ held at the Westin Seoul Parnas…

  49. Real World, Kim Minjeong Presidency in Shin-im COO, Korea IMT Consulting
    KRAI Times (Korea) Translated from Korean

    ROBOTICS Foundation Model Startup Real World (CEO Ryujung Hee) announced that it will be elected the CEO of the Korean IBM Consulting before Kim Minjeong. Kim Min-jeong Shin-im COO was a member of the Korean ISM Consulting, McKinji &…

  50. Apple, iOS 27.3 Beta release.. .Advanced AI and Syringe App Control
    KRAI Times (Korea) Translated from Korean

    Apple introduced Siri AI that enhances the AI function and app control ability for Apple Watch to remember and summarize conversations. With this update, developers can integrate their apps directly with the syringe, and the utilization…

  51. LGU+, 3Q sales profit 84% increase annual motivation ratio...AIDC Growth Traction
    KRAI Times (Korea) Translated from Korean

    LG UPlus announced that it recorded KRW 3,9813 trillion and KRW 2976 trillion in sales profits in Q3 of the year. Sales decreased by 0.7% year-long motivation, but sales profit increased by 84.1%. This is an explanation of the reduction…

  52. LLM-Augmented Agent-Based Simulation of Cyber Social Agents
    USarXiv cs.AI

    arXiv:2610.32988v2 Announce Type: replace-cross Abstract: Social media bots are often portrayed as malicious automations to be detected and removed, but automation on social media platforms encompasses a wide spectrum of roles, behaviors…

  53. SkillDRE: Dual-Stage Red-Team Evolution of Agent Skills via Pre-Execution and Runtime Feedback
    USarXiv cs.AI

    arXiv:2609.32400v2 Announce Type: replace-cross Abstract: Agent skills package instructions, executable code, and task-specific resources into reusable artifacts that agents can improve using execution feedback. The same mechanism also…

  54. OntoKG-EQ: Evidence-boundary auditing for analytical knowledge graphs
    USarXiv cs.AI

    arXiv:2609.08869v2 Announce Type: replace-cross Abstract: An analytical knowledge-graph result can be supported in its ambient graph while its delivered explanation omits deciding observations. OntoKG-EQ connects competency-question…

  55. Matryoshka Hash Representations for Model-Aware Compact Semantic Retrieval
    USarXiv cs.AI

    arXiv:2609.07276v2 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) depends on dense retrieval: each document is stored as a learned vector, and queries are answered by nearest-neighbor search. Keeping one…

  56. Reasoning Consensus: Structural Ensembling of LLM Reasoning via Weighted DAG Aggregation
    USarXiv cs.AI

    arXiv:2607.27783v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) reason in free-form chain-of-thought (CoT) without revealing which steps were well supported and which alternatives were rejected. In high-stakes…

  57. ClinFusion: A Vision-Centric Multimodal LLM System for Holistic Medical Understanding
    USarXiv cs.AI

    arXiv:2607.24743v3 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) hold immense potential to revolutionize clinical practice, yet deploying them in the medical domain is fundamentally a vision-centric…

  58. Routing Subspaces: Auditing Evaluation-to-Deployment Mismatch in Fine-Tuned Language Models
    USarXiv cs.AI

    arXiv:2607.20436v2 Announce Type: replace-cross Abstract: Safety evaluations often assume that behavior observed during testing reflects behavior in ordinary use, but fine-tuning can break this assumption. A checkpoint can appear fixed…

  59. AIGB-R1: Self-Evolving Generative Auto-Bidding via Hierarchical Planner-Executor Optimization
    USarXiv cs.AI

    arXiv:2607.17281v2 Announce Type: replace-cross Abstract: Auto-bidding plays an essential role in online advertising, automatically adjusting bids for advertisers to optimize their commercial goals. The emerging AI-Generated Bidding (AIGB)…

  60. Towards Continual Motion-Language Agents: LoRA Variants for Incremental Motion Understanding and Generation
    USarXiv cs.AI

    arXiv:2606.30266v2 Announce Type: replace-cross Abstract: Motion-language agents must possess the bidirectional capability to both understand human movement (motion-to-text, M2T) and generate it from natural language (text-to-motion, T2M)…

  61. SemFlowRAG: Directed Semantic Flow from Abstraction to Evidence for Complex Reasoning
    USarXiv cs.AI

    arXiv:2606.28447v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) enhanced by Knowledge Graphs has shown promise in complex multi-hop reasoning tasks. However, existing graph-based retrieval methods typically…

  62. Dual Dimensionality for Local and Global Attention
    USarXiv cs.AI

    arXiv:2606.18587v2 Announce Type: replace-cross Abstract: Decoder-only Transformers compute attention over the KV cache of preceding tokens. Keys and Values are typically represented with the same dimensionality, regardless of their…

  63. EV9V: An Echocardiographic Video Benchmark with Uncertainty Guided View Classification
    USarXiv cs.AI

    arXiv:2606.17437v2 Announce Type: replace-cross Abstract: Automated classification of standard echocardiographic views is crucial for efficient clinical workflow but faces three main challenges. First, publicly available datasets are…

  64. Non-frontal face recognition using GANs and memristor-based classifiers
    USarXiv cs.AI

    arXiv:2606.12074v2 Announce Type: replace-cross Abstract: Face recognition systems have advanced significantly through deep learning techniques, delivering high performance and robustness in complex scenarios. However, these approaches…

  65. GitInject: Real-World Prompt Injection Attacks in AI-Powered CI/CD Pipelines
    USarXiv cs.AI

    arXiv:2606.09935v2 Announce Type: replace-cross Abstract: AI agents are increasingly embedded in continuous integration and continuous delivery (CI/CD) workflows to autonomously review pull requests, triage issues, and maintain codebases…

  66. Adversarial Robustness of Activation Steering in Large Language Models
    USarXiv cs.AI

    arXiv:2606.07696v2 Announce Type: replace-cross Abstract: Activation steering has become a popular training-free method to control LLMs' behavior by injecting precomputed direction vectors into the model's residual stream at inference…

  67. NIV: Neural Axis Variations for Variable Font Generation
    USarXiv cs.AI

    arXiv:2606.05261v2 Announce Type: replace-cross Abstract: Variable fonts enable continuous variation of glyph geometry along semantic design axes such as weight, width, slant, and optical size. However, constructing a variable font from a…

  68. A Systematic Analysis of Linguistic Features in AI-Generated English Text Detection Across Domains and Models
    USarXiv cs.AI

    arXiv:2606.04177v2 Announce Type: replace-cross Abstract: Interpretable linguistic features offer a promising approach for explaining why a given text appears machine-generated, particularly for non-expert users. However, existing findings…

  69. Improvise, Adapt, Overcome: An On-The-Fly Multifidelity Algorithm for Efficient Machine Learning
    USarXiv cs.AI

    arXiv:2606.02662v2 Announce Type: replace-cross Abstract: Machine learning has accelerated quantum chemistry but is hindered by the prohibitive cost of generating high fidelity training data. Multifidelity machine learning (MFML) mitigates…

  70. Selective Linkage Precision for Released Medical Vision-Language Models
    USarXiv cs.AI

    arXiv:2606.02276v2 Announce Type: replace-cross Abstract: Hospitals may release de-identified chest radiographs while restricting access to their reports. An attacker who already has reports from the same population and their metadata can…

  71. Authenticity Debt and the Synthetic Content Threat Landscape: A Layered Framework for Trust, Provenance, and IP Governance in the Generative AI Era
    USarXiv cs.AI

    arXiv:2606.00621v2 Announce Type: replace-cross Abstract: Generative artificial intelligence has fundamentally changed how content is now produced. It has enabled how high-fidelity text, images, audio, and videos are created, modified, and…

  72. Beyond Augmentation: Score-Guided Pathological Prior for EEG-based Depression Detection
    USarXiv cs.AI

    arXiv:2606.00180v3 Announce Type: replace-cross Abstract: Deep learning-based Major Depressive Disorder (MDD) detection using Electroencephalography (EEG) is fundamentally constrained by the "small-sample dilemma." Prevailing generative…

  73. Curriculum Learning for Safety Alignment
    USarXiv cs.AI

    arXiv:2605.26315v2 Announce Type: replace-cross Abstract: Direct Preference Optimization (DPO) is a widely used approach for safety alignment that aims to reduce harmful behaviors in large language models. However, prior work shows that it…

  74. Deep learning-based self-supervised detection and localization of magnetic resonance imaging anomalies in artificial intelligence-supported radiotherapy workflows
    USarXiv cs.AI

    arXiv:2605.24609v3 Announce Type: replace-cross Abstract: Artificial intelligence is increasingly integrated into radiotherapy workflows, yet it remains vulnerable to out-of-distribution image data that can introduce unexpected behavior in…

  75. Beyond Isotropy in JEPAs: Hamiltonian Geometry and Symplectic Prediction
    USarXiv cs.AI

    arXiv:2605.20107v2 Announce Type: replace-cross Abstract: JEPAs often regularize one-view embeddings toward an isotropic Gaussian, implicitly baking Euclidean symmetry into the representation. We show that this is not merely a benign…

  76. Efficient Factorized Volume Synthesis with Lifted Inter-Slice Feature Trajectories
    USarXiv cs.AI

    arXiv:2605.19060v2 Announce Type: replace-cross Abstract: High-resolution 3D medical image generation remains challenging because fully volumetric models are computationally expensive, while efficient 2D slice generators often fail to…

  77. ReLATE: Learning Relative Representations for Post-Hoc Multimodal Alignment
    USarXiv cs.AI

    arXiv:2605.16834v2 Announce Type: replace-cross Abstract: Post-hoc multimodal alignment connects independently pretrained vision and language encoders while keeping them frozen, offering an alternative to joint pretraining when paired data…

  78. Acceptance Cards:A Four-Diagnostic Standard for Safe Fine-Tuning Defense Claims
    USarXiv cs.AI

    arXiv:2605.10575v2 Announce Type: replace-cross Abstract: Safe fine-tuning defenses are often endorsed on the basis of a held-out gap reduction, but the same reduction can come from sampling noise, subject artifacts, capability loss, or a…

  79. Pacing Branch Parallelism in LLM Serving
    USarXiv cs.AI

    arXiv:2605.06914v2 Announce Type: replace-cross Abstract: Recent methods expose intra-request parallelism in LLM outputs, allowing independent branches to decode concurrently. Existing serving systems execute these branches eagerly or…

  80. Quantizing With Randomized Hadamard Transforms: Efficient Heuristic Now Proven
    USarXiv cs.AI

    arXiv:2605.06014v2 Announce Type: replace-cross Abstract: Uniform random rotations (URRs) are a common preprocessing step in modern quantization approaches used for gradient compression, inference acceleration, KV-cache compression, model…

  81. Bolek: A Multimodal Language Model for Molecular Reasoning
    USarXiv cs.AI

    arXiv:2605.02745v2 Announce Type: replace-cross Abstract: Molecular property models increasingly support high-stakes drug-discovery decisions, but their outputs are often difficult to audit: classical predictors return scores without…

  82. Reinforcement Learning with Markov Risk Measures and Multipattern Risk Approximation
    USarXiv cs.AI

    arXiv:2605.00654v2 Announce Type: replace-cross Abstract: For a risk-averse finite-horizon Markov Decision Process, we introduce a special class of Markov coherent risk measures, called mini-batch measures. We also define the class of…

  83. RaMP: Runtime-Aware Megakernel Polymorphism for Mixture-of-Experts
    USarXiv cs.AI

    arXiv:2604.26039v2 Announce Type: replace-cross Abstract: LLM serving systems autotune their GPU kernels offline by timing candidate kernel configurations, like tile sizes, at a range of batch sizes. This works when the number of tokens…

  84. IndicDBBench - Benchmarking Multilingual Text-to-SQL Capabilities in Indian Languages
    USarXiv cs.AI

    arXiv:2604.13686v2 Announce Type: replace-cross Abstract: While Large Language Models (LLMs) have significantly advanced Text-to-SQL performance, existing benchmarks largely focus on Western contexts leaving a gap in real-world…

  85. SYN-DIGITS: A Synthetic Control Framework for Calibrated Digital Twin Simulation
    USarXiv cs.AI

    arXiv:2604.07513v2 Announce Type: replace-cross Abstract: Persona simulation based on large language models---often referred to as digital twin simulation---is increasingly used for market research, recommender systems, and social science…

  86. Fooling Reward Models with Gibberish: Adversarial Examples Beyond the Semantic Regime
    USarXiv cs.AI

    arXiv:2604.02686v2 Announce Type: replace-cross Abstract: Reward models (RMs) are widely used as optimization targets in reinforcement learning from human feedback (RLHF). We show that leading RMs can assign higher scores to gibberish than…

  87. Sub-Band Spectral Matching with Localized Score Aggregation for Robust Anomalous Sound Detection
    USarXiv cs.AI

    arXiv:2603.13749v2 Announce Type: replace-cross Abstract: Detecting subtle deviations in noisy acoustic environments is central to anomalous sound detection (ASD). A common training-free ASD pipeline temporally pools frame-level…

  88. Revisiting RAG Retrievers: An Information Theoretic Benchmark
    USarXiv cs.AI

    arXiv:2602.21553v2 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) systems rely critically on the retriever module to surface relevant context for large language models. Although numerous retrievers have…

  89. StretchTime: Adaptive Time Series Forecasting via Symplectic Attention
    USarXiv cs.AI

    arXiv:2602.08983v2 Announce Type: replace-cross Abstract: Transformer architectures have established strong baselines in time series forecasting, yet they typically rely on positional encodings that assume uniform, index-based temporal…

  90. Multi-layer Cross-attention is Provably Optimal for Multi-modal In-context Learning
    USarXiv cs.AI

    arXiv:2602.04872v4 Announce Type: replace-cross Abstract: Recent progress has rapidly advanced our understanding of the mechanisms underlying in-context learning in modern attention-based neural networks. However, existing results focus…

  91. Trustworthy Blockchain-based Federated Learning for Electronic Health Records: Identity-Gated Participation against Sybil Attacks
    USarXiv cs.AI

    arXiv:2602.02629v2 Announce Type: replace-cross Abstract: Federated learning allows hospitals to train a shared clinical model without exchanging patient records, but its aggregation assumes that each participant is a distinct institution…

  92. Low-Dimensional Adaptation of Rectified Flow: A Diffusion and Stochastic Localization Perspective
    USarXiv cs.AI

    arXiv:2601.15500v4 Announce Type: replace-cross Abstract: In recent years, Rectified flow (RF) has gained considerable popularity largely due to its generation efficiency and state-of-the-art performance. In this paper, we investigate how…

  93. Beyond Single-shot Writing: Deep Research Agents are Unreliable at Multi-turn Report Revision
    USarXiv cs.AI

    arXiv:2601.13217v2 Announce Type: replace-cross Abstract: Existing benchmarks for Deep Research Agents (DRAs) treat report generation as a single-shot writing task, which fundamentally diverges from how human researchers iteratively draft…

  94. PentestEval: Benchmarking LLM-based Penetration Testing with Modular and Stage-Level Design
    USarXiv cs.AI

    arXiv:2512.14233v2 Announce Type: replace-cross Abstract: Penetration testing is essential for assessing and strengthening system security against real-world threats, yet traditional workflows remain highly manual, expertise-intensive, and…

  95. Hierarchy-Aware Multimodal Unlearning for Medical AI
    USarXiv cs.AI

    arXiv:2512.09867v4 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) are increasingly used in sensitive domains such as medical AI, where privacy regulations, including HIPAA and GDPR, require the removal of…

  96. Prompt Less, Smile More: Semantic Engineering in Lieu of Prompt Engineering
    USarXiv cs.AI

    arXiv:2511.19427v2 Announce Type: replace-cross Abstract: AI-integrated applications increasingly rely on large language models (LLMs), but their intended behavior is not always fully specified by code. Meaning Typed Programming (MTP)…

  97. Beyond MACs: Deployment-Constrained Design Space Exploration for Tensorized Neural Networks
    USarXiv cs.AI

    arXiv:2511.17971v3 Announce Type: replace-cross Abstract: Tensor-train (TT) compression reduces neural-network parameters, but the latency of a tensorized layer is set by its contraction order and its mapping onto hardware: a contraction…

  98. InTAct: Interval-based Task Activation Consolidation for Continual Learning
    USarXiv cs.AI

    arXiv:2511.17439v3 Announce Type: replace-cross Abstract: Continual learning seeks to acquire new knowledge while preserving previously learned representations. Despite the success of prompt-based methods, they remain fragile in…

  99. AQBuc: Air Pollution Forecasting in Bucharest
    USarXiv cs.AI

    arXiv:2511.00532v2 Announce Type: replace-cross Abstract: In recent years, air pollution, especially the particulate matter $2.5\,\mu\text{m}$ (PM2.5), has become a growing concern mainly in urban areas. Exposure to air pollution is linked…

  100. AuthRank: Scoring Source Credibility to Counter Poisoning in CTI LLM Pipelines
    USarXiv cs.AI

    arXiv:2510.20768v3 Announce Type: replace-cross Abstract: Large Language Model (LLM) pipelines have become essential for processing Cyber Threat Intelligence (CTI) to quickly extract updated threat information from open source feeds and…

  101. AgentTypo: Adaptive Typographic Prompt Injection Attacks against Black-box Multimodal Agents
    USarXiv cs.AI

    arXiv:2510.04257v2 Announce Type: replace-cross Abstract: Multimodal agents built on large vision-language models are increasingly deployed in open-world settings but remain highly vulnerable to prompt injection. Typographic text-in-image…

  102. DELTA: Robustly Training Label-Conditional Diffusion Models with Weak Annotations
    USarXiv cs.AI

    arXiv:2510.03016v3 Announce Type: replace-cross Abstract: While label-conditional diffusion models have recently exhibited remarkable generative capabilities, their success heavily relies on massive, cleanly labeled datasets. In practice…

  103. Functional Critics for Off-Policy Actor-Critic: From Stability to Exploration
    USarXiv cs.AI

    arXiv:2509.22964v5 Announce Type: replace-cross Abstract: Actor-critic methods are central to modern reinforcement learning, yet off-policy actor-critic theory remains far removed from practical algorithms, due in part to the interaction…

  104. Agentic Software Engineering: Foundational Pillars and a Research Roadmap
    USarXiv cs.AI

    arXiv:2509.06216v4 Announce Type: replace-cross Abstract: Agentic Software Engineering (SE 3.0) represents a new era where intelligent agents are tasked not with simple code generation, but with achieving complex, goal-oriented SE…

  105. EAvatar: Expression-Aware Head Avatar Reconstruction with Generative Geometry Priors
    USarXiv cs.AI

    arXiv:2508.13537v2 Announce Type: replace-cross Abstract: High-fidelity head avatar reconstruction plays a crucial role in AR/VR, gaming, and multimedia content creation. Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated…

  106. Generative Logic: A Deterministic Reasoning Engine
    USarXiv cs.AI

    arXiv:2508.00017v5 Announce Type: replace-cross Abstract: We present Generative Logic (GL), a deterministic computer architecture for automated mathematical reasoning. GL starts from axiomatic definitions written in a minimalist…

  107. Large Language Models for Crash Detection in Video: A Survey of Methods, Datasets, and Challenges
    USarXiv cs.AI

    arXiv:2507.02074v3 Announce Type: replace-cross Abstract: Crash detection from video feeds is a critical problem in intelligent transportation systems. Recent developments in large language models (LLMs) and vision-language models (VLMs)…

  108. Learning to Charge More: How Q-Learning Firms Lock In Supracompetitive Prices
    USarXiv cs.AI

    arXiv:2505.22909v2 Announce Type: replace-cross Abstract: There is growing experimental evidence that $Q$-learning agents can learn to charge supracompetitive prices. We ask when such behavior persists, whether it emerges, and which…

  109. Narrative-Centered Emotional Reflection: An Early Prototype for AI-Supported Emotional Self-Reflection
    USarXiv cs.AI

    arXiv:2504.20342v3 Announce Type: replace-cross Abstract: Reflexion is an AI-powered prototype designed to explore structured emotional self-reflection. By integrating emotion detection, layered reflective prompting, and metaphorical…

  110. Speculative Ad-hoc Querying
    USarXiv cs.AI

    arXiv:2503.00714v2 Announce Type: replace-cross Abstract: Large language models (LLMs) invite substantial effort and investment, yet agentic systems can spend much of their time waiting for tools. This paper explores how a small auxiliary…

  111. Large Language Model Guided Self-Debugging Code Generation
    USarXiv cs.AI

    arXiv:2502.02928v3 Announce Type: replace-cross Abstract: Automated code generation is gaining significant importance in intelligent computer programming and system deployment. However, current approaches often face challenges in…

  112. MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data
    USarXiv cs.AI

    arXiv:2403.11207v3 Announce Type: replace-cross Abstract: Reconstructions of visual perception from brain activity have improved tremendously, but the practical utility of such methods has been limited. This is because such models are…

  113. Multiscale Policy Optimization for Long-Horizon LLM Agents
    USarXiv cs.AI

    arXiv:2610.35013v2 Announce Type: replace Abstract: Long-horizon LLM agents interact with their environments through complete assistant turns while generating each action token by token. We observe that critic values vary primarily across…

  114. GraphEvoResearcher: A Cost-Matched Mechanism Analysis of Explicit Reasoning Graphs in Training-Free Self-Reflection
    USarXiv cs.AI

    arXiv:2610.25978v2 Announce Type: replace Abstract: Training-free self-reflective protocols (generate, critique, revise) improve LLM reasoning but operate on a flat token stream: revisions cannot identify which prior step is unreliable…

  115. A Survey on the Linear Representation Hypothesis
    USarXiv cs.AI

    arXiv:2609.22695v2 Announce Type: replace Abstract: The term "linear representation hypothesis" (LRH) has appeared across diverse subfields of artificial intelligence, neuroscience, and cognitive science. But previous works have not…

  116. GRATE: Temporal Extensions for Inductive KG Foundation Models via Gated Rotary Attention
    USarXiv cs.AI

    arXiv:2607.10197v2 Announce Type: replace Abstract: Knowledge graph foundation models such as Ultra and Trix achieve strong inductive transfer by learning relation-graph representations that generalise to unseen entities and relations…

  117. Harnessing Agent Skills: Architectural Patterns and a Reference Architecture for Skill-Mediated LLM Agents
    USarXiv cs.AI

    arXiv:2606.20631v2 Announce Type: replace Abstract: CONTEXT: Agent skills package reusable behavioural guidance and resources that large language model agents load or apply during a run. Within the agent harness, agent skill harnessing…

  118. Rubric-Guided Process Reward for Stepwise Model Routing
    USarXiv cs.AI

    arXiv:2605.29310v3 Announce Type: replace Abstract: Stepwise model routing improves the efficiency of Large Reasoning Models (LRMs) by assigning each reasoning step to a suitable model. Recent methods formulate routing as a sequential…

  119. Anchor: Preventing Artifact Drift for Reliable Agent Verification
    USarXiv cs.AI

    arXiv:2605.26321v2 Announce Type: replace Abstract: Reliable evaluation of long-horizon agents requires verifiers that match what the task actually asks. Verifiable rewards have driven progress in domains like math and coding, yet building…

  120. Isomorphism Is Not Enough: An Agentic Audit of Verified Code Generation in Lean
    USarXiv cs.AI

    arXiv:2605.23772v2 Announce Type: replace Abstract: Agentic systems have recently emerged as state-of-the-art approaches for automated theorem proving in formal mathematics. To assess how far these capabilities extend to program…

  121. Teacher-Aware Evolution of Heuristic Programs from Learned Optimization Policies
    USarXiv cs.AI

    arXiv:2605.10634v2 Announce Type: replace Abstract: LLM-based automatic heuristic design has shown promise for generating executable heuristics for combinatorial optimization, but existing methods mainly rely on delayed endpoint…

  122. Can You Break RLVER? Probing Adversarial Robustness of RL-Trained Empathetic Agents
    USarXiv cs.AI

    arXiv:2605.07138v2 Announce Type: replace Abstract: Reinforcement learning from verifiable emotion rewards (RLVER) yields strong empathetic language models, but is evaluated with cooperative, honest simulated users. Real users gaslight…

  123. The Circuit Basis of Metacognitive Failure in Language Models: Interference and Hallucination
    USarXiv cs.AI

    arXiv:2605.05686v4 Announce Type: replace Abstract: Language models often answer with high confidence, as measured by output entropy, even when they are wrong. We study two situations that lead to such errors, knowledge conflict, where the…

  124. Foundation Models to Unlock Real-World Evidence from Nationwide Medical Claims
    USarXiv cs.AI

    arXiv:2605.02740v3 Announce Type: replace Abstract: Real-world data increasingly inform regulatory and clinical decisions. Administrative claims, among their major sources, continuously track care and its cost across providers throughout…

  125. WMIP: A Working-Memory-Inspired Programmable Probe of Cumulative State Tracking in Large Language Models
    USarXiv cs.AI

    arXiv:2603.27343v3 Announce Type: replace Abstract: Human working memory lets people maintain and update information. Large language models (LLMs) face a related demand when tracking updates within a single prompt without writing out…

  126. Representational Acquisition Under Bounded Resources
    USarXiv cs.AI

    arXiv:2603.21736v3 Announce Type: replace Abstract: A learner's obligation to keep acquired content usable imposes a storage-access tradeoff, with exact endpoints that force different physical organizations. For an exact concept-learning…

  127. Bi-directional digital twin prototype anchoring with multi-periodicity learning for few-shot fault diagnosis
    USarXiv cs.AI

    arXiv:2603.07054v2 Announce Type: replace Abstract: Intelligent fault diagnosis (IFD) typically requires abundant labeled data, which are difficult to obtain from physical assets in practical industrial applications. Digital twins (DTs)…

  128. When Should LLMs Be Less Specific? Selective Abstraction for Reliable Long-Form Text Generation
    USarXiv cs.AI

    arXiv:2602.11908v4 Announce Type: replace Abstract: LLMs are widely used, yet they remain prone to factual errors that erode trust and limit adoption in high-risk settings. One approach to mitigate this risk is to equip models with…

  129. Dynamic Mixed-Precision Routing for Efficient Multi-step LLM Interaction
    USarXiv cs.AI

    arXiv:2602.02711v3 Announce Type: replace Abstract: Large language models (LLMs) achieve strong performance in long-horizon agentic tasks through multi-step interaction and reasoning at test time. However, repeatedly invoking these models…

  130. ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge
    USarXiv cs.AI

    arXiv:2512.20276v2 Announce Type: replace Abstract: Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. However, their…

  131. Surrogate Modeling and Explainable Artificial Intelligence for Complex Systems: A Workflow for Automated Simulation Exploration
    USarXiv cs.AI

    arXiv:2510.16742v3 Announce Type: replace Abstract: Complex systems are increasingly explored through simulation-driven workflows that combine physics-based and empirical models with optimization and analytics. Despite their power, these…

  132. SynBench: A Benchmark for Differentially Private Text Generation
    USarXiv cs.AI

    arXiv:2509.14594v3 Announce Type: replace Abstract: Synthetic text generation with Differential Privacy (DP) guarantees emerges as a principled approach that can enable the sharing of sensitive datasets across institutional and regulatory…

  133. A Comprehensive Review of Multimodal Large Language Models: Performance and Challenges Across Different Tasks
    USarXiv cs.AI

    arXiv:2408.01319v2 Announce Type: replace Abstract: In an era defined by the explosive growth of data and rapid technological advancements, Multimodal Large Language Models (MLLMs) stand at the forefront of artificial intelligence (AI)…

  134. Pixel: A Co-Designed Full-Stack Robot System for Learning General-Purpose Mobile Manipulation
    USarXiv cs.AI

    arXiv:2611.03648v1 Announce Type: cross Abstract: General-purpose mobile manipulation requires a combination of capable robot hardware, scalable demonstration collection, expressive whole-body control, and effective policy learning. Yet…

  135. Tracing the Scientific Discovery Capacity of LLMs to Their Training Data
    USarXiv cs.AI

    arXiv:2611.03617v1 Announce Type: cross Abstract: Large language models (LLMs) show growing promise for scientific discovery, yet basic questions remain about what enables reliable discovery. How can we trace the scientific discovery…

  136. Evaluating Uncertainty Calibration in Probabilistic Time Series Foundation Models
    USarXiv cs.AI

    arXiv:2611.03592v1 Announce Type: cross Abstract: Time series foundation models (TSFMs) are increasingly used for probabilistic forecasting across diverse domains, yet the reliability of their uncertainty estimates remains poorly…

  137. Sonar-MoE: In-Backward Expert Updates for Memory-Efficient Mixture-of-Experts Training
    USarXiv cs.AI

    arXiv:2611.03577v1 Announce Type: cross Abstract: Most parameters of a Mixture-of-Experts (MoE) model belong to its experts, and standard training keeps every expert gradient in GPU memory until the optimizer step. This memory grows with…

  138. Contrastive Adapter Training and Consensus Knowledge Distillation for Multi-Source-Free Domain Adaptation in Skin Cancer Diagnosis
    USarXiv cs.AI

    arXiv:2611.03528v1 Announce Type: cross Abstract: Skin cancer diagnosis, particularly differentiating melanoma from benign nevi, is challenging due to the visual similarity between lesions. Although convolutional neural networks (CNNs) and…

  139. PanoTac: A Dense Three-Axis Tactile Sensor for Robust Touch-Centric Manipulation
    USarXiv cs.AI

    arXiv:2611.03520v1 Announce Type: cross Abstract: Contact-rich manipulation benefits from tactile feedback that captures both detailed contact geometry and interaction forces. When scooping grains, contact patterns help infer spoon…

  140. A Cross-Domain Matrix Engine for Signal Processing and Deep Neural Net Inference
    USarXiv cs.AI

    arXiv:2611.03503v1 Announce Type: cross Abstract: Matrix multiplication dominates both PHY signal processing and Deep Neural Network (DNN) inference, yet these domains place opposing demands on the matrix engine. Signal processing requires…

  141. From Reactive to Agentic Resilience: Multi-Timescale AI-RAN Orchestration Under Signaling Storms
    USarXiv cs.AI

    arXiv:2611.03406v1 Announce Type: cross Abstract: Next-generation open radio access network (Open RAN) and AI-RAN deployments turn the RAN into a disaggregated, compute-aware platform whose service continuity depends on many control loops…

  142. TrafficInstruct: A Multi-Task Network Traffic Dataset and Benchmark for Instruction Tuning
    USarXiv cs.AI

    arXiv:2611.03319v1 Announce Type: cross Abstract: Instruction tuning for multi-task network traffic analysis requires data with clearly defined sources, supervision targets, and evaluation conditions. We introduce TrafficInstruct, a newly…

  143. SQLoom: Set-Contrastive Learning of Syntactic, Semantic, and Logical Dependencies for Text-to-SQL
    USarXiv cs.AI

    arXiv:2611.03298v1 Announce Type: cross Abstract: Text-to-SQL translates natural language questions into SQL queries, enabling database access without writing SQL. Fine-tuning open-weight large language models (LLMs) is cost-effective, but…

  144. Rethinking Fairness Metrics for Multi-Class and Multi-Group Assessment: Not Everything is Binary
    USarXiv cs.AI

    arXiv:2611.03257v1 Announce Type: cross Abstract: Artificial Intelligence, and Machine Learning in general, is becoming pervasive across all domains and disciplines. By means of this data-driven approach, algorithms can learns complex task…

  145. CERBO: Certified Recourse at a Requested Rate for Gradient-Boosted Trees
    USarXiv cs.AI

    arXiv:2611.03222v1 Announce Type: cross Abstract: Recourse-aware learning trains models so that people who receive an adverse decision can reverse it with an affordable change. For gradient-boosted decision trees (GBDTs), the existing…

  146. Reducing SNAP Payment Errors at Lower Cost With AI: A Simulation Study
    USarXiv cs.AI

    arXiv:2611.03213v1 Announce Type: cross Abstract: New federal legislation will require states with high payment error rates in the Supplemental Nutrition Assistance Program (SNAP) to pay a share of benefit costs, potentially costing some…

  147. LaBraM at Scale: Scaling Data and Models for EEG Pretraining
    USarXiv cs.AI

    arXiv:2611.03165v1 Announce Type: cross Abstract: LaBraM, an early EEG foundation model, demonstrated that large-scale EEG pretraining can yield strong representations for diverse downstream tasks. Building on this foundation, we expand…

  148. Visual-to-Executable Procedural Reconstruction of Buildings
    USarXiv cs.AI

    arXiv:2611.03143v1 Announce Type: cross Abstract: Procedural modeling offers compact, editable, and reusable representations for architectural content, but building such models still depends heavily on manually authored grammars. We study…

  149. An Experimental Study of Hybrid Retrieval, NLI and Open-source LLMs for Scientific Claim Verification
    USarXiv cs.AI

    arXiv:2611.03093v1 Announce Type: cross Abstract: Claim verification is the task of deciding whether a source text supports a claim, contradicts it, or lacks sufficient evidence. Natural Language Inference (NLI) models can judge whether a…

  150. MIRA: Reinforcing Multimodal Reasoning via Deceptive Contextual Augmentation
    USarXiv cs.AI

    arXiv:2611.03087v1 Announce Type: cross Abstract: While Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced multimodal reasoning, existing frameworks suffer from a critical "perception-reasoning gap." Due to an…