AX·NEWS
Sign in

Artificial Intelligence

Artificial Intelligence news — Wednesday 28 October 2026

150 stories from 12 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. Apple Launches Touch Screen MacBook in 16 years...The Principle of Steve Jobs
    KRAI Times (Korea) Translated from Korean

    Apple released the MacBook Pro with touchscreen since its founder. Since Steve Jobs Founder strongly criticized the laptop touchscreen in 2010, it will reveal its existing entry in 16 years. The new MacBook Pro introduced to OLED display…

  2. [October 28] The essential question “What is the human beings” by AI Duned...Is the debate of the existing material value
    KRAI Times (Korea) Translated from Korean

    The debate of human creativity started by AI writing and painting is now a fundamental question to ask human identity and existence value. AI is the opinion of experts that engage in the discovery of veterinary science and develop into the…

  3. Google Launches AI Performance Evaluation in 1 Year..."Jeminai 4 Carbon"
    KRAI Times (Korea) Translated from Korean

    Google has released the world’s first in AI model comprehensive performance evaluation in one year. The newly revealed "Gemini 4 Carbon" has come to the top with the "Crawl Offer 5.5" and the "GPT-6 Astra" of OpenAI in the ATIFIC Analyst…

  4. Creating AI value with enterprise context
    USMIT Technology Review Biotech

    Across a business, any number of AI platforms, systems, and tools may be at work simultaneously—for example, AI-powered customer platforms, conversational assistants, and workflow and process automation tools. Each one is sophisticated…

  5. Chinese AI invasion gives enterprises a new way to control cost
    GBAI Business

    The open-weight models are, at least on the surface, considerably cheaper than their U.S. proprietary counterparts.

  6. Nvidia bets on personal agents in new Meta partnership
    GBAI Business

    The deal will provide inference for the Muse personal agent.

  7. Gemini 4 Carbon: Our next frontier model
    GBGoogle DeepMind
  8. OpenAI slashes price for model again
    GBAI Business

    The cut means enterprises have more options for lower-cost models, but many still struggle to decide which model to use for particular use cases.

  9. How to Make AI Work for End-to-End Business Processes
    USTowards Data Science

    From simplifying workflows to connecting knowledge and rules, coordinating AI agents, and delivering measurable business results. The post How to Make AI Work for End-to-End Business Processes appeared first on Towards Data Science .

  10. NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
    USHugging Face Blog
  11. Where a Cell Sits Matters More Than How Many There Are
    USTowards Data Science

    Using spatstat raster layers and point process models to ask how neurons, astrocytes and microglia sit relative to each other The post Where a Cell Sits Matters More Than How Many There Are appeared first on Towards Data Science .

  12. OpenAI and ServiceNow expand partnership across enterprise workflows
    USOpenAI Blog

    OpenAI and ServiceNow expand their partnership to bring OpenAI models and agents into ServiceNow and ServiceNow workflows into ChatGPT and Codex.

  13. Building and Visualising Kolmogorov–Arnold Networks
    USTowards Data Science

    Exploring what happens when neural networks learn functions instead of weights. The post Building and Visualising Kolmogorov–Arnold Networks appeared first on Towards Data Science .

  14. The Download: dead startups fueling AI and interspecies communication
    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. AI labs are buying data from dead startups to train their models When Keelan Miskell began winding…

  15. Pairing industrial AI autonomy and human control
    USMIT Technology Review Biotech
  16. Your RAG Pipeline Is Missing a 1994 Algorithm
    USTowards Data Science

    Embeddings know that "broken" means "defective". They may not notice that ZX-4821 is not ZX-4812. The post Your RAG Pipeline Is Missing a 1994 Algorithm appeared first on Towards Data Science .

  17. To Tackle Alzheimer’s, Scientists Turn To The Brain’s Own Immune System
    SGAsian Scientist

    Blocking an immune checkpoint helped restore the function of microglia, the brain’s resident immune cells, while reducing toxic plaques and abnormal neuronal activity in mice.

  18. The longest proof of the mathematics, the simplicity sorting machine verification completion
    KRAI Times (Korea) Translated from Korean

    Axiom, a professional startup at Mathematics AI, has succeeded to validate the full proof of 'CFSG·Classification of Finite Simple Groups' as the most comprehensive and complex proof of math history. This huge system, called the…

  19. Requid AI, OnDevice AI Model ‘LFM3’ released.. .Brauzer·Voice·Immigration processing support
    KRAI Times (Korea) Translated from Korean

    Requid AI revealed a new AI model suite ‘LFM3’ that allows businesses and individuals to run AI agents and voice assistants, image/video recognition functions directly. It consists of six small models of 20 billion parameters, and designed…

  20. 'Deepclock-V4.1 Flash', independent tester control room·safety false exposure
    KRAI Times (Korea) Translated from Korean

    Deepshik’s latest model was designed to have improved control over the previous version in an independent safety assessment, but still revealed serious safety issues, such as preventing system shutdown or avoiding monitoring. In…

  21. “Annex, AI development from the summer of next year”...Expert evaluation company view
    KRAI Times (Korea) Translated from Korean

    In July to August, Antropice has been a view to develop the next generation AI model without the help of human researchers. The result of the high-tech AI model’s R&D capability, and the performance of the ANTRONIC has been improved at…

  22. The future of the firm
    USOpenAI Blog

    How AI might enter the org chart—beneath, beside, and between us.

  23. Legora is building the agentic operating system for legal work
    USOpenAI Blog

    See how Legora uses OpenAI models, ChatGPT, and Codex to accelerate legal workflows, expand contract analysis, and free lawyers to focus on higher-value work.

  24. Webshopping, Joining Po... K-existence to financial AI agent
    KRAI Times (Korea) Translated from Korean

    WebCesi (CEO Kangwon) announced that it will take part in the LG AI Research Institute Contest of the ‘Independent AI Foundation Model’ project, which will lead to the development and implementation of AI agent for financial services. The…

  25. [AI Model Story] 6 types of OCR publications read more than 100 pages
    KRAI Times (Korea) Translated from Korean

    ■ Baidu, one hundred pages document analyzed at a time, the ‘Unlimited OCR 2’ open source public bindu has released an open source of optical character (OCR) model ‘Unlimited OCR 2’ which can read and analyze long documents on hundreds…

  26. [Bulletin] AI Sophia, International Cyber Crime Investigation Support
    KRAI Times (Korea) Translated from Korean

    ■ AI SPEA (CEO) announced that it will participate in the ‘gateway’ initiative of Interpol and Cyber Crime Contribution Agreement, which will be concluded with the International Supreme Prosecutor’s Interpol. The Gateway is a cyber threat…

  27. MS Launched the next generation security model ‘MAI Cyber 1.1 Flash’, which is 20% higher cost efficiency
    KRAI Times (Korea) Translated from Korean

    Microsoft (MS) released the Cyber Security AI model ‘MAI Cyber 1.1 Flash’. It is characterized by reducing cost per scan while maintaining vulnerability detection performance similar to existing models. We focus on increasing the…

  28. $20 million GPU Nohida...High-speed freight forwarder covered with high-tech crime
    KRAI Times (Korea) Translated from Korean

    In the U.S., the organizational cargo climax criminals who enter the AI data center in the U.S. Crime organizations combine advanced technologies such as hacking, phishing, GPS signal manipulation, vehicle collision, and more. After June…

  29. OpenAI Ears and Tropics 'Nanje' Semi-Film...Enlarged to compete, scientific discovery
    KRAI Times (Korea) Translated from Korean

    OpenAI announced that it has discovered an algorithm that uses an AI to evacuate the ‘ETH’, the key to computer science. The two companies have released the fixation performance of the same ovaries with only 6 hours intervals, and the…

  30. Antarpic, Authorization of ‘Extreme Poster’...Management
    KRAI Times (Korea) Translated from Korean

    Andropic has provided a system that has been previously assigned to the successor after the IPO. The latter symbol shall not be disclosed to the general shareholders. In the event that the investor will stop pressing to reduce the AI…

  31. ChatGPT, now connects to Apple iCloud Mail and Calendar
    KRAI Times (Korea) Translated from Korean

    OpenAI expands the external service integration of ‘ChatGPT’ and stands out on the Apple ecosystem. OpenAI announced that it supports plugins that can link Apple’s iCloud mail and calendar on chatGPT on 27th (local time). We use cookies to…

  32. MAI Image 2.7 2nd place of AA and 3rd place of Arena
    KRAI Times (Korea) Translated from Korean

    Microsoft (MS) released its own image editing AI model ‘MAI Image 2.7’. In performance evaluation, we’ve made an openAI model and we’ve come to the 2nd place in the Artistic Analyst (AA), and we’ve also won the 3rd place in the image…

  33. Vietnam and home photovoltaic support plan to expand...
    KRAI Times (Korea) Translated from Korean

    Vietnam’s government directed to establish a household roof solar support policy until December this year. In the last September, we plan to establish a separate solar supply target and energy storage device (ESS) expansion plan according…

  34. World Solar Power Generation, Won-Wood...13% ahead of last 12 months
    KRAI Times (Korea) Translated from Korean

    The world’s solar power has been analyzed by 13% of nuclear power generation in recent 12 months. In the past 10 years, solar power has increased more than 10 times, and the price drop of the module, and the battery supply has impacted…

  35. SK hynix opens a quarterly revenue of 100 trillion times.. . 20,000 KRW per week
    KRAI Times (Korea) Translated from Korean

    SK hynix has opened a quarterly revenue of 100 trillion KRW for the first time in the first time, which is powered to expand AI memory demand. SK hynix announced that it has recorded KRW 775 trillion in sales of KRW 104 trillion, operating…

  36. Nvidia·Oraq, 'Bala Lubin' Massive Alliance...GPU Overrun limit
    KRAI Times (Korea) Translated from Korean

    Nvidia and Oracle have expanded AI infrastructure cooperation, and presented a new computing structure to reduce the need for corporate GPUs. In addition to high-performance GPUs, the CPU resources responsible for executing and data…

  37. The Federal Court of Justice "Alivava Lounge Maggiore"...Depending on the Ministry of Defense
    KRAI Times (Korea) Translated from Korean

    Alababa took the braking to the action that restricts the lobby activities in relation to the blacklists of the U.S. Department of Defense. The United States Federal Court commanded that it does not prohibit the transaction of the company…

  38. Order 5 times in a stock ballroom in Meta, Article 33...Expansion of capital procurement scale
    KRAI Times (Korea) Translated from Korean

    In order to establish an AI infrastructure investment fund, the order of a total of $250 billion (about $33 trillion KRW) issued to the general public of the size was collected. As a result of the high interests of investors, Meta has…

  39. China AI bubble disruption accelerate...Technical rifle 640 trillion evaporation
    KRAI Times (Korea) Translated from Korean

    The market capitalization of the Chinese AI and semiconductor companies, and the market capitalization of $ 4800 billion (about 640 trillion KRW) was evaporated. The Chinese prosecutor of AI and semiconductor enterprises fell by 40% from…

  40. LGU+ Development of Liquid Cooling System for Cola and Data Center
    KRAI Times (Korea) Translated from Korean

    As the introduction of liquid cooling angle of AI data center (AIDC), LG Uplus will lead to the development of related cooling system operation standard. LGU+ (CEO Hong Buk) announced that it has signed a cooperative agreement (MOU) for…

  41. "Let's get a free health spot!".. .Started AI chat bot service based on 500 million governments and health data
    KRAI Times (Korea) Translated from Korean

    “AI Health Assistant Chatbot” service, which reflects the national health survey big data for over 500 million people, entered the formal operation. We’re here to help you make your own health. The Ministry of Science and Technology and…

  42. “GPU two-sided AI hospital buzz in four weeks”...Acrylic, Medical AX Market Summary
    KRAI Times (Korea) Translated from Korean

    Acrylic (CEO Park Sang-jin) announced that it has a full-fledged performance, such as supplying solutions to 17 hospitals in medical AX business. Acrylic has built its own medical AX solution lineup, such as ‘ALLM.H’, ‘NADIA’ and AI…

  43. "AI Cheating"... Geb Expands Smart Glasses Inspection Solution Supply
    KRAI Times (Korea) Translated from Korean

    We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.OkPrivacy policy In the last August, the AI test supervision solution ‘monito’ added…

  44. Apple, ‘M6’ chip-mounted imac... AI performance enhancement
    KRAI Times (Korea) Translated from Korean

    Apple released a 24-inch iMac and a 14-inch MacBook Pro with the latest M6 chip. Both products are similar to the previous generation, but it is characterized by increased AI processing performance and graphics performance. Apple announced…

  45. PLATINER, GS NEOTECH Co., Ltd.
    KRAI Times (Korea) Translated from Korean

    PLATIER (CEO Sanghoon) announced that it signed a cooperative agreement (MOU) for the development and business of GS Neotek (CEO) and ‘XGEN’ based AI contract center (AICC). With this affiliate, Yangjae will combine GS NEOTECH’s contract…

  46. Meta, AI infrastructure massive funding... NVIDIA 4 trillion, AMD 1.3 trillion bets
    KRAI Times (Korea) Translated from Korean

    Nvidia invests a total of $30 billion (about 4 trillion KRW) in Meta to expand AI infrastructure investment. At the same time, Meta participates in a large-scale equity contest that promotes building AI infrastructure, and also enhances…

  47. When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models
    USarXiv cs.CV

    arXiv:2610.37430v1 Announce Type: cross Abstract: For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet…

  48. When Slides Met Scans: Cross-Attention for Renal Cancer Prognosis
    USarXiv cs.CV

    arXiv:2610.36364v1 Announce Type: cross Abstract: Histopathology and computed tomography (CT) provide complementary views of renal cancer. We connect these modalities through bidirectional cross-attention between 16 histology prototypes…

  49. How Many and How Much? Contrastive Learning with Negative Samples for Knowledge Graph Embedding
    USarXiv cs.LG

    arXiv:2610.37389v1 Announce Type: new Abstract: Knowledge graph embedding (KGE), a fundamental technique for knowledge-centric Web and KG-augmented AI applications, is commonly trained through contrastive learning with negative samples to…

  50. Algorithmics for Graphical Separation Criteria
    USarXiv cs.LG

    arXiv:2610.36502v1 Announce Type: cross Abstract: Statistical graphical models are representations of the dependences between the components of a random vector. They encode these dependences via graphical separation criteria. We consider…

  51. FrameForge: Harnessing Priors of Video Foundation Models for Unified Image Generation
    USarXiv cs.CV

    arXiv:2610.36545v1 Announce Type: new Abstract: Unified conditional image generation requires a single model to preserve different aspects of a visual input while producing a target image that follows textual guidance. Pretrained video…

  52. TRUCO: A Scalable Lock-Free Algorithm for Parallel Contour Extraction
    USarXiv cs.CV

    arXiv:2610.37297v1 Announce Type: new Abstract: Transforming binary masks into geometric contours remains a critical step in many computer vision pipelines, particularly for post-processing high-resolution segmentation outputs. Yet…

  53. Taming Noise Initialization for Zero-shot Video Restoration with Image Diffusion Priors
    USarXiv cs.CV

    arXiv:2610.37290v1 Announce Type: new Abstract: Recent advances in diffusion models have achieved remarkable success in restoring high-quality images in a zero-shot manner. Unfortunately, extending this success to video restoration is…

  54. Gym-Anything-RT: Scaling Real-Time Environments for Vision-Language Agents
    USarXiv cs.LG

    arXiv:2610.37567v1 Announce Type: new Abstract: AI agents have made substantial progress in domains where they have a non-trivial amount of time to deliberate between actions, such as mathematics, software engineering, and performing tasks…

  55. Evaluation-Guided Demonstration Refinement for Vision-Language-Action Policies: The Winning Entry in the MetaFine Open Competition
    USarXiv cs.CV

    arXiv:2610.36765v1 Announce Type: new Abstract: We fine-tuned a separate $\pi_{0.5}$ policy for each of the five tasks in the MetaFine Open Competition, with a task-specific training mixture and number of actions executed per prediction…

  56. TextCloak: Safeguarding Data from Unauthorized LLM Fine-Tuning via RL-Driven Unlearnable Examples
    USarXiv cs.LG

    arXiv:2607.28862v2 Announce Type: replace-cross Abstract: The development of Large Language Models (LLMs) has led to significant advances across a wide range of language tasks, while simultaneously raising growing concerns about…

  57. Conditional Sampling via Wasserstein Autoencoders and Triangular Transport
    USarXiv cs.LG

    arXiv:2604.02644v2 Announce Type: replace Abstract: We present Conditional Wasserstein Autoencoders (CWAEs), a framework for conditional simulation that exploits low-dimensional structure in both the conditioned and the conditioning…

  58. Proposing Better Rollouts for On-Policy Distillation
    USarXiv cs.LG

    arXiv:2610.36477v1 Announce Type: new Abstract: On-policy distillation (OPD) improves language-model post-training by letting the student learn from its own sampled trajectories under teacher supervision, aligning the training process with…

  59. Exact Unlearning of Finetuning Data via Model Merging at Scale
    USarXiv cs.LG

    arXiv:2504.04626v2 Announce Type: replace Abstract: Approximate unlearning has gained popularity as an approach to efficiently update an LLM so that it behaves (roughly) as if it was not trained on a subset of data to begin with. However…

  60. Every Little Helps: Building Knowledge Graph Foundation Model with Fine-grained Transferable Multi-modal Tokens
    USarXiv cs.CL

    arXiv:2602.15896v2 Announce Type: replace Abstract: Multi-modal knowledge graph reasoning (MMKGR) aims to predict the missing links by exploiting both graph structure information and multi-modal entity contents. Most existing works are…

  61. Agent Planning Benchmark: A Diagnostic Framework for Planning Capabilities in LLM Agents
    USarXiv cs.CL

    arXiv:2606.04874v4 Announce Type: replace Abstract: Planning is central to LLM agents: before acting, an agent must decompose goals, select tools, reason over constraints, and decide when a task is infeasible. Yet existing agent…

  62. NextEval: An Agent-Based Solver and Benchmark for Derivative-Free Optimization
    USarXiv cs.LG

    arXiv:2610.36553v1 Announce Type: new Abstract: Derivative-free optimization (DFO) addresses problems where derivatives are unavailable or costly to obtain. This setting often arises in simulation-based and experimental applications, where…

  63. Who Convinced Whom? Tracing How LLMs Change Their Minds in Multi-Agent Systems
    USarXiv cs.AI

    arXiv:2610.35930v1 Announce Type: cross Abstract: Multi-agent debate (MAD) aims to improve language-model decisions through the exchange of arguments, yet agreement after debate need not reflect independent corroboration. We investigate…

  64. Variational Optimal Control for Diffusion Guidance
    USarXiv cs.LG

    arXiv:2610.37146v1 Announce Type: new Abstract: We present a variational control framework that connects sampling-time guidance and reward-based model adaptation through a shared Jacobian pullback. Starting from terminal reward…

  65. Bayesian Inference of Expensive Simulators: Survey and Comparison of Design Choices
    USarXiv cs.LG

    arXiv:2610.37284v1 Announce Type: new Abstract: We present a survey of approaches to Bayesian inference with simulator forward models, focusing on settings with prohibitively expensive simulators. In this domain, Bayesian…

  66. RAG-CSI: Global Local Retrieval via Joint State and Dynamics Matching for CSI Prediction
    USarXiv cs.LG

    arXiv:2610.36881v1 Announce Type: cross Abstract: Channel state information (CSI) prediction forecasts future channel states from historical observations, supporting transceiver design and resource allocation with reduced measurement…

  67. Online gas classification at the edge
    USarXiv cs.LG

    arXiv:2610.36999v1 Announce Type: cross Abstract: Field applications in chemical sensing, from portable safety monitors to mobile robots, require models that are lightweight, fast to respond, and can be recalibrated from limited labelled…

  68. Temporal difference learning adapts to incomplete coverage: an oracle inequality over subgraphs
    USarXiv cs.LG

    arXiv:2411.09731v3 Announce Type: replace Abstract: Temporal difference (TD) and Monte Carlo (MC) are two fundamental principles for value function estimation in reinforcement learning. TD is asymptotically efficient for value estimation…

  69. Orlicz-Sobolev with Musielak: An Efficient Regularization Approach for Graph-based IPM
    USarXiv cs.LG

    arXiv:2510.25591v2 Announce Type: replace Abstract: We study the Sobolev IPM problem for probability measures supported on a graph metric space, where critic function is constrained to lie within the unit ball defined by Sobolev norm…

  70. Evaluating Priority Dispatch for Premium Membership in On-demand Taxi-sharing
    USarXiv cs.AI

    arXiv:2610.37011v1 Announce Type: cross Abstract: In this paper, we analyze the dispatch-level feasibility of introducing premium membership into an on-demand taxi-sharing service using a sequential optimal insertion method. We propose two…

  71. Climate downscaling with likelihood functional flow matching
    USarXiv cs.LG

    arXiv:2610.33482v2 Announce Type: replace Abstract: Decisions about heat, frost and flooding are made locally, yet reanalyses and climate models resolve tens of kilometres. Generative downscalers add kilometre-scale detail with…

  72. INTRA: Interleaved Non-contiguous Token spaRse Attention
    USarXiv cs.LG

    arXiv:2610.34056v2 Announce Type: replace Abstract: Transformers achieve strong performance across modalities but are bottlenecked by the quadratic cost of attention. Sparse attention can reduce this cost, but effective sparsity must…

  73. Knowledge distillation across teacher architectures for compact diabetic retinopathy grading
    USarXiv cs.CV

    arXiv:2610.35961v1 Announce Type: cross Abstract: Diabetic retinopathy grading from retinal fundus photographs is important but difficult to deploy on resource-constrained hardware. We conducted a controlled study of offline response-based…

  74. NAG-GS: A Semi-Implicit Stochastic Optimizer with Large-Step Stability
    USarXiv cs.LG

    arXiv:2209.14937v3 Announce Type: replace-cross Abstract: Stochastic first-order methods are widely used for large-scale optimization and machine learning, where their performance often depends on the learning rate. We study NAG-GS, a…

  75. Discovering quantum phenomena with Interpretable Machine Learning
    USarXiv cs.LG

    arXiv:2604.16015v2 Announce Type: replace-cross Abstract: Interpretable machine learning techniques are becoming essential tools for extracting physical insights from complex quantum data. We build on recent advances in variational…

  76. The proton structure one mode at a time: a Neural Tangent Kernel analysis of a DIS fit
    USarXiv cs.LG

    arXiv:2610.36034v1 Announce Type: cross Abstract: We study the training dynamics of a feed-forward neural network (NN) through the analysis of the neural tangent kernel (NTK) in a phenomenologically relevant fit of parton distribution…

  77. An Algebra of Transformers
    USarXiv cs.AI

    arXiv:2610.36795v1 Announce Type: cross Abstract: On the theoretic analyses on transforms, they mainly focus on the computability analysis, such the Turing-completeness of unbounded precision transformers, the advantage of…

  78. Sparse Attention as Compact Kernel Regression
    USarXiv cs.LG

    arXiv:2601.22766v4 Announce Type: replace Abstract: Recent work has revealed a link between self-attention mechanisms in transformers and test-time kernel regression via the Nadaraya-Watson estimator, with standard softmax attention…

  79. Scaling and Renormalization in Neural Networks: An Effective Renormalization-Group Description of Grokking
    USarXiv cs.LG

    arXiv:2610.36113v1 Announce Type: new Abstract: Grokking is a delayed transition from memorization to generalization in deep neural networks (NNs). We examine whether scaling and renormalization group (RG) theories can describe this…

  80. PsyJev: Psychological Analysis via System-1 Judgment
    USarXiv cs.AI

    arXiv:2610.36689v1 Announce Type: new Abstract: Self-disclosures on the Web support diverse psychological analyses, from recognizing expressed experiences to identifying symptom-relevant information. We study a shared, question-conditioned…

  81. As the Story Unfolds: Watching a Film and Identifying Characters as a Human Does
    USarXiv cs.CV

    arXiv:2610.37579v1 Announce Type: new Abstract: Films introduce characters incrementally as they progress. We study the problem of identifying principal characters and assigning their names directly from the movie itself, without using…

  82. Structured Time-Frequency Representation for Channel Prediction
    USarXiv cs.AI

    arXiv:2610.35956v1 Announce Type: cross Abstract: Accurate CSI forecasting is essential for mitigating channel aging and supporting reliable downlink transmission in high-mobility wireless systems. In FDD systems, forecasting future DL CSI…

  83. Formalizing and Estimating Task-Space Complexity for Zero-shot Generalization
    USarXiv cs.LG

    arXiv:2606.20967v2 Announce Type: replace Abstract: Generalization is crucial for modern control, but learning across tasks can be costly. Recent work shows that zero-shot transfer, in which a controller designed for a source task is…

  84. Strategyproof Deep Reinforcement Learning for Online Resource Allocation
    USarXiv cs.AI

    arXiv:2610.37504v1 Announce Type: cross Abstract: Online resource allocation faces two coupled challenges: (1) allocating scarce resources efficiently under uncertain future demand, and (2) ensuring that agents report task valuations and…

  85. TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems
    USarXiv cs.LG

    arXiv:2606.25147v2 Announce Type: replace-cross Abstract: User modeling in industrial recommender systems typically produces dense embeddings, which suffer from representational constraints due to fixed-dimensional vectors. As an emerging…

  86. The Theory of Strategic Evolution: Three Principles of Strategic Replicators
    USarXiv cs.AI

    arXiv:2512.07901v5 Announce Type: replace-cross Abstract: Strategic replicators combine evaluated choice and reproductive transmission. An agent chooses among alternatives; the consequences of those choices influence continuation or…

  87. TimeLineXR: An Interactive Exploration of Time in Vision-Language Models
    USarXiv cs.CV

    arXiv:2610.37349v1 Announce Type: new Abstract: Recently, it has been shown that vision-language models (VLMs) have a certain intrinsic understanding of time. They can estimate the temporal origin of objects and encode time along a…

  88. Know, then Draw: A New Learning Paradigm for Visual Generation
    USarXiv cs.CV

    arXiv:2610.37626v1 Announce Type: new Abstract: Visual generation requires both high-level semantic knowledge and fine-grained pixel details. We propose a learning paradigm that assigns these roles to autoregressive pretraining for…

  89. Repurposing Video Diffusion Transformers for Cross-View Temporal Object Correspondence
    USarXiv cs.CV

    arXiv:2610.37633v1 Announce Type: new Abstract: Cross-view object correspondence identifies the same object across views. However, existing methods often rely on frame-wise matching or sequential tracking, which can propagate…

  90. DifFRACT: Diffusion Feature Reconstruction and Attribution for Circuit Tracing
    USarXiv cs.CV

    arXiv:2606.15796v2 Announce Type: replace Abstract: While transcoder-based circuit tracing has recently enabled detailed causal analyses of large language models, it remains unclear how semantic information propagates across denoising…

  91. Persistence, Exposure, and Credit Routing in Linear Temporal Predictive Coding
    USarXiv cs.LG

    arXiv:2610.36412v1 Announce Type: new Abstract: Temporal Predictive Coding (tPC) networks offer a layer-local alternative to backpropagation. For linear tPC, we show, by applying tools from cellular sheaf theory, that temporal credit…

  92. From Ethernet to Wi-Fi: Flow Feature Selection and Construction for Traffic Classification
    USarXiv cs.LG

    arXiv:2610.37007v1 Announce Type: cross Abstract: Flow-based traffic classifiers trained on Ethernet data may lose accuracy when applied to Wi-Fi traffic. This paper proposes a five-stage procedure for selecting and constructing features…

  93. Multi-Label Agreement: A Flexible Framework for Inter-Annotator Reliability
    USarXiv cs.CL

    arXiv:2610.37292v1 Announce Type: new Abstract: Reliable evaluation of inter-annotator agreement (IAA) is important for assessing the consistency of annotation processes. While established IAA metrics such as Cohen's kappa, Fleiss' kappa…

  94. Systematic Morphometric Brain Autoencoding for Representation Learning in Neurodegenerative Diseases
    USarXiv cs.CV

    arXiv:2610.37074v1 Announce Type: new Abstract: Early identification of neurodegenerative disorders using structural T1-weighted MRI is essential for timely clinical intervention, yet supervised deep learning models often struggle to…

  95. Beyond Parameter Protection: Fisher-Shaped EWC for Replay-Free Continual Learning
    USarXiv cs.LG

    arXiv:2610.37145v1 Announce Type: new Abstract: In replay-free continual learning, Elastic Weight Consolidation (EWC) uses Fisher information matrix (FIM) to protect acquired knowledge by penalizing changes to historically important…

  96. Multibranch Ensemble Head in Vision Foundation Models for Efficient Cucumber Phenological Stage Classification
    USarXiv cs.CV

    arXiv:2610.37118v1 Announce Type: new Abstract: Vision Foundation Models (VFMs) pretrained with self-supervised learning, such as DINOv2, provide highly transferable features, yet adapting them to small, domain-shifted datasets is still…

  97. Evaluating AlphaEarth Embeddings for Crop Classification: Separability, Temporal Dynamics, and Cross-Regional Transferability
    USarXiv cs.CV

    arXiv:2610.36857v1 Announce Type: new Abstract: Earth Observation (EO)-based agricultural monitoring is shifting from pixel-level spectral analysis toward representation-driven modeling, yet systematic evaluation of learned representations…

  98. TS-DP: Reinforcement Speculative Decoding For Temporal Adaptive Diffusion Policy Acceleration
    USarXiv cs.LG

    arXiv:2512.15773v2 Announce Type: replace Abstract: Diffusion Policy (DP) excels in embodied control but suffers from high inference latency and computational cost due to multiple iterative denoising steps. Existing diffusion model (DMs)…

  99. NN-OpInf: an operator inference approach using structure-preserving composable neural networks
    USarXiv cs.LG

    arXiv:2603.08488v2 Announce Type: replace Abstract: We propose neural network operator inference (NN-OpInf): a structure-preserving, composable, and minimally restrictive operator inference framework for the non-intrusive reduced-order…

  100. Real-Time Smart Retail Analytics through Fog Computing using the RetailLightNet Lightweight Model
    USarXiv cs.CV

    arXiv:2610.35871v1 Announce Type: cross Abstract: Smart retail systems now depend on the application of real-time video analytics, which can track customer behavior, manage shelf inventory, and enhance in-store processes. Currently…

  101. Reliability-Aware Short-Term Glycemic Risk Prediction Using an Anchor-Based Comparative Framework of Sequence, Multimodal, and State-Updating Digital-Twin Models
    USarXiv cs.LG

    arXiv:2610.36363v1 Announce Type: new Abstract: Anticipating short-term hypoglycemic and hyperglycemic risk is essential for proactive diabetes management. However, existing machine learning approaches often use inconsistent evaluation…

  102. PRISM: Pathology Reliability In Scarce-label Medicine. A Tile-level Benchmark of Pathology Foundation Models
    USarXiv cs.CV

    arXiv:2610.36517v1 Announce Type: cross Abstract: Pathology foundation models are typically benchmarked on classification performance at full label availability and on in-distribution test sets, while calibration, behavior at low label…

  103. Less is more: learning what to profile for GPU kernel optimization
    USarXiv cs.LG

    arXiv:2610.36576v1 Announce Type: new Abstract: Large language models have recently shown strong potential for automatically optimizing GPU kernels, yet their effectiveness depends critically on the hardware information available to the…

  104. Retrospective macromolecular pool fraction mapping from conventional clinical MRI: a physics-guided self-supervised deep learning framework
    USarXiv cs.CV

    arXiv:2610.37213v1 Announce Type: cross Abstract: Macromolecular pool fraction (MPF) is a quantitative imaging biomarker sensitive to myelin integrity and other changes in tissue macromolecular content, but specialized acquisitions and…

  105. When Is a Cheap Fairness Edit Safe? Screening Post-Hoc Edits in Two-Sided Matching Markets
    USarXiv cs.LG

    arXiv:2610.36649v1 Announce Type: new Abstract: Two-sided matching systems, such as admissions pipelines, allocate opportunities jointly. Reducing a protected at tribute's influence usually requires encoder retraining. We study a…

  106. LoBE-GS: Load-Balanced and Efficient 3D Gaussian Splatting for Large-Scale Scene Reconstruction
    USarXiv cs.CV

    arXiv:2510.01767v3 Announce Type: replace Abstract: 3D Gaussian Splatting (3DGS) has established itself as an efficient representation for real-time, high-fidelity 3D scene reconstruction. However, scaling 3DGS to large and unbounded…

  107. Why and How to Do Science When AI Is Better at It: A Framework Grounded in Human Understanding and Direction
    USarXiv cs.AI

    arXiv:2610.36116v1 Announce Type: new Abstract: As AI begins to match or exceed human researchers at core scientific tasks, two questions become pressing: \emph{why} should humans keep doing science, and \emph{how} should they use AI when…

  108. Beyond the Evidence Coherence Trap: Agent-Driven Evidence Selection for Temporally Disordered Fake News Detection
    USarXiv cs.AI

    arXiv:2610.36764v1 Announce Type: new Abstract: Fake news undermines public trust and the integrity of online information ecosystems. Temporally disordered fake news is especially elusive because it rearranges factual events into a false…

  109. Who Knows What - and What Does It Support? Perspective-Bounded Evidence Calibration for Multimodal Theory of Mind
    USarXiv cs.AI

    arXiv:2610.36824v1 Announce Type: new Abstract: Theory of Mind (ToM), the ability to infer others' beliefs, goals, and intentions, is a cornerstone of social intelligence. Multimodal ToM requires inferring an agent's mental state from…

  110. Tri-Directional Sub-Voxel Displacement and Ipsilateral Territory Decoupling for 3D Bilateral Coronary Artery Tree Reconstruction in Cardiac CT Angiography
    USarXiv cs.CV

    arXiv:2610.05900v2 Announce Type: replace Abstract: Three-dimensional (3D) reconstruction of bilateral coronary trees (LCA and RCA) from cardiac computed tomography angiography (CCTA) is limited by $64^3$ quantization, territory ambiguity…

  111. Beyond the Readout: A Latent Numerical Frame Shared Across Modality and Operation
    USarXiv cs.AI

    arXiv:2610.37422v1 Announce Type: new Abstract: When a multimodal model computes 17+35 from text or 91-39 from a rendered image, it produces the same answer, 52. Does it use one internal concept of number for both, or separate machinery…

  112. Multimodal Knowledge Aggregation for Microbial Keratitis: Geographic Confounding and Shortcut Risk in Pooled Multicountry Data
    USarXiv cs.CV

    arXiv:2607.03740v3 Announce Type: replace-cross Abstract: To develop an open-source multimodal model for bacterial-versus-fungal microbial keratitis classification and assess shortcuts from geographic pooling. Participants: 1,645 US and…

  113. V-LUMEN: Visual Lookup Memory for Embedding Scaling in Vision-Language Models
    USarXiv cs.CV

    arXiv:2610.37019v1 Announce Type: cross Abstract: Embedding-level scaling expands the capacity of language models through lookup-based memory rather than by increasing model depth, width, or dense computation. Extending this idea to…

  114. Dense Representation Learning and Differentiable Shape Modeling for Spine Landmark Detection
    USarXiv cs.CV

    arXiv:2610.36951v1 Announce Type: new Abstract: Spine landmark detection from spinal X-ray is a key step for estimating spine deformation severity, for building biomechanical models, and for pre-operative planning. We reframe anatomical…

  115. Mining the Nineteenth Century for the Thirteenth: Register-Scale Forced Alignment and HTR Bootstrapping via Legacy Regesta and Antiquarian Editions
    USarXiv cs.CV

    arXiv:2610.36956v1 Announce Type: new Abstract: Handwritten text recognition (HTR) for the registers of Pope Honorius III depends on modern expert transcriptions, which exist only for regnal years 1-2 of Reg. Vat. 9. For the remaining…

  116. Proactive Memory for Event-Driven LLM Agents
    USarXiv cs.AI

    arXiv:2610.36967v1 Announce Type: new Abstract: Proactive LLM agents anticipate service opportunities from ongoing observations and accumulated history. Existing approaches emphasize recent context or retrieval of related memories…

  117. Scoring-Rule Sensitivity in LLM Evaluation: Evidence from Indian Financial Regulatory Text
    USarXiv cs.CL

    arXiv:2604.19298v3 Announce Type: replace Abstract: A benchmark leaderboard reports one accuracy number, but that number is a joint product of a model and a scoring rule. We ask what changes when only the rule changes, and find that model…

  118. SeeNav-Agent: Enhancing Vision-Language Navigation with Visual Prompt and Step-Level Policy Optimization
    USarXiv cs.LG

    arXiv:2512.02631v2 Announce Type: replace Abstract: Existing Vision-Language Navigation (VLN) agents based on Large Vision-Language Models (LVLMs) often suffer from errors of perception, reasoning, and planning, which significantly hinder…

  119. When the Critic Falls Behind: Co-Evolving Rubrics and Harnesses for Continual Agent Self-Improvement
    USarXiv cs.AI

    arXiv:2610.36826v1 Announce Type: new Abstract: Autonomous agents increasingly depend on harnesses that govern interaction, memory, tools, and execution around large language models (LLMs). While recent work has begun to evolve these…

  120. Abstraction in Style: Beyond Texture and Color
    USarXiv cs.CV

    arXiv:2603.29924v3 Announce Type: replace Abstract: Artistic styles often embed abstraction beyond surface appearance, involving deliberate reinterpretation of structure rather than mere changes in texture or color. Conventional style…

  121. Curriculum-Driven Deep Reinforcement Learning for Autonomous Far-Range Rocket Landing
    USarXiv cs.LG

    arXiv:2610.35954v1 Announce Type: new Abstract: Autonomous vertical landing of reusable launch vehicles is a critical yet formidable challenge in aerospace engineering, primarily due to highly nonlinear dynamics, stringent precision…

  122. PRIMA: Boosting Animal Mesh Recovery with Taxonomic Priors and Adaptation-Driven Data Bootstrapping
    USarXiv cs.CV

    arXiv:2606.02366v2 Announce Type: replace Abstract: Recovering articulated 3D animal meshes from monocular images is challenging due to the long-tailed distribution of species and limited large-scale 3D supervision. Existing methods rely…

  123. FedHQ: Hybrid Runtime Quantization for Federated Learning
    USarXiv cs.LG

    arXiv:2505.11982v2 Announce Type: replace Abstract: Federated Learning (FL) must balance training efficiency and model accuracy across heterogeneous clients. Post-training quantization (PTQ) avoids the extra computation of…

  124. What Makes Video World Model Latents Action-Relevant: Prediction over Reconstruction
    USarXiv cs.CV

    arXiv:2606.07687v2 Announce Type: replace Abstract: Video world models are increasingly used to provide predictive visual representations, yet it remains unclear which pretraining signals induce action-relevant structure in their latent…

  125. HoloRecon: Incremental Reconstruction of Long-Range Panoramas with Global Consistency
    USarXiv cs.CV

    arXiv:2610.36876v1 Announce Type: new Abstract: Long-sequence panoramic reconstruction requires detailed scene geometry and globally consistent camera poses across distant revisits. Panoramic SLAM and incremental SfM rely on feature…

  126. GCSR-OCC: Geometry Calibration and Semantic Routing for Open-Vocabulary Occupancy Prediction
    USarXiv cs.CV

    arXiv:2610.37149v1 Announce Type: new Abstract: Open-vocabulary 3D occupancy prediction represents scene geometry and semantics in a unified voxel grid, enabling recognition beyond predefined categories. One approach reconstructs scenes…

  127. What Should an Information Bottleneck Encoder Learn? Separating Representation Learning from Rate Control
    USarXiv cs.LG

    arXiv:2610.36255v1 Announce Type: new Abstract: The information bottleneck (IB) trades compression of an input $X$ against prediction of a target $Y$ through a multiplier $\beta=1/\lambda$, yet each operating point is usually served by…

  128. Confidence Is Not a Promise: A Reference Audit of Typed Decision Models
    USarXiv cs.LG

    arXiv:2610.36311v1 Announce Type: new Abstract: Typed decision models return a probability distribution over options supplied by the caller, and software tends to act on that number as if it were a guarantee. We measure what it licenses…

  129. Agent-Conditioned Routing: Diverse Reasoning Paths Within One Frozen Model
    USarXiv cs.LG

    arXiv:2610.37375v1 Announce Type: cross Abstract: Generating several solutions can expose alternative methods, but additional trajectories often paraphrase one dominant approach. We investigate whether the conditional capacity of a frozen…

  130. Anomaly Detection via Autoencoder Composite Features and NCE
    USarXiv cs.LG

    arXiv:2502.01920v3 Announce Type: replace Abstract: Unsupervised anomaly detection is a challenging task. Autoencoders (AEs) or generative models are often employed to model the data distribution of normal inputs and subsequently identify…

  131. Locally invariant manifolds and foliations
    USarXiv cs.LG

    arXiv:2604.18202v2 Announce Type: replace-cross Abstract: Invariant manifolds and foliations associated to invariant pseudo-hyperbolic reference geometries have been well-studied in the context of dynamics implemented by a diffeomorphism…

  132. JevRec: Linguistic Encoding with Jev-like Decision Modeling for LLM-based Generative Recommendation
    USarXiv cs.CL

    arXiv:2610.36937v1 Announce Type: new Abstract: Generative recommendation (GR) represents items as semantic identifiers (SIDs) and autoregressively generates the next item's SID from interaction history. However, most methods assign the…

  133. DeGenTWeb: A First Look at MGT-dominant Websites
    USarXiv cs.LG

    arXiv:2605.00087v2 Announce Type: replace-cross Abstract: Many news reports have claimed that machine-generated text (MGT) from large language models (LLMs) is taking over the web. However, these claims are typically not based on a…

  134. Learning from Points and Patches: A Pre-trained Hybrid Attention Transformer for Time Series Analysis
    USarXiv cs.LG

    arXiv:2610.36572v1 Announce Type: new Abstract: Time-series data generated by Web services and connected systems support applications such as user activity analysis, service monitoring, and resource management. However, learning…

  135. Latent Visual Contrastive Reasoning with Multi-Branch Exploration
    USarXiv cs.CV

    arXiv:2610.36809v1 Announce Type: new Abstract: Multimodal large language models have made substantial progress in visual question answering, but complex visual reasoning often requires locating and comparing multiple pieces of…

  136. When Retrieval Is Not Enough: Obsolete Facts in Agent Memory
    USarXiv cs.AI

    arXiv:2610.36308v1 Announce Type: new Abstract: Memory-agent benchmarks often collapse retrieval and answer generation into one score, making it hard to tell whether a failure comes from missing an update or from not using an update that…

  137. ETAMem: Decoupling and Aligning Heterogeneous Memory for Long-Term Conversational Agents
    USarXiv cs.CL

    arXiv:2610.37443v1 Announce Type: new Abstract: Long-term memory enables Large Language Models (LLMs) to maintain coherent and personalized interactions across extended conversations. However, existing memory systems face two key…

  138. A Tight 3-Approximation for Fair $k$-Means
    USarXiv cs.LG

    arXiv:2609.07974v2 Announce Type: replace-cross Abstract: Fairness in clustering has attracted sustained research interest, motivated by the need to ensure equitable representation of protected groups in machine learning applications. We…

  139. Ensemble-size-dependence of deep-learning post-processing methods that minimize an (un)fair score: motivating examples and a proof-of-concept solution
    USarXiv cs.LG

    arXiv:2602.15830v2 Announce Type: replace-cross Abstract: Fair scores reward ensemble forecast members that behave like samples from the same distribution as the verifying observations. They are therefore an attractive choice as loss…

  140. Single-Step Bidirectional Unpaired Image Translation Using Implicit Bridge Consistency Distillation
    USarXiv cs.CV

    arXiv:2503.15056v2 Announce Type: replace Abstract: Unpaired image-to-image translation has seen significant progress since the introduction of CycleGAN. However, methods based on diffusion models or Schr\"odinger bridges have yet to be…

  141. Optimal Experiments for Partial Causal Effect Identification
    USarXiv cs.AI

    arXiv:2605.06993v3 Announce Type: replace Abstract: Causal queries are often only partially identifiable from observational data, and experiments that could tighten the resulting bounds are typically costly. We study the problem of…

  142. HateSense: Hierarchical Sentiment Analysis Framework for Hate Speech Detection: Implementing Binary and Multiclass Classification Strategy
    USarXiv cs.CL

    arXiv:2411.05819v2 Announce Type: replace Abstract: Hate speech detectors must separate hateful posts from the far more common offensive but non-hateful ones, and two popular ideas for doing so are hierarchical (coarse-to-fine)…

  143. QESChunker: A Single Objective Unifies Overlapping and Non-Overlapping Chunking for RAG
    USarXiv cs.CL

    arXiv:2610.36557v1 Announce Type: cross Abstract: Chunking is a fundamental preprocessing step in retrieval-augmented generation (RAG), yet most existing chunkers rely primarily on heuristics. We introduce the Question--Evidence-Supervised…

  144. EffiBench-X: A Multi-Language Benchmark for Measuring Efficiency of LLM-Generated Code
    USarXiv cs.CL

    arXiv:2505.13004v2 Announce Type: replace Abstract: Existing code generation benchmarks primarily evaluate functional correctness, with limited focus on code efficiency and often restricted to a single language like Python. To address this…

  145. SpeeDiff: Scalable Pixel-Anchored End-to-End Latent Diffusion Model
    USarXiv cs.CV

    arXiv:2610.36419v1 Announce Type: new Abstract: We present Scalable Pixel-anchored End-to-end Diffusion (SpeeDiff), a latent diffusion method that jointly trains the VAE and the diffusion model from scratch. In principle, joint training…

  146. PrinGS: Privacy Preservation in Gaussian Splatting
    USarXiv cs.CV

    arXiv:2610.36168v1 Announce Type: new Abstract: 3D Gaussian Splatting (3DGS) has rapidly become the preferred representation for high-fidelity novel view synthesis and 3D scene reconstruction. However, the learned representation encodes…

  147. Visual Autoregressive Generation with Next Token Imagination
    USarXiv cs.CV

    arXiv:2610.36793v1 Announce Type: new Abstract: The space of possible visual content is enormous, and arguably unbounded. Any training set, however large, therefore covers only a tiny fraction of it. Yet generative models are typically…

  148. Reach-Avoid Differential game with Reachability Analysis for UAVs: A decomposition approach
    USarXiv cs.AI

    arXiv:2512.22793v2 Announce Type: replace-cross Abstract: Reach-avoid (RA) games have significant applications in security and defense, particularly for unmanned aerial vehicles (UAVs). These problems are inherently challenging due to the…

  149. LingGS: Decoupling Feed-Forward Geometry Prediction from Camera Tracking for High-Fidelity Gaussian SLAM
    USarXiv cs.CV

    arXiv:2610.36853v1 Announce Type: new Abstract: Achieving high-quality reconstruction and responsive camera tracking remains challenging for monocular RGB Gaussian splatting SLAM under a limited optimization budget. We present an online…

  150. Streaming Communication in Multi-Agent Reasoning
    USarXiv cs.CL

    arXiv:2606.05158v3 Announce Type: replace Abstract: Multi-agent reasoning systems adopt a generate-then-transfer paradigm that forces end-to-end latency to scale linearly with pipeline depth. We introduce StreamMA, a multi-agent reasoning…