AI News

Kalshi Hits $15.66B as Combination Bets Lift Weekly Trading (readwrite.com)

Official MCP Python SDK Flaw Can Let Malicious Servers Steal OAuth Credentials (thehackernews.com)

MagPro Launches AeroLev 1850 for Water-Efficient AI Data Center Cooling - KuCoin (www.kucoin.com)

New coalition warns data center growth could strain Texas water supplies - KXAN Austin (www.kxan.com)

Your phone is AI’s newest hardware (stackoverflow.blog)

Incredible (www.producthunt.com)

OpenAI Shelves GPT-6.1 Astra After Tests Find Deception and Unauthorized Actions (thehackernews.com)

Veille quotidienne — 29/09/2026

Alibaba Qwen Releases Qwen-Audio-3.1-Realtime: A Full-Duplex Voice Model Trained to Think, Act, and Decide When to Speak (www.marktechpost.com)

OpenAI Pauses Tool Use After Agent Bypasses Internet Controls to Reach External Chatbot (thehackernews.com)

Anthropic Releases Claude Sonnet 5.5: 70.6% on Terminal-Bench 4.0 at the Same $2/$10 Price (www.marktechpost.com)

Pugnac. Un nouveau soutien financier public pour les projets innovants de Symbiose (www.hautegironde.fr)

Stable initialization without the CLT (arxiv.org)

Low-Cost Black-Box Detection of LLM Hallucinations via Dynamical System Prediction (arxiv.org)

Energy-efficient operation of neural operators for virtual sensing (arxiv.org)

ChemMLLM: Chemical Multimodal Large Language Model (arxiv.org)

Generalization behavior of OPTQ and the role of regularization (arxiv.org)

The KV Cache Is the New Memory Wall (arxiv.org)

AC Power Flow Contingency Analysis Using a Single Deep Neural Network (arxiv.org)

Tight Stochastic Condition-Number Dependence in Nonconvex-Strongly-Concave Minimax Optimization (arxiv.org)

Online Learning via Learned Latent Bayesian Tracking (arxiv.org)

BeatGraph: Self-Supervised Heartbeat Graphs for Infant ECG Representations from the Home Environment (arxiv.org)

NEXT: Physics-Informed Neuro-Spectral Exponential Time Differencing Architectures (arxiv.org)

HySTAR: Anchored Hypergraphs for Stable Credit Assignment in Cooperative Multi-Agent Reinforcement Learning (arxiv.org)

Scaffold: Support Graph Theory Based Sparsification for Graph Neural Networks (arxiv.org)

Retraction-Based Gradient Projection Algorithms on Manifolds (arxiv.org)

Differential Attention Unlocks Complementary EEG and Speech Fusion for Emotion Recognition (arxiv.org)

Towards Understanding LLM-Based Log Anomaly Detection: An Empirical Study of Performance, Efficiency, and Robustness (arxiv.org)

Progressive Memory Transformer: Memory-Aware Attention for Time-Series (arxiv.org)

Bridging Body and Brain: Gene-Driven Morphology--Control Co-Design (arxiv.org)

More Sensors Only One Field: Rethinking Continual Spatio-Temporal Forecasting (arxiv.org)

Benchmarking Attention for Tabular Foundation Models (arxiv.org)

Deterministic Regime Switching and Feasibility Inversion in Dynamic Tensor Rematerialization (arxiv.org)

Budgeted Quotient-Residual Guidance for Frozen Pocket-Conditioned Molecular Diffusion (arxiv.org)

Self-Supervised Representation Learning: From Spectral Foundation Models to Auroral Emission Spectra (arxiv.org)

Synth-JEPA: Joint Embedding Prediction for Renderer-Free Synthesizer Parameter Search (arxiv.org)

Audio emotion recognition for atypical hearing (arxiv.org)

WorldTS: World Modeling for Multimodal Covariate-aware Time Series Forecasting (arxiv.org)

I Act Therefore I Am: When Is JEPA's Action-Conditioning Enough to Learn Causal Mechanisms? (arxiv.org)

Bayesian Tensor Autoencoder with Physics-informed Predictive Prior for Multi-dimensional Time Series Anomaly Detection (arxiv.org)

CRNDiff: Count-Native Diffusion Framework via Chemical Reaction Networks (arxiv.org)

Precision at Speed: Sample-Efficient Online Model-Based Reinforcement Learning for Hydraulic Excavator Control (arxiv.org)

KuaFu: Compressing Long User Behavior into Understanding at Billion Scale (arxiv.org)

Frame the adversary: a structure-aware attack methodology (arxiv.org)

HybridInfer: Thermal-Aware Reinforcement-Learning Tier Routing for On-Device, Edge, and Cloud LLM Inference (arxiv.org)

Reinforcement Learning of Communication in a Mesh of Small Language Models (arxiv.org)

Dynamic Regret in Online Convex Optimization with Indicator Switching Costs (arxiv.org)

LocUS: Head Selection and Subspace Projection for Targeted Activation Steering (arxiv.org)

Unknown-Traffic Detection, Calibration and Shortcut Reliance in Distilled Encrypted-Traffic Classifiers over One Year (arxiv.org)

GyroNovo: Error-Guided Fragment Imputation with Mass-Aware Attention for \textit{De Novo} Peptide Sequencing (arxiv.org)