AI News

Get a GRIP, this will be a long TRIP: A Quantifiable Long-Range Framework for Verifying Over-squashing (arxiv.org)

Cephalonauts One: A deep fMRI dataset for decoding naturalistic speech in the human brain (arxiv.org)

A Path Integral Surrogate for Multi-Step Gradient Inversion in Federated Learning (arxiv.org)

Mastering Atari 2600 Games with Discovered Options (arxiv.org)

DyadMem: A Long-Term Memory Benchmark of How Agents Work with Users (arxiv.org)

Verifiable, Articulable, and Tacit Components of Preference (arxiv.org)

SoftGene: Protein Language Model-Enhanced Soft Prompting for Interpretable Gene Set Annotation (arxiv.org)

When Numbers Start Talking: Numerical Signalling and Strategic Behaviour Among LLMs (arxiv.org)

UniIntervene++: An Adaptive Intervention Agent for Efficient Real-World Reinforcement Learning (arxiv.org)

Confidence-Gated Cloud-Edge Cascade Triage via Variational Risk Minimization for Medical Imaging (arxiv.org)

IDRF: Inverse-Distilled Reward Fine-tuning of Masked Discrete Diffusion Models (arxiv.org)

On the Convergence of Success Conditioning for Policy Optimization (arxiv.org)

Forecasting from Counterfactual Simulator Rollouts: A Sim2Real Evaluation (arxiv.org)

hacktrace: behavior-supervised detection of reward hacking during code generation (arxiv.org)

MOF-VERIFY: A Failure-Aware Agentic Harness for MOF Hypothesis Verification (arxiv.org)

RIFAR: Reliability and Forgetting-Aware Replay for Continual Robot Learning (arxiv.org)

Peer Influence across Heterogeneous AI Models (arxiv.org)

Flow Matching for Fast Posterior Sampling in Bayesian Inverse Problems (arxiv.org)

Predictor-Guided Latent Space Codon Optimization for Maximizing Protein Expression (arxiv.org)

Trading Strategy Optimization via Textual Gradient (arxiv.org)

Keeping JEPA World Models Plannable When Little of the Frame Moves (arxiv.org)

Gains and Collapse in On-Policy Distillation:A Reinforcement Learning Perspective (arxiv.org)

KV$^2$: A Self-Refining KV Cache (arxiv.org)

Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Models (arxiv.org)

LESSER: Post-Training Data Selection with Output-Layer Gradients (arxiv.org)

RNADyn: A Benchmark for Generating and Understanding RNA Dynamics (arxiv.org)

What Should World Models Forget? Stratified Retention for Continual Adaptation (arxiv.org)

MiDShip: Multimodal Dataset of Ship Cargo Hold Structures for Engineering Design (arxiv.org)

Toward SLM-based agentic task-tool intent matching (arxiv.org)

Learning a Fact Is Not Learning How to Retrieve It (arxiv.org)

EVOL: Simulator-Guided Evolutionary Expert Synthesis for Deployment-Free Learning Path Recommendation (arxiv.org)

JOVE: Joint Execution and Verification for Resource-Aware LLM Task Graphs (arxiv.org)

RINS: Residual-Image Neural Subspace Solvers for Large Sparse Linear Systems (arxiv.org)

Generalizable single-cell perturbation response prediction using energy-guided flow matching (arxiv.org)

HyperThink: Text-to-Parameter Hypernetworks for Efficient Reasoning (arxiv.org)

Budgeted Cache Repair for Cross-Context KV-Cache Reuse (arxiv.org)

Hardware-Native Joint Sparse-Quantization for Trillion-Scale Mixture-of-Experts (arxiv.org)

S$^{2}$-PINN: Stochastic Separable Physics-Informed Neural Networks (arxiv.org)

Optimal Planning in a Dynamic World (arxiv.org)

Coherence-Driven Belief Formation and Population Dynamics of Contagion in LLM Agents (arxiv.org)

EmTech Future 2026: When AI Meets Everything (www.technologyreview.com)

Intent-Hiding Jailbreaks: An Information-Theoretic Framework for Compositional Attacks (arxiv.org)

SimuVerity: Benchmarking Agents for Engineering-Grade Simulink Model Generation (arxiv.org)

Lightweight, Rubric-Guided Trajectory Evaluation for Production AI Agents (arxiv.org)

Multi-Task Evolution for Zero-Shot Cross-Problem Generalization using LLMs (arxiv.org)

Refinement Buys Intelligibility, Search Buys Identity: What Test-Time Compute Buys in Masked-Diffusion TTS (arxiv.org)

Preserving Mathematical Reasoning in Compressed Diffusion Language Models via Trajectory-Aware Low-Rank Approximation (arxiv.org)

ReFract: Benchmarking Perspective Awareness in Language Model Agents with Text World Models (arxiv.org)

Geometry Meets Physics: Data-Efficient Pre-Training for Unstructured Neural PDE Solvers (arxiv.org)

Multilingual GSM-Symbolic: What determines capability transfer across languages? (arxiv.org)