AI News

WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution (arxiv.org)

Two Routes to Interoceptive Prioritisation: Dissociating Likelihood Precision and Preference Reweighting in Active Inference (arxiv.org)

AReaL-RSI: Adaptive Scaling and Coordination of Recursive Self-Improvement (arxiv.org)

Chain of Thought Monitoring Can Fail Even When CoT Is Necessary (arxiv.org)

One-Sided Verification Under Limited Expert Review (arxiv.org)

AM-Bench: Understanding Agents under Task Underspecification across Models and Harnesses (arxiv.org)

LifeSearch: Benchmarking Deep Research Agents for Real-World Life and Productivity (arxiv.org)

CompanionSafe: From Multi-turn Safety Evaluation to Guardrails for AI Companions (arxiv.org)

Fresh at Decision, Stale at Action: Routing New Information to LLM Decisions Already Made (arxiv.org)

Monitoring latent reasoning in looped language models (arxiv.org)

Mitigating Cross-Modal Hallucinations by Correcting Modality Misrouting in Omnimodal LLMs (arxiv.org)

RadixFlow: Filling Execution Bubbles in Prefix-Cached Hybrid LLM Serving (arxiv.org)

Beyond Answer Sharing: TrueMAD for Targeted Peer Engagement in Multi-Agent Debate (arxiv.org)

When and How Order-Flow Interventions Work After Dealers Adapt (arxiv.org)

Back to the Future: Reverse-Consequence Modeling for Long-Horizon Particle Dynamics (arxiv.org)

Neuro-Symbolic Graph Transformers for Hierarchical Intention Recognition from Multimodal Human Activity Sequences (arxiv.org)

QORH: Learning Query-Specific Ordered Rule Hypergraphs for Inductive Knowledge Graph Link Prediction (arxiv.org)

Consider, Then Choose: Order-Invariant Decisions in System One Models (arxiv.org)

Output Discipline Sets the Floor: A Falsification-Guided, Model-Agnostic Agent Protocol for Auditable Hypothesis-Driven Decision Making (arxiv.org)

A dataset of formalised theorem statements from the Annals of Mathematics (arxiv.org)

AssetForge: Asset-Augmented Task Synthesis for Business Automation Agents (arxiv.org)

Choosing LLMs for Strategic Agent Interactions (arxiv.org)

ABot-ER: Unified Embodied Reasoning through Agentic Data-Model Co-Evolution (arxiv.org)

Delegating Earth science model calibration to knowledge-guided LLM agents (arxiv.org)

The Singularity Has a Distribution: Exact Timing Laws and Detection Limits for a Stochastic Intelligence Explosion (arxiv.org)

Thinker-WM: A Unified World Model with Shared Multimodal Representations for Embodied Intelligence (arxiv.org)

Predict Before You Modify: Learning Improvement Dynamics for Self-Evolving Agents (arxiv.org)

Autonomous Science Bench: Evaluating AI Agents for Autonomous Scientific Discovery (arxiv.org)

ReTerm: Recovery Training for Self-Improving Terminal Agents (arxiv.org)

Judges Follow the Reference: The Derivation Gap in Verifying Tool-Call Arguments (arxiv.org)

The Smart Buildings Challenge: Learning Foundation Models for Real-World Optimization at Scale (arxiv.org)

Input-Conditioned Search for Efficient Test-Time Scaling of Recursive Reasoners (arxiv.org)

Rethinking Quality in Large Language Model Routing: Security-Conditioned Model-and-Budget Selection (arxiv.org)

Configuring RAG Systems: What Breaks, Where Originates, and How to Fix (arxiv.org)

Order-Consistent Latent Geometry for Reliable Short-Horizon Planning in World Models (arxiv.org)

Agentic4PDM: Host-Governed Evidence Admissibility for Industrial Fault Diagnosis (arxiv.org)

Beyond the Verdict: Agentic Code Judging with Verifiable Feedback (arxiv.org)

Jeva: System One Models from First Principles (arxiv.org)

BLADE: Auditing Information Use for Failure Attribution in LLM Multi-Agent Systems (arxiv.org)

Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents (arxiv.org)

Cross-Market Meta-Learning for Electricity Price Forecasting (arxiv.org)

Self-Evolving Agent Harness (arxiv.org)

Read the Room Before Your Model Generates (arxiv.org)

Who Improves the Robot? Toward Physical Recursive Self-Improvement (arxiv.org)

MFG: Multi-turn Feedback Guided Molecular Optimization (arxiv.org)

AesCode: Aesthetic Code Generation with Decoupled Cross-Modal Rewards (arxiv.org)

Better Capturing Interactions Between Products in~Retail: Revisited Negative Sampling for~Basket Choice Modeling (arxiv.org)

EgoOmniWorld: Unifying Egocentric Understanding via Ego-Action-World Interactions (arxiv.org)

A Global View of How LLMs Value Human Time (arxiv.org)

The Geometry of Language Model Logits: Generic Structure and Learned Departures (arxiv.org)