AI News

Recursive Reward Improvement via Adversarial Meta-Prompt Evolution (arxiv.org)

Learning When to Withdraw: Curriculum Coverage and Rule-Specific Failures in Fine-Tuned Language Agents (arxiv.org)

Athanor: An Agent-Native System for Autonomous Post-Training Research (arxiv.org)

OmicsBench: a long-horizon multi-turn benchmark of AI agents on research-study-scale bioinformatics (arxiv.org)

Sci$\Pi$vot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents (arxiv.org)

Honesty is the Best Policy ... or is it? Demand Management for Probabilistic Service Options (arxiv.org)

Same Target, Different Origins: Aligning Latent Rollout for Consistent PDE Emulation (arxiv.org)

Memorilla: Latent Semantic Memory for LLMs (arxiv.org)

Does Compression Imply Generalization? A Minimum Description Length Perspective (arxiv.org)

SHRED: Self-Play for Heterogeneous Road-Users without Expert Demonstrations (arxiv.org)

Clinically Interpretable Biological Age Estimation Using CatBoost and SHAP with Continuous and Categorical Features from NHANES (arxiv.org)

CogDrift-4D: Hierarchical Multimodal Modeling of Temporal Behavioral Decay for MCI Detection (arxiv.org)

Do Emergent Misalignment Models Share a Circuit? (arxiv.org)

Wrong Votes, Right Gradients: Why Majority-Vote Rewards Track Verifiable RL (arxiv.org)

Sign Betting for Misspecified Regression: Variance Adaptivity and Optimal Approximation (arxiv.org)

ABNAT: Attention-Based Noise-Aware Training for Robust Transformers on Analog Compute-in-Memory Systems (arxiv.org)

Probe-and-Resume: When and How to Evaluate Agent Execution (arxiv.org)

MemComposer: Learning to Assemble Memories for LLM Agents via Reinforcement Learning (arxiv.org)

Planning Commitment of LLM under Partial Observability (arxiv.org)

Kolab: Benchmarking Human-Agent Collaboration in Long-Horizon Workflows (arxiv.org)

CleanRM: Reward Modeling from Noisy Feedback with a Few Clean Feedback for RLHF (arxiv.org)

Task-Centric Multimodal Learning for Cognitive Impairment Detection (arxiv.org)

Replicable Dueling Bandits (arxiv.org)

Recursive Offline Distillation via Entropy Gain Sampling and Answer Mixing (arxiv.org)

How to Train Your Little Dragon: Teaching a Small Language Model to Construct and Refine GPU Optimization Scripts (arxiv.org)

Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models (arxiv.org)

All-in-One Memory: Process-Rewarded Reinforcement Learning for Memory Agents (arxiv.org)

GRADS: Graph-Regulated Agentic Data Synthesis with Controllable Difficulty (arxiv.org)

MemScope: Evolving Agent Memory Architectures through Localized Code Repair (arxiv.org)

AgentEval: Evidence-Bounded Assessment of Stateful Agents (arxiv.org)

Decoding Naturalistic Speech from fMRI with Temporal Modeling and Multimodal Alignment (arxiv.org)

Does Latent Chain-of-Thought Actually Reason? A Single, Unifying Taxonomy for Latent Reasoning (arxiv.org)

Mine, Sample, Render: Test-Time Diversity Alignment Of Generative Models (arxiv.org)

Condition-Contrastive Extrapolation for On-Policy Distillation of Language Models (arxiv.org)

Who May Release an Agent's Promise? Standing over Commitments in Multi-Party Chat (arxiv.org)

Entertain, then Cancel: How Thinking Makes Reasoning Models Read Some and Or Literally (arxiv.org)

Imperfect Is Perfect: Learned Column Prefilling for Branch-Price-and-Cut (arxiv.org)

CausalEvo: Causal Skill Self-Evolution for Health LLM Agents (arxiv.org)

INCONVENIENT: A Benchmark to Assess LLM Generalization Capabilities from Convenience Samples (arxiv.org)

Poisoning the corpora of protein design systems during pre-training and retrieval (arxiv.org)

Rubric Judgments Lack Uncertainty: Conformal Weighting for LLM Evaluation (arxiv.org)

AutoOnboarder: Autonomously Onboarding LLM Agents (arxiv.org)

Auditing Escalation Reliability in Healthcare LLM Agents: A Perturbation-Controlled Benchmark of Workflow-State Failures (arxiv.org)

Diviner: Agentic RCA with Human-Governed Adaptation for Large-Scale Infrastructure (arxiv.org)

Mallea: Building, Steering, and Memorizing Worlds with Coding Agent (arxiv.org)

Hard Way, The Only Way: Provable Merits of Training on Hardest Examples (arxiv.org)

Learning and Forecasting Multiscale Spatiotemporal Epidemic Dynamics with Lagrangian-Eulerian Graph Neural ODEs (arxiv.org)

Not All Change Is Damage: Update-Matched Unlearning for Large Language Models (arxiv.org)

Compatibility Matters: A Framework for Token-Level Teacher Selection in On-Policy Distillation (arxiv.org)

Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL (arxiv.org)