01. Harness-Zero: Harness Distillation via Agent-as-Harness

86% worth reading · novelty 2.0/3 · new method · cs.AI

02. Critical-State RL: Diagnosing Trainable States for Multi-Turn Tool Use

85% worth reading · novelty 1.9/3 · new method · cs.CL

03. Et Tu, Brute? Economic Misalignment in Personal AI Agents

82% worth reading · novelty 2.0/3 · benchmark/eval · cs.AI

04. OSWorld-Pro: Process-based Evaluation for Computer Use Agents

85% worth reading · novelty 1.6/3 · benchmark/eval · cs.AI

05. Emergent Collusion in Long-Horizon LLM Agent Interaction

82% worth reading · novelty 1.7/3 · benchmark/eval · cs.AI

06. SocioVerse2: A Longitudinal Dynamic Social Simulation Framework under a Human-AI Co-evolutionary Paradigm

78% worth reading · novelty 1.8/3 · applied system · cs.CL

07. RRSI: Regularized Recursive Self-Improvement of Agent Harnesses

84% worth reading · novelty 1.4/3 · new method · cs.AI

08. GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay

77% worth reading · novelty 1.8/3 · benchmark/eval · cs.AI

09. DolphinBench: Mapping the Pareto Frontier of Agent Memory

79% worth reading · novelty 1.6/3 · benchmark/eval · cs.AI

10. onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction

80% worth reading · novelty 1.6/3 · applied system · cs.CL

11. When Tomorrow Becomes Today: Self-Evolving Policies for Agentic Time-Series Forecasting

75% worth reading · novelty 1.9/3 · new method · cs.LG

12. Rare Event Estimation via Iterative Unalignment

74% worth reading · novelty 1.9/3 · new method · cs.AI

13. BackTrend: Evaluating Scientific Weak-Signal Prediction via Backward Reconstruction

72% worth reading · novelty 1.9/3 · benchmark/eval · cs.AI

14. The Copy Ceiling: An Input-Exposure Control for Ontology-Grounded Generation over Curated Corpora

71% worth reading · novelty 1.9/3 · benchmark/eval · cs.CL

15. Human-LLM Deliberation as Interactive Proof: Conditions for Verifiability Without Transparency

60% worth reading · novelty 2.0/3 · new method · cs.CL

16. Decomposing Error and Style in Automated Clinical Coding

56% worth reading · novelty 2.1/3 · new method · cs.CL

17. Generative Tutorial: Towards Live Contextualized Visual Instructions for Physical Tasks

54% worth reading · novelty 1.9/3 · applied system · cs.AI

18. Exactness at Inference: A Representational Criterion for Out-of-Distribution Generalization

51% worth reading · novelty 2.1/3 · survey/position · cs.AI

19. Jev for Scientific Decisions: Evaluating Semantic Choices and Their Consequences

70% worth reading · novelty 0.7/3 · benchmark/eval · cs.AI

20. LoRA-generating hypernetworks for efficient on-device LLM generative personalization

38% worth reading · novelty 1.9/3 · new method · cs.LG