01. MolLean: A Process-Verifiable Language for Chemical Reasoning

86% worth reading · novelty 1.9/3 · new method · cs.CL

02. Same Claim, Different Interpretations: How LLM Agents Resolve Ambiguity in Scientific Claims

86% worth reading · novelty 1.8/3 · new method · cs.CL

03. ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps

81% worth reading · novelty 2.0/3 · new method · cs.LG

04. One Task, Many Interfaces: Simple Cross-Interface Consistency Training Improves LLM Agent Robustness

85% worth reading · novelty 1.7/3 · new method · cs.AI

05. PhySGA: A Physics-Space Grounded Agentic Framework for Generalizable and Interactive 3D Composition

82% worth reading · novelty 1.9/3 · new method · cs.AI

06. Co-Evolving Acting and Forecasting for Language Agents

83% worth reading · novelty 1.8/3 · new method · cs.AI

07. Scaling Discovery Requires Forgetting

80% worth reading · novelty 1.9/3 · new method · cs.AI

08. Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization

81% worth reading · novelty 1.7/3 · new method · cs.AI

09. Player-1: Learning to Play from Synthetic Experience

76% worth reading · novelty 1.6/3 · new method · cs.AI

10. Price of Anarchy of First-Price Auctions in Agentic Markets

73% worth reading · novelty 1.7/3 · survey/position · cs.AI

11. AdaRSI: On-Demand Recursive Self-Improvement for Embodied Agents

72% worth reading · novelty 1.7/3 · new method · cs.CL

12. PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis

64% worth reading · novelty 1.9/3 · new method · cs.AI

13. PerSo: Investigating Facets of LLMs' Social Intelligence in Real-World Persuasion Scenarios

63% worth reading · novelty 1.5/3 · benchmark/eval · cs.AI

14. Recovering Reliable Supervision via Critique-Guided On-Policy Distillation

59% worth reading · novelty 1.6/3 · new method · cs.AI

15. Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR

47% worth reading · novelty 1.8/3 · benchmark/eval · cs.CL

16. Latent Compaction with Test-Time Training

43% worth reading · novelty 1.9/3 · new method · cs.CL

17. ReaDy: Recurrent States as Dynamic Context for Speculative Drafting in Hybrid Linear Attention Large Language Models

44% worth reading · novelty 1.8/3 · new method · cs.CL

18. Lost in Translation: Why Lower Bias Scores Don't Mean Fairer Models

38% worth reading · novelty 1.9/3 · benchmark/eval · cs.CL

19. Your Transformer Is Secretly Elastic

39% worth reading · novelty 1.8/3 · new method · cs.AI

20. VASPO: Self-Distillation with Privileged Value Estimation

42% worth reading · novelty 1.4/3 · new method · cs.CL