01. Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models

85% worth reading · novelty 1.9/3 · benchmark/eval · cs.AI

02. SciExam for ENSO: Can AI Agents Build Climate Models?

83% worth reading · novelty 2.0/3 · benchmark/eval · cs.AI

03. RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing

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

04. CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution

84% worth reading · novelty 1.8/3 · new method · cs.CL

05. RunningTab: Direct Workspace Interaction with Environment-Side Tabs

85% worth reading · novelty 1.6/3 · new method · cs.CL

06. Know the Shape, Find the Fault: Topology-Conditioned Diagnosis of Multi-Agent LLM Failures

82% worth reading · novelty 1.8/3 · new method · cs.MA

07. Recursive Game Creator: An Agentic Product-Level Experience-Oriented Game Harness

82% worth reading · novelty 1.6/3 · new method · cs.MA

08. A Good Self-Teacher Meets the Student Where They Are: Joint On-Policy Learning and Teaching

79% worth reading · novelty 1.8/3 · new method · cs.AI

09. A Society of Researchers: Designing Institutions for Populations of Autonomous Research Agents

73% worth reading · novelty 2.1/3 · survey/position · cs.AI

10. Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models

77% worth reading · novelty 1.8/3 · new method · cs.CL

11. EmbodiedRSI: Active Continual Robot Learning Through Hypothesis-Guided Co-Evolution

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

12. Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station

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

13. CANDO: Cooperative Agentic Network for Layout Design Optimization

77% worth reading · novelty 1.7/3 · new method · cs.MA

14. When the Governor Becomes the Disturbance: Control-Generated Disturbance and Cost-Aware Backoff in Governed Tool-Using Agents

77% worth reading · novelty 1.6/3 · new method · cs.MA

15. AGAR: a reinforcement learning substrate for LLM program evolution

73% worth reading · novelty 1.9/3 · new method · cs.MA

16. A Case Study in Assuring AI-Written Software

78% worth reading · novelty 1.5/3 · applied system · cs.MA

17. Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision

83% worth reading · novelty 1.0/3 · new method · cs.CL

18. How assigned AI use before class shapes active student engagement in class

66% worth reading · novelty 1.5/3 · applied system · cs.AI

19. Never Look Back: Understanding Persistence in 3D Object Memory from Egocentric Videos

63% worth reading · novelty 1.5/3 · new method · cs.AI

20. FoldBack: Self-Correcting Masked Generative Policy for Long-Horizon Garment Folding

56% worth reading · novelty 1.9/3 · new method · cs.AI