ReNIO: Reweighting Negative Trajectory Importance for LLM On-Policy Distillation

Skill0.5: Joint Skill Internalization and Utilization for Out-of-Distribution Generalization in Agentic Reinforcement Learning