01Entry
RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning
Self-retiring on-policy distillation for agentic RL — the distillation schedule adapts and stops when the student stops benefiting, instead of following a fixed plan.