# On-Policy Learning

Co-Evolving Harnesses and Models: On-Policy Correction Helps Weaker Models Catch Up Where Imitation Fails

Harness 为弱模型量身演化后，直接模仿强模型的完整轨迹会破坏二者的配合。本文沿着两条真实失败轨迹，拆解 on-policy correction 为什么只改学生自己的第一个错误 turn。