Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Modelswith Yecheng Wu, Song Han1 versionW_g85rufqhv1 · current