Submitted by Yecheng Wu 4 Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models Jet-AI 5 2