RP-OPSD: Reasoning-Pivot-Guided On-Policy Self-Distillation for Multilingual Reasoning Transfer

Part of

Advanced Inverse Reward And Prompt Engineering