arXiv cs.AI / cs.LG / cs.CL·2d agoSelf-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight#reinforcement-learning#rlvr#distillationAI research 2 sources1