arXiv · Information Retrieval· Haoxiang Zhang, Qinglin Chen, Hiroaki Hayashi, Zhuofeng Li, Siming Zhang, Jiaxin Zhang, Jixuan Chen, Fang Wu, Pan Lu, Silvio Savarese, Julian McAuley, Chien-Sheng Wu·· 4 hr agoAI score44
Self-Retrospection Distillation：把事后经验转化为事前预见
Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight
AI brief
论文提出 prospective learning 与 Self-Retrospection Distillation（SRD），用事后轨迹监督智能体在交互前的预见预测，把 hindsight 蒸馏为同策略的 foresight，且 foresight 只作训练目标、无需在推理时显式生成。
Source: arXiv · Information Retrieval · arxiv.org