arXiv Artificial Intelligence· Kunwang Zheng, Xinyu Sun, Yu Cao, Shuocheng Li, Lei Cheng, Ying Liu, Qilong Zhang, Jinlong Chen, Hua Yang, Qianglong Chen·· 18 小时前AI 评分22
RELAY：面向评分标准强化学习的自适应交接选择性自蒸馏
RELAY: Selective Self-Distillation with Adaptive Hand-off for Rubric-Based Reinforcement Learning
AI 导读
RELAY 提出一种选择性自蒸馏与自适应交接方法，用于基于评分标准的强化学习：第一阶段用冻结教师模型结合 GRPO 做选择性蒸馏，平台期检测器触发后移除教师，第二阶段仅用 GRPO。
来源：arXiv Artificial Intelligence · arxiv.org