作者位次 #5/5NeurIPS 2026Poster· Reinforcement Learning / Deep RL
Distribution-Adaptive Policy Optimization
Yuxiao He · Ziqi Wang · Xingzhou Lou · Xiaoqian Liu · Junge Zhang
强化学习与后训练OpenReview
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Yuxiao He · Ziqi Wang · Xingzhou Lou · Xiaoqian Liu · Junge Zhang
Ziqi Wang · Xingzhou Lou · Meiqi Wu · Zhengqi Wen · Junge Zhang
Group Relative Policy Optimization (GRPO) enhances LLM reasoning but often induces overconfidence, where incorrect responses yield lower perplexity than correct ones, degrading relative calibration as described by the Area Under the Curve (AUC). Existing approaches either yield limited improvemen...