作者位次 #4/7ICLR 2026Poster· Deep Learning / Everything Else
LaSeR: Reinforcement Learning with Last-Token Self-Rewarding
Wenkai Yang · Weijie Liu · Ruobing Xie · Yiju Guo · Lulu Wu · Saiyong Yang · Yankai Lin
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。
Wenkai Yang · Weijie Liu · Ruobing Xie · Yiju Guo · Lulu Wu · Saiyong Yang · Yankai Lin
Yiju Guo · Tianyi Hu · Zexu Sun · Yankai Lin
Reinforcement Learning with Verifiable Rewards (RLVR) has advanced LLM reasoning, but remains constrained by inefficient exploration under limited rollout budgets, leading to low sampling success and unstable training in complex tasks. We find that many exploration failures arise not from problem...
Shengda Fan · Xuyan Ye · Yupeng Huo · Zhi-Yuan Chen · Yiju Guo · Shenzhi Yang · Wenkai Yang · Shuqi Ye · Jingwen Chen · Haotian Chen · Xin Cong · Yankai Lin