arXiv preprint / 奖励驱动进化

Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning

作者：Kongcheng Zhang, Qi Yao, Shunyu Liu, Yingjie Wang, Baisheng Lai, Jieping Ye, Mingli Song, Dacheng Tao · 单位：Zhejiang University, Alibaba Cloud Computing, Nanyang Technological University · 会议/期刊：arXiv preprint · 方向：奖励驱动进化 · 发布日期：2025-06-10

自我进化智能体奖励驱动进化大语言模型智能体奖励学习强化学习
新标签打开 AlphaXiv
Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning figure
AlphaXiv 中文概览（可滚动查看）