arXiv / Working Memory / Token-level

MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning

作者：Qianhao Yuan, Jie Lou, Zichao Li, Jiawei Chen, Yaojie Lu, Hongyu Lin, Le Sun, Debing Zhang, Xianpei Han · 单位：Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China · 来源：arXiv · 发布日期：2025/11

工作记忆词元级记忆
新标签打开 AlphaXiv
MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning figure
AlphaXiv 中文概览（可滚动查看）