MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent
Tsinghua University · University of the Chinese Academy of Sciences · Shanghai AI Lab · ByteDance Seed · Tsinghua University ByteDance · AIR, Tsinghua University · Microsoft · , Chinese Academy of Sciences · Tsinghua University, Tsinghua University
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
Despite improvements by length extrapolation, efficient attention and memory modules, handling infinitely long documents without performance degradation during extrapolation remains the ultimate challenge in long-text processing. To solve this problem, We introduce a novel agent workflow, \method, which processes text in segments and updates memory through an overwrite strategy, addressing the challenge of long-context task through enhanced memory management. We further extend the DAPO algorithm to directly optimize memory ability in an end-to-end fashion, facilitating training via independent-context multi-conversation generation. Experimental results demonstrate that MemAgent has superb long-context capabilities, being able to extrapolate from an 8K context to a 3.5M QA task with a performance loss of less than 10\% and achieving over 95\% on the 512K NIAH test.