arXiv 2503.080262025-07-28cs.CLcs.AI

展望与回顾：面向长期个性化对话智能体的反思式记忆管理

In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents

Arizona State University（亚利桑那州立大学）；Google Cloud AI Research（谷歌云AI研究院）；Google Cloud AI（谷歌云AI）；UNC Chapel Hill（北卡罗来纳大学教堂山分校）

Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long T. Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Iyer, Tianlong Chen, Huan Liu, Chen… 展开作者

Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long T. Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, Tomas Pfister

备注：: Accepted to ACL 2025

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对长期个性化对话中的记忆粒度僵化与检索固定问题，提出 RMM，通过多粒度前瞻记忆构建和基于引用证据的在线 RL 回顾式检索优化，在 LongMemEval 等基准上显著提升性能。

摘要：

大语言模型（LLM）在开放式对话中取得了显著进展，但其无法从长期交互中保留并检索相关信息，限制了它们在需要持续个性化的应用中的效果。已有研究提出外部记忆机制来解决这一局限，使 LLM 能够维持对话连续性。然而，现有方法面临两个关键挑战：第一，僵化的记忆粒度无法捕捉对话的自然语义结构，导致表征碎片化且不完整；第二，固定的检索机制无法适应多样的对话上下文和用户交互模式。本文提出反思式记忆管理（Reflective Memory Management，RMM），这是一种面向长期对话智能体的新型机制，融合前瞻与回顾两类反思：（1）Prospective Reflection（前瞻性反思）跨话语、轮次和会话等粒度动态总结交互，将其写入个性化记忆库，以便未来有效检索；（2）Retrospective Reflection（回顾性反思）基于 LLM 引用的证据，以在线强化学习（RL）方式迭代优化检索。实验表明，RMM 在多项指标和基准上均取得持续提升；例如，在 LongMemEval 数据集上，RMM 相比无记忆管理基线的准确率提升超过 10%。

英文摘要：

Large Language Models (LLMs) have made significant progress in open-ended dialogue, yet their inability to retain and retrieve relevant information from long-term interactions limits their effectiveness in applications requiring sustained personalization. External memory mechanisms have been proposed to address this limitation, enabling LLMs to maintain conversational continuity. However, existing approaches struggle with two key challenges. First, rigid memory granularity fails to capture the natural semantic structure of conversations, leading to fragmented and incomplete representations. Second, fixed retrieval mechanisms cannot adapt to diverse dialogue contexts and user interaction patterns. In this work, we propose Reflective Memory Management (RMM), a novel mechanism for long-term dialogue agents, integrating forward- and backward-looking reflections: (1) Prospective Reflection, which dynamically summarizes interactions across granularities-utterances, turns, and sessions-into a personalized memory bank for effective future retrieval, and (2) Retrospective Reflection, which iteratively refines the retrieval in an online reinforcement learning (RL) manner based on LLMs' cited evidence. Experiments show that RMM demonstrates consistent improvement across various metrics and benchmarks. For example, RMM shows more than 10% accuracy improvement over the baseline without memory management on the LongMemEval dataset.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。