arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2601.01280cs.CLcs.AI
收藏

记忆是否需要图?一种统一框架和长期对话记忆的实证分析

Does Memory Need Graphs? A Unified Framework and Empirical Analysis for Long-Term Dialog Memory

Peking University(北京大学);Georgia Tech(佐治亚理工学院);ZJU(浙江大学);MBZUAI(马克斯·普朗克智能系统研究所);UCAS(中国科学院大学)

Sen Hu, Yuxiang Wei, Jiaxin Ran, Zhiyuan Yao, Xueran Han, Huacan Wang, Ronghao Chen, Lei Zou

AI总结:

本文提出一种统一框架,分析长期对话记忆系统,发现性能差异主要由基础设置驱动,而非特定架构创新,并提供了稳定基线。

摘要:

图结构在对话记忆系统中被日益广泛应用,但关于其有效性的实证发现仍不一致,使得不清楚哪些设计选择真正重要。我们对长期对话记忆架构进行了实验性和系统性分析。我们引入了一个统一框架,将对话记忆系统分解为核心组件,并支持基于图和非图的方法。在此框架下,我们在LongMemEval和HaluMem上进行了受控、分阶段的实验,比较了记忆表示、组织、维护和检索中的常见设计选择。我们的结果表明,许多性能差异是由基础系统设置而非特定架构创新驱动的。基于这些发现,我们确定了未来对话记忆研究的稳定且可靠的强基线。代码可在https://github.com/AvatarMemory/UnifiedMem获取。

英文摘要:

Graph structures are increasingly used in dialog memory systems, but empirical findings on their effectiveness remain inconsistent, making it unclear which design choices truly matter. We present an experimental, system-oriented analysis of long-term dialog memory architectures. We introduce a unified framework that decomposes dialog memory systems into core components and supports both graph-based and non-graph approaches. Under this framework, we conduct controlled, stage-wise experiments on LongMemEval and HaluMem, comparing common design choices in memory representation, organization, maintenance, and retrieval. Our results show that many performance differences are driven by foundational system settings rather than specific architectural innovations. Based on these findings, we identify stable and reliable strong baselines for future dialog memory research. Code are available at https://github.com/AvatarMemory/UnifiedMem

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。

↑