arXiv cs.CL· Yanyu Yao, Shangze Li, Zhi Zheng, Hui Zheng, Qi Liu, Tong Xu, Enhong Chen·· 12 小时前AI 评分38
AtomMem：用原子事实为 LLM 智能体构建简单高效的记忆系统
AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts
AI 导读
针对 LLM 固定上下文窗口限制长期信息积累的问题，AtomMem 提出一种长期记忆系统，通过 Fact Executor 从长对话中抽取高价值原子事实作为高效记忆表示，并组织为层级事件结构与时间画像，检索时激活关联记忆图连接碎片化记忆。在 LoCoMo benchmark 上，AtomMem 在多项推理任务中达到 SOTA，为部署个性化智能体提供了可扩展且经济可行的方案。
正文 · 原文
Abstract:Large language models (LLMs) demonstrate strong reasoning and generation abilities, but their fixed context windows limit long-term information accumulation and reuse across multi-session interactions. Existing memory-augmented systems often construct memory in a coarse and unstable manner, relying on inefficient memory representations or unstable unconstrained updates. To address these challenges, we propose AtomMem, a long-term memory system designed for value-dense storage and stable memory evolution. AtomMem introduces a Fact Executor, which selectively extracts high value atomic facts from long form interactions to serve as highly efficient memory representations. Subsequently, AtomMem organizes these facts into hierarchical event structures and temporal profiles, capturing coherent episodic contexts and tracking dynamically evolving user attributes over time. During retrieval, the system activates an associative memory graph to connect fragmented memories. Experiments on the LoCoMo benchmark confirm that AtomMem achieves state-of-the-art performance across various reasoning tasks, offering a scalable and economically viable solution for deploying intelligent personalized agents.
| Comments: | 23 pages, 10 figures, 8 tables |
| Subjects: | Computation and Language (cs.CL) |
| ACM classes: | I.2.7 |
| Cite as: | arXiv:2606.19847 [cs.CL] |
| (or arXiv:2606.19847v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.19847 arXiv-issued DOI via DataCite |
Submission history
From: Yanyu Yao [view email]
[v1]
Thu, 18 Jun 2026 06:56:15 UTC (2,547 KB)
[v2]
Sat, 31 Oct 2026 14:38:22 UTC (2,504 KB)
来源：arXiv cs.CL · arxiv.org