发表机构
Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing, School of Artificial Intelligence, Beihang University; Beijing academy of blockchain and edge computing; National University of Defense Technology; MemTensor (Shanghai) Technology Co., Ltd.（北京航空航天大学人工智能学院未来区块链与隐私计算北京高精尖创新中心; 北京区块链与边缘计算研究院; 国防科技大学; MemTensor（上海）科技有限公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对多智能体系统记忆易受噪声污染且缺乏结构的问题，提出CoMem架构，通过私有经验沉淀、集体智慧策展和并行双流检索实现集体-个体记忆协同，在ALFWorld和PDDL基准上取得强性能并避免记忆污染。
AI 中文摘要
设计有效的记忆机制对于推进LLM驱动的多智能体系统（MAS）至关重要，有助于智能体共同学习并随时间表现得更好。尽管近期工作已带来强大的协作能力，但大多数方法仍使用扁平、非结构化的记忆，这些记忆容易充满噪声并抹平智能体之间的差异。为解决这一问题，我们引入了集体-个体记忆协同的概念，并提出了CoMem，一种统一了私有经验和共享知识的多智能体学习架构。CoMem的特点包括：（i）私有经验沉淀，使每个智能体能够随时间保留并更新自身有用的记忆；（ii）集体智慧策展，精心挑选仅被广泛验证的想法在智能体间共享；（iii）并行双流检索，允许智能体同时从自身记忆和群体智慧中提取信息，利用聚类确保高效检索。在ALFWorld和PDDL基准上的实验表明，CoMem实现了强大的整体性能，并稳健地避免了记忆污染。
英文摘要
Designing effective memory mechanisms is crucial for advancing LLM-driven Multi-Agent Systems (MAS), helping agents learn together and perform better over time. While recent work has led to strong cooperation skills, most methods still use flat, unstructured memories, which easily get filled with noise and erase differences between agents. To address this, we introduce the concept of collective-individual memory synergy and propose CoMem, an architecture that unifies both private experience and shared knowledge for multi-agent learning. CoMem features:(i) Private Experience Sedimentation, which lets each agent keep and update its own useful memories over time;(ii) Collective Wisdom Curation, which carefully selects only widely proven ideas to be shared among agents;(iii)Parallel Dual-Stream Retrieval, which allows agents to draw both from their own memory and the group's wisdom, using clustering to ensure diversity.Experiments on ALFWorld and PDDL benchmarks show that CoMem achieves strong overall performance and robustly avoids memory pollution.
CommentsSubmitted to AAAI 2027.9 pages,4 figures
