从计算机架构视角重构多智能体记忆系统，解决协同难题。
Multi-Agent Memory from a Computer Architecture Perspective: Visions and Challenges Ahead
- 提出三层记忆层级：输入输出、缓存与主存
- 发现跨智能体缓存共享与结构化访问控制两大协议缺口
- 聚焦记忆一致性问题，适合系统设计与智能体研究者
随着大模型智能体演变为协作式多智能体系统，其记忆需求迅速复杂化。本文将多智能体记忆视为计算机架构问题，区分共享与分布式记忆范式，提出包含输入输出、缓存和主存的三层记忆层次结构，并识别出两个关键协议缺陷：跨智能体缓存共享机制缺失，以及结构化记忆访问控制不足。文章强调，当前最紧迫的开放挑战是多智能体记忆一致性。该架构视角为构建可靠、可扩展的多智能体系统提供了基础框架。
原文摘要 · Abstract (English)
As LLM agents evolve into collaborative multi-agent systems, their memory requirements grow rapidly in complexity. This position paper frames multi-agent memory as a computer architecture problem. We distinguish shared and distributed memory paradigms, propose a three-layer memory hierarchy (I/O, cache, and memory), and identify two critical protocol gaps: cache sharing across agents and structured memory access control. We argue that the most pressing open challenge is multi-agent memory consistency. Our architectural framing provides a foundation for building reliable, scalable multi-agent systems.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。