A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agents

A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agents

• ID: ce60e01f
• 原文链接: https://arxiv.org/abs/2602.06052
• PDF: https://arxiv.org/pdf/2602.06052v4
• 作者: Wei-Chieh Huang, Weizhi Zhang, Yueqing Liang, Yuanchen Bei, Yankai Chen, Tao Feng, Xinyu Pan, Zhen Tan, Yu Wang, Tianxin Wei, Shanglin Wu, Ruiyao Xu, Liangwei Yang, Rui Yang, Wooseong Yang, Chin-Yuan Yeh, Hanrong Zhang, Haozhen Zhang, Siqi Zhu, Henry Peng Zou, Wanjia Zhao, Song Wang, Wujiang Xu, Zixuan Ke, Zheng Hui, Dawei Li, Yaozu Wu, Langzhou He, Chen Wang, Xiongxiao Xu, Baixiang Huang, Juntao Tan, Shelby Heinecke, Huan Wang, Caiming Xiong, Ahmed A. Metwally, Jun Yan, Chen-Yu Lee, Hanqing Zeng, Yinglong Xia, Xiaokai Wei, Ali Payani, Yu Wang, Haitong Ma, Wenya Wang, Chenguang Wang, Yu Zhang, Xin Eric Wang, Yongfeng Zhang, Jiaxuan You, Hanghang Tong, Xiao Luo, Xue Liu, Yizhou Sun, Wei Wang, Julian McAuley, James Zou, Jiawei Han, Philip S. Yu, Kai Shu
• 日期: 2026-01-14
• 更新: 2026-08-04
• 分类: agents
• 来源类型: arxiv
• 标签: agent-memory, survey, long-horizon-agents, self-evolving, arxiv
• 质量评分: 5/5
• 抓取时间: 2026-08-15T04:45:00Z

中文导读

综述：AI 进入"下半场"后，研究重心从模型创新与跑分转向问题定义与真实世界的严格评估。agentic coding、deep research、computer use 等长程、动态、依赖用户上下文的场景面临固定上下文窗口之外的"上下文爆炸"，agent 必须在扩展交互中持续积累、管理并选择性复用信息——记忆由此成为弥合实用性差距的关键方案（仅 2025 年就有数百篇记忆相关论文）。记忆不再只是被动存储，而是 agent 自我进化的载体：短期记忆控制在执行中哪些经验被感知与抽象，长期记忆把它们固化为可复用的知识与技能，构成 agent 从自身经验中改进的闭环。综述沿三个维度给出基础 agent 记忆的统一视图：记忆基质（内部参数化状态与外部检索增强存储）、认知机制（感知、工作、情景、语义、程序性记忆）、记忆主体（以用户为中心的个性化与以 agent 为中心的经验）。随后分析单/多智能体拓扑下的记忆操作，指出记忆管理本身正成为可训练能力——强化学习驱动的上下文策展、决策时经验固化、以及可移植可共享的 agent 技能生态。最后梳理记忆效用评测基准与开放挑战。

为什么值得关注

agent memory 方向的最新全景综述，已被 TMLR 接收并获 Survey Certification；配项目页 github.com/AgentMemoryWorld/Awesome-Agent-Memory。三个维度（substrate × cognitive mechanism × subject）的分类法适合作为该方向检索与选型的入口框架；"memory management as trainable capability"一节把 RL context curation、experience consolidation、portable skills 三条线收拢，是当前 agent 记忆工程化的直接参照。

关键信息

• 论文标题：A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agents
• arXiv：2602.06052（v4，2026-08-04 更新）
• 发表：Accepted at Transactions on Machine Learning Research (TMLR) with Survey Certification
• 主分类：cs.CL；跨类：cs.AI
• 项目页：https://github.com/AgentMemoryWorld/Awesome-Agent-Memory
• 三维分类法：memory substrate（parametric internal state / external retrieval-augmented stores）× cognitive mechanism（sensory, working, episodic, semantic, procedural）× memory subject（user-centric / agent-centric）

English Abstract

Research in artificial intelligence is shifting from model innovations and benchmark scores towards problem definition and rigorous real-world evaluation. As the field enters the "second half," the central challenge becomes real utility in long-horizon, dynamic, and user-dependent settings such as agentic coding, deep research, and computer use, where LLM-based agents face context explosion beyond fixed context windows and must continuously accumulate, manage, and selectively reuse information across extended interactions. Memory, with hundreds of papers released in 2025, therefore emerges as the critical solution to fill this utility gap. Beyond passive storage, memory is increasingly the substrate through which agents self-evolve: short-term memory gates which experiences are perceived and abstracted during execution, while long-term memory consolidates them into reusable knowledge and skills, forming the loop through which agents improve from their own experience. In this survey, we provide a unified view of foundation agent memory along three dimensions: memory substrate (internal parametric state and external retrieval-augmented stores), cognitive mechanism (sensory, working, episodic, semantic, and procedural), and memory subject (user-centric personalization and agent-centric experience). We then analyze how memory is operated under single- and multi-agent topologies and highlight learning policies over memory operations, showing how memory management itself is becoming a trainable capability spanning reinforcement-learned context curation, experience consolidation at decision time, and the emerging ecosystem of portable, shareable agent skills. Finally, we review evaluation benchmarks and metrics for memory utility, and outline open challenges and future directions.

Obsidian Notes

• 内容由 opencli arxiv paper 2602.06052 -f json 拉取的 arXiv 元数据与摘要生成；TMLR 接收信息来自 arXiv comment 字段。
• 中文导读与价值判断锚定在摘要声明与条目已有摘要上；未补充摘要之外的实验细节。
• 相关条目：0a9a891f（KV cache 层的 agent 记忆下沉）、b7d52f88（记忆篡改攻击）。