arXiv cs.CL· Haozhen Zhang, Quanyu Long, Jianzhu Bao, Tao Feng, Weizhi Zhang, Haodong Yue, Wenya Wang·· 11 小时前AI 评分38
MemSkill：为自进化智能体学习并演化记忆技能
MemSkill: Learning and Evolving Memory Skills for Self-Evolving Agents
AI 导读
MemSkill 将 LLM 智能体的记忆操作重构为可学习、可演化的记忆技能，由 controller 学习选择少量相关技能、LLM executor 生成技能引导的记忆，并通过 designer 定期复盘失败案例来改进和新增技能。
正文 · 原文
Abstract:Most Large Language Model (LLM) agent memory systems rely on a small set of static, hand-designed operations for extracting memory. These fixed procedures hard-code human priors about what to store and how to revise memory, making them rigid under diverse interaction patterns and inefficient on long histories. To this end, we present \textbf{MemSkill}, which reframes these operations as learnable and evolvable memory skills, structured and reusable routines for extracting, consolidating, and pruning information from interaction traces. Inspired by the design philosophy of agent skills, MemSkill employs a \emph{controller} that learns to select a small set of relevant skills, paired with an LLM-based \emph{executor} that produces skill-guided memories. Beyond learning skill selection, MemSkill introduces a \emph{designer} that periodically reviews hard cases where selected skills yield incorrect or incomplete memories, and evolves the skill set by proposing refinements and new skills. Together, MemSkill forms a closed-loop procedure that improves both the skill-selection policy and the skill set itself. Experiments on LoCoMo, LongMemEval, HotpotQA, and ALFWorld demonstrate that MemSkill improves task performance over strong baselines and generalizes well across settings. Further analyses shed light on how skills evolve, offering insights toward more adaptive, self-evolving memory management for LLM agents. Code is available at this https URL
| Comments: | Accepted by NeurIPS 2026. Code is available at this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2602.02474 [cs.CL] |
| (or arXiv:2602.02474v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2602.02474 arXiv-issued DOI via DataCite |
Submission history
From: Haozhen Zhang [view email]
[v1]
Mon, 2 Feb 2026 18:53:28 UTC (1,321 KB)
[v2]
Sun, 24 May 2026 19:01:09 UTC (1,327 KB)
[v3]
Fri, 6 Nov 2026 16:31:39 UTC (1,314 KB)
来源：arXiv cs.CL · arxiv.org