arXiv cs.AI· Tianyi Xu, Huazheng Wang·· 12 小时前AI 评分38
MemCryst：通过记忆与技能双向学习实现终身上下文适应
Lifelong Context Adaptation through Bidirectional Memory-Skill Learning
AI 导读
MemCryst 提出一种在冻结语言模型上双向学习记忆与技能的终身上下文适应框架，记忆中的重复片段支持技能创建与修订，技能引导的交互又反哺新记忆。在涵盖具身控制、长文档阅读、对话、数学与代码的十项设置中，MemCryst 在七项对比中领先；ALFWorld 上联合服务成功率 63.17%，高于仅记忆的 59.83% 和仅技能的 50.50%。代码已开源。
正文 · 原文
Abstract:Lifelong LLM agents must learn from an expanding history of interactions while making each decision within a bounded context window. Memory makes past experience available, but the agent may still need to infer how to solve each new task from the retrieved records. Skills provide reusable procedures, yet may omit the task-specific facts needed to apply them. Because retrieved memory and skills shape the agent's behavior, they also influence the experience available for future learning. We introduce MemCryst, a framework for lifelong context adaptation through bidirectional learning of memory and skills with a frozen language model. Recurring episodes stored in memory support skill creation and revision, and the interactions guided by those skills supply new memory for subsequent learning. Facts retrieved from memory and learned skills jointly form the agent's context, while task outcomes adjust the future retrieval influence of the records actually served without rewriting their content. Across ten settings spanning embodied control, long-document reading, conversation, mathematics, and code, MemCryst leads seven reported comparisons. On ALFWorld, joint serving achieves 63.17% success, compared with 59.83% for memory-only and 50.50% for skill-only serving. Code is available at this https URL
| Comments: | 42 pages, including references and appendices. 18 figures, 13 tables |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.32295 [cs.AI] |
| (or arXiv:2610.32295v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.32295 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tianyi Xu [view email]
[v1]
Fri, 23 Oct 2026 07:53:18 UTC (308 KB)
来源：arXiv cs.AI · arxiv.org