让大模型像人一样记东西、删无用记忆，提升长期对话能力。
SCM: Sleep-Consolidated Memory with Algorithmic Forgetting for Large Language Models
- 模仿人类记忆机制，设工作记忆与分阶段睡眠式巩固
- 十轮对话完美回忆，噪声降低90.9%，检索延迟低于1毫秒
- 适合研究长时记忆、智能体长期学习的学者与开发者
我们提出SCM（Sleep-Consolidated Memory），一种基于神经科学原理的大语言模型记忆架构，旨在解决当前系统缺乏持久、结构化且生物合理记忆的根本问题。现有方法依赖截断上下文窗口、无限增长向量数据库或分层存储，均缺乏整合与遗忘机制。SCM受人类记忆启发，实现五个核心组件：有限容量的工作记忆、多维重要性标记、具区分性的非快速眼动（NREM）与快速眼动（REM）睡眠阶段离线巩固、有意识的价值导向遗忘，以及支持自我反思的计算自模型。在涵盖八项测试的标准基准中，原型在十轮对话中实现完美召回率，通过自适应遗忘将记忆噪声降低90.9%。即使存储数百个概念，内存搜索延迟仍保持在1毫秒以下。该工作为可整合、可优先排序、可遗忘的记忆系统奠定架构基础，提供可验证的研究平台以推进大模型记忆研究。
原文摘要 · Abstract (English)
We present SCM (Sleep-Consolidated Memory), a research preview of a memory architecture for large language models that draws on neuroscientific principles to address a fundamental limitation in current systems: the absence of persistent, structured, and biologically plausible memory. Existing approaches rely on truncating context windows, growing vector databases without bound, or tiered storage systems that lack consolidation and forgetting mechanisms. SCM implements five core components inspired by human memory: a limited-capacity working memory, multi-dimensional importance tagging, offline sleep-stage consolidation with distinct NREM and REM phases, intentional value-based forgetting, and a computational self-model enabling introspection. Across a standardized benchmark suite of eight tests, the prototype achieves perfect recall accuracy over ten-turn conversations while reducing memory noise by 90.9% through adaptive forgetting. Memory search latency remains below one millisecond even with hundreds of stored concepts. This work establishes the architectural foundations for memory systems that consolidate, prioritize, and forget, offering a testable platform for advancing LLM memory research.
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。