让大模型记忆更新更智能，避免覆盖有用知识。
The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory
- 用动态触发机制判断记忆更新是否异常
- 在六组实验中提升准确率2.7至4.6个百分点
- 无需修改原有更新规则，适合各类记忆系统
序列演化的大语言模型记忆可让智能体复用过往经验，但现有系统通常无差别采纳每个本地生成的记忆更新，导致有益于当前任务的更新可能覆盖已有知识，引入过度特定规则或使最终记忆偏向近期样本。我们提出Janus——一种插件式记忆控制器，决定是否接受候选更新或保留旧记忆。为高效决策，Janus使用内存动量触发器识别记忆更新轨迹中的可疑偏差，并在紧凑的混合评估集（涵盖覆盖率、边界任务和新任务）上对比新旧记忆，而非重放完整历史。Janus方法无关，可封装现有更新器而不改变其更新规则。在六个数据集、两种主干LLM和两种记忆更新器上，Janus相较基线更新器平均准确率提升2.7至4.6个百分点。
原文摘要 · Abstract (English)
Sequentially evolving LLM memory enables agents to reuse past experience, but existing systems usually deploy each locally generated memory update without checking whether it improves future behavior. As a result, updates that help the current task may overwrite useful knowledge, introduce over-specific rules, or bias the final memory toward recent examples. We propose Janus, a plug-in memory controller that decides whether to accept a candidate memory update or retain the previous memory. To make this decision efficient, Janus uses a Memory Momentum Trigger to identify suspicious deviations in the memory-update trajectory, and compares old and new memories on a compact hybrid evaluation set of coverage, boundary, and fresh tasks instead of replaying the full history. Janus is method-agnostic and wraps existing updaters without changing their update rules. Across six datasets, two backbone LLMs, and two memory updaters, Janus improves average accuracy by +2.7 to +4.6 points over the corresponding base updaters.
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。