发表机构
University of Science and Technology of China（中国科学技术大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对LLM智能体持久记忆将局部推理错误固化为持久事实的问题，提出CASK提交规则，通过保留话语所有权信号区分共享事实与临时内容，防止污染后续回答。
AI 中文摘要
持久记忆使LLM智能体能够跨对话携带经验，但同时也将局部推理错误转变为持久性错误。在思考过程中，智能体可能考虑一个计划、模拟一个工具结果、报告另一个说话者的信念，然后全部拒绝。如果记忆仅保留最终生成的句子，那些曾经有用的可能性日后可能作为事实返回。这些记录既非捏造也非无关，只是脱离了其有效时的上下文。我们将这一缺失的上下文识别为"话语所有权"：即许可一个命题的世界、分支或说话者。我们的第一个发现是反直觉的。语言模型已经携带了一个因果活跃的所有权信号，但传统记忆接口在将推理转换为记录时会丢弃该信号。我们引入CASK（因果锚定作用域键），一种提交规则，保留该信号，使共享世界的事实进入持久记忆，而临时内容仅在其原始作用域内可用。我们的第二个发现是，保留该信号最明显的方式——存储发现的内部坐标——并不可靠，因为等效表示不必保持相同的坐标。CASK转而保留表达所有权的稳定关系。受控的长对话冲突和工具智能体轨迹表明，该设计改善了记忆准入，并防止临时内容污染后续回答，同时补充了运行时溯源。由此产生的提交边界使智能体能够探索更多可能性，而不赋予每个中间句子对未来行为的权威。
英文摘要
Persistent memory allows an LLM agent to carry experience across conversations, but it also turns a local reasoning mistake into a durable one. During deliberation, an agent may consider a plan, simulate a tool result, report another speaker's belief, and then reject all of them. If memory retains only the resulting sentences, those once-useful possibilities can later return as facts. The record is neither fabricated nor irrelevant; it has simply been detached from the context in which it was valid. We identify this missing context as \emph{discourse ownership}: the world, branch, or speaker that licenses a proposition. Our first finding is counterintuitive. Language models already carry a causally active signal for ownership, yet conventional memory interfaces discard it when they convert reasoning into records. We introduce CASK (Causally Anchored Scoping Keys), a commit rule that preserves this signal so that shared-world facts enter durable memory while provisional content remains available only within its original scope. Our second finding is that the most obvious way to preserve the signal---storing the discovered internal coordinates---is unreliable because equivalent representations need not keep the same coordinates. CASK instead preserves the stable relations that express ownership. Controlled long-conversation conflicts and tool-agent traces show that this design improves memory admission and prevents provisional content from contaminating later answers while complementing runtime provenance. The resulting commit boundary lets agents explore more possibilities without granting every intermediate sentence authority over future behavior.
Comments20 pages, 2 figures
Journal refNeurIPS 2026 Workshop LP4FM Poster, https://openreview.net/forum?id=jXE0A1Pv0T
