为智能体记忆设计事务边界，确保更新可靠、故障可恢复。
MemTxn: A Transaction Boundary for Source-Supported Updates and Complete-State Recovery in Agent Memory
- 外部治理层验证更新来源，保障记忆修改可信
- 在多键故障下恢复完整状态，无需知道实际写入集
- 适合长期运行的智能体系统，提升记忆一致性
持久化内存使长时运行的大语言模型智能体可在会话与任务间复用信息。然而，可写内存中的错误可能持续存在并破坏后续行为。现有系统虽改进了存储与检索，但未提供可靠的更新事务边界与恢复机制。为此，我们提出 MemTxn，一个位于答案模型之外的治理层。MemTxn 验证更新是否由源支持，冲突时选择可见版本，并在故障后恢复应用可见状态。系统采用有序补丁测试（Ordered PatchTest）验证写入，时间解析器（Temporal Resolver）选择版本，以及持久快照日志实现状态恢复。在项目无关审计中，MemTxn 接受全部 60 个支持原始项，拒绝所有 179 个强负例。在 LongMemEval-S 与 LoCoMo 状态下的持续多键故障场景中，它在不知实际物理写入集的情况下恢复了完整的声明活跃映射。在 MemoryAgentBench FactConsolidation 任务上，MemTxn 在十二种答案模型配置中取得最高平均 F1；在五个代表性设置中，相比 Dense 提升 17.06–24.07 分。
原文摘要 · Abstract (English)
Persistent memory lets long-running large language model agents reuse information across sessions and tasks. Yet errors in writable memory can persist and corrupt future behavior. Existing systems improve storage and retrieval, but they do not provide a transaction boundary for reliable updates and recovery. We therefore propose MemTxn, a governance layer outside the answer model. MemTxn verifies whether an update is supported by its source. It also selects the visible version when facts conflict and restores the application-visible state after a fault. The system uses Ordered PatchTest to validate writes, a Temporal Resolver to select versions, and a durable snapshot journal to recover state. On an item-disjoint audit, MemTxn accepts all 60 supported originals and rejects all 179 hard negatives. Under persistent multi-key faults on LongMemEval-S and LoCoMo states, it restores the complete declared active map without knowing the actual physical write set. On MemoryAgentBench FactConsolidation, MemTxn achieves the highest average F1 across all twelve answer-model configurations. It outperforms Dense by 17.06--24.07 points in five representative settings.
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。