用于智能体内存优化的机制性注意力引导
Mechanistic Attention Guidance for Agent Memory Refinement
- Tsinghua University（清华大学）
- Beijing Institute of Mathematical Sciences and Applications（北京应用数学科学研究院）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
研究如何优化智能体内存，提出注意力引导内存优化框架AGMR，利用检索头注意力揭示内存使用模式，指导段级内存更新，经实验验证其能提升任务性能与内存效率。
AI中文摘要：
现有的自我进化内存系统主要基于文本输出（如任务轨迹和反思）来改进智能体内存。然而，这种基于文本的范式很少纳入内部机制信号，导致在任务执行期间实际如何利用检索到的内存未得到充分探索。这一局限性可能导致不可靠的错误归因和幻觉性的内存修改。在这项工作中，我们表明检索头注意力提供了一个机制信号来揭示段级内存利用情况。通过在内存段和决策步骤上聚合注意力，我们构建了一个上下文利用矩阵，该矩阵揭示了反复出现的内存使用模式并指示相应的优化策略。在此基础上，我们提出了注意力引导内存优化（AGMR）框架，该框架利用注意力揭示的使用模式来指导有针对性的段级内存更新。AGMR对失败的执行进行内存纠正或增强，对成功的执行简化内存，并通过重新执行验证每次更新。在交互式决策基准上的实验表明，与仅基于文本的内存优化基线相比，AGMR提高了任务性能和内存效率。代码可在该https URL获取。
英文摘要：
Existing self-evolving memory systems mainly improve agent memory based on textual outputs, such as task trajectories and reflections. However, this text-based paradigm rarely incorporates internal mechanistic signals, leaving how retrieved memory is actually utilized during task execution underexplored. This limitation can lead to unreliable error attribution and hallucinated memory modifications. In this work, we show that retrieval-head attention provides a mechanistic signal for revealing segment-level memory utilization. By aggregating attention over memory segments and decision steps, we construct a context utilization matrix that exposes recurring memory-use patterns and indicates corresponding refinement strategies. Building on this observation, we propose Attention-Guided Memory Refinement (AGMR), a framework that uses utilization patterns revealed by attention to guide targeted segment-level memory updates. AGMR corrects or enhances memory for failed executions, simplifies memory for successful executions, and verifies each update through re-execution. Experiments on interactive decision-making benchmarks show that AGMR improves both task performance and memory efficiency over text-only memory refinement baselines. Code is available at https://anonymous.4open.science/r/AGMR_code-3262/
