MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents· MeClear: 长时LLM代理的合作归因与风险记忆清理
Long horizon Large Language Model (LLM) agents rely on external memory systems to preserve user preferences and task knowledge across extended interactions. Conventional retrieval mechanisms optimize semantic compatibility rather than downstream utility, frequently introducing outdated, misleading, or conflicting evidence into the active context. We present MeClear, a task conditioned memory clearance framework that identifies memories featuring negative downstream utility through cooperative attribution and selectively suppresses them from agent execution. MeClear combines Leave One Out screening with sampled cooperative Shapley attribution to distribute utility across interacting evidence, effectively resolving redundant conflict masking where single removal evaluations fail. Utilizing a
提出MeClear框架，解决长时LLM代理中外部记忆系统的负效用问题。
- 核心方法
- MeClear通过合作归因技术评估记忆条目的效用，并使用Leave One Out筛选和采样合作Shapley值归因来识别和抑制负效用记忆。
- 适合谁读
- 研究者、工程师
- 要解决的问题
- 长时大语言模型代理在长时间交互中，传统检索机制导致外部记忆系统引入过时、误导或冲突的信息。
- 关键实验
- 未提供
- 主要贡献
- 提供了一种有效的记忆清理方法，减少了长时交互中LLM代理的错误决策。
- 意义与局限
- MeClear提高了大语言模型代理的可靠性和效率，但其适用性和局限性需进一步实验验证。