发表机构
College of Computer Science and Artificial Intelligence, Fudan University; School of Cyberspace Science and Technology, Beijing Institute of Technology（复旦大学计算机与人工智能学院; 北京理工大学网络空间科学与技术学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
MeClear提出合作博弈归因与风险感知记忆清除框架，通过识别负效用记忆并选择性抑制，在长程LLM智能体中实现85.9%召回率和82.3%任务恢复率，较LOO基线提升25.5个百分点。
AI 中文摘要
长程大语言模型（LLM）智能体依赖外部记忆系统来在扩展交互中保留用户偏好和任务知识。传统的检索机制优化语义兼容性而非下游效用，经常将过时、误导或冲突的证据引入活动上下文。我们提出MeClear，一种任务条件化的记忆清除框架，通过合作归因识别具有负下游效用的记忆，并选择性地将其从智能体执行中抑制。MeClear将留一法筛选与采样合作Shapley归因相结合，以在交互证据间分配效用，有效解决单一移除评估失效的冗余冲突遮蔽问题。利用归因排名，MeClear在嵌套过滤上执行查询范围的最小清除策略，验证清除上下文上的任务恢复，而不永久改变持久记忆库。在十个长对话记忆池上的全面实验评估表明，MeClear实现了85.9%的目标召回率和82.3%的整体任务恢复率，比留一法（LOO）基线提高了25.5个百分点。
英文摘要
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retrieval mechanisms optimize semantic compatibility rather than downstream utility, frequently introducing outdated, misleading, or conflicting evidence into the active context.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 attribution rankings, MeClear executes a query scoped minimal clearance strategy over a nested filtration, verifying task recovery on the cleared context without permanently altering the persistent memory bank. Comprehensive experimental evaluations across ten long dialogue memory pools demonstrate that MeClear achieves a target recall of 85.9% and an overall task recovery rate of 82.3%, representing a 25.5 percentage point improvement over Leave One Out (LOO) baselines.
Comments15 pages, 12 figures
