Reliable Post-Retrieval Assembly for Agent Memory: Separating Evidence Extraction from Policy Execution
不要询问LLM追踪新鲜度:一种确定性的内存冲突解决策略
机构 * IIT Kgp(印度理工学院科钦分校)
专题命中 长上下文与记忆 :LLM(title_cn,summary_cn);分类 cs.CL、cs.AI
AI总结 针对基于LLM的内存系统中事实冲突解决性能低下的问题,提出用候选提取加Python max(serial)的确定性聚合替代LLM判断,在单跳任务上提升10.8个百分点,并扩展到多跳任务。
Comments 11 pages, 5 tables. Accepted as a poster at the Lifelong Agent Workshop at COLM 2026. Code: https://github.com/cvikasreddy/memory-conflict-resolution