RippleMem：从孤立检索到智能体长期记忆的关联回忆
RippleMem: From Isolated Retrieval to Associative Recollection for Long-Term Agent Memory
浏览论文内容
中文总结 AI 辅助
RippleMem是一种智能体长期记忆系统，以自适应关联回忆替代孤立检索，在LoCoMo、LongMemEval-S数据集上提升LLM-as-a-Judge准确率并降低图构建成本，性能优于现有方法。
中文摘要 AI 辅助
基于大语言模型（LLM）的智能体越来越依赖外部记忆来支持长周期推理与交互，但其主要瓶颈并非简单存储过往经验，而是当相关信息分布在多次交互中时，能正确恢复所需证据集合。现有方法在该访问问题上存在不足：全上下文方法需进行嘈杂的长上下文搜索，平面检索常返回孤立且不完整的记录，基于图的记忆系统构建成本高且会压缩丰富的事件上下文。本文提出RippleMem，一种长期记忆系统，用自适应关联回忆替代一次性检索。受线索依赖型情景记忆与关联补全启发，RippleMem将交互历史存储为富含线索的情景记忆单元，并以事件为中心组织成记忆图。给定查询时，它先通过混合线索召回相关记忆锚点，再沿语义与结构关联从这些锚点扩展以恢复缺失的支撑证据。如此，初始召回的记忆既作为答案上下文，又作为补全回答所需证据的线索。在LoCoMo与LongMemEval-S上的实验表明，RippleMem在所有评估设置中取得最佳整体性能，在LoCoMo上提升LLM-as-a-Judge准确率3.95%，在LongMemEval-S上提升达11.87%，同时将图构建成本降低约30倍。
英文摘要
LLM-based agents increasingly rely on external memory to support long-horizon reasoning and interaction. However, the main bottleneck is not simply storing past experience, but recovering the right set of evidence when relevant information is distributed across many interactions. Existing approaches struggle with this access problem. Full-context methods require noisy long-context search, flat retrieval often returns isolated and incomplete records, and graph-based memory systems can be expensive to construct while compressing rich event context. We introduce RippleMem, a long-term memory system that replaces one-shot retrieval with adaptive associative recollection. Inspired by cue-dependent episodic retrieval and associative completion, RippleMem stores interaction history as cue-rich episodic memory units and organizes them in an event-centric memory graph. Given a query, it first recalls relevant memory anchors through hybrid cues, then expands from these anchors along semantic and structural associations to recover missing supporting evidence. In this way, initially recalled memories serve not only as answer context, but also as cues for completing the evidence needed to answer. Experiments on LoCoMo and LongMemEval-S show that RippleMem achieves the best overall performance across evaluated settings, improving LLM-as-a-Judge accuracy by 3.95% on LoCoMo and up to 11.87% on LongMemEval-S, while reducing graph construction cost by about 30x.
发表机构
- Communication University of China（中国传媒大学）
- Zhilian Yinghe Technology Co., Ltd.（智联映和科技有限公司）
- State Key Laboratory of Media Convergence and Communication（媒体融合与传播国家重点实验室）
机构由 AI 辅助整理，请以论文原文为准。
