论文

论文提出 BRIDGE:极简 RAG 记忆系统在多个基准上追平甚至超过复杂方案

RAG in a Trenchcoat: When Minimal Memory Is Enough for Agentic Systems, and When It Isn't

精选理由

做智能体记忆的朋友可以省下不少功夫:论文用 BRIDGE 证明 BM25 加余弦检索就够打平 LangMem 这些复杂记忆框架,还能在 EverMemBench 上高出 5.7-6.3 分。更关键的是它设计了五个对抗探测,解释了不同记忆方案各自在哪里失效。正在给 Agent 选型记忆方案的人应该看看。

一篇 arXiv 论文提出 BRIDGE,一个刻意极简的 RAG 记忆系统:只用链式观察笔记、BM25 加余弦混合检索和按时间顺序渲染,每次查询约消耗 5-7K reader-prompt tokens,摄入和查询各只需一次 LLM 调用。

正文

在 LongMemEval-S 和 BEAM 上,BRIDGE 的成绩与此前 SOTA 在 bootstrap CI 内持平;在 EverMemBench 上以相同 reader 超过最强记忆基线 5.7-6.3 分。作者认为这种持平本身就是结论:现有标准基准衡量的只是记忆检索到了什么,而不是记忆能让智能体实际做到什么。

为检验后者,论文设计了五个对抗性探测任务(如规则应用、模式归纳、动态多轮),共 5 探测 x 20 场景,用跨厂商验证链在生成时评分。结果显示没有任何架构全面胜出:LangMem 这类原子事实抽取器在规则应用上占优,BRIDGE 的块摘要记忆在模式归纳上占优,且该结论在四个领域、三个评分器和三种 reader 上均成立。原因在于原子抽取器丢弃了示范模式的形式,而块摘要检索到了规则但 reader 没有执行,更突出地呈现规则即可基本弥合这一差距。

原文 · arXiv cs.AI

RAG in a Trenchcoat: When Minimal Memory Is Enough for Agentic Systems, and When It Isn't

Agentic memory research has accumulated extraction, supersession, reflection, and graph synthesis on top of retrieval-augmented generation. However, we found that almost none of those primitives earn their cost in our tested memory benchmarks: we show this with BRIDGE, a deliberately minimal RAG memory with chained observation notes, hybrid BM25 with cosine retrieval, and chronological rendering, which reaches parity within bootstrap CI with the prior SOTA on LongMemEval-S and BEAM, and beats the best memory baseline on EverMemBench by 5.7-6.3 points at matched reader. It runs at ~5-7K reader-prompt tokens per query, using a single LLM call for each ingestion or query. We argue that this parity is itself the result: standard benchmarks score what memory retrieves, not what memory enables when the agent acts. To test the latter, we introduce five adversarial probes (Rule Application +-, Pattern Induction Style / Procedural, Dynamic Multiturn) judged at generation time by a cross vendor verification chain, where each query can only be solved by applying a remembered rule or pattern it never mentions. Across 5 probes x 20 scenarios, no architecture wins universally: atomic fact extractors (LangMem) win rule application, while BRIDGE's block-summary memory wins pattern induction. This result holds across four domains, three graders, and three reader families. Each design fails where the other wins, for reasons a recall benchmark cannot see: atomic-fact extractors discard the form a demonstrated pattern takes, while block-summary memory retrieves the rule but the reader does not act on it, a gap that presenting the rule more prominently largely closes. What separates memory designs is not what they retrieve but what they enable an agent to do, and that is what agentic memory should be evaluated on.

产业坐标与信源核验

本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:RAG、BRIDGE、LangMem、LongMemEval,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Jingyu Liu, Zongze Li, Zach Xu, Zhanhui Zhou, Tahseen Rabbani, Dawn Song, Ce Zhang),经 AITOP 多维评分管线结构化提炼。