cs.AI提交于 2026-10-28已译

检索并非足够：智能体记忆中的过时事实

When Retrieval Is Not Enough: Obsolete Facts in Agent Memory

Oleeviya Babu Poikarayil · Tony Basil Babu · Abdulrahman Nahhas · Het Darshan Mehta · Jawid Ahmad Baktash · Ernesto William De Luca · et al.

中文摘要

记忆智能体基准常将检索与答案生成合并评分，难以区分失败源于遗漏更新还是未利用已检索的更新。在 MemoryAgentBench 的单跳 FactConsolidation 任务中，使用 Qwen2.5-7B-Instruct 时，当检索证据预算从 64 增至 512 tokens，6K 历史下的 BM25 准确率从 45% 降至 32%，尽管事后审计发现每个检索集均包含最新陈述。多数错误出现在包含过时陈述版本的问题中。在固定提示和四句证据大小下，将一句无关陈述替换为过时的竞争陈述，导致 Qwen2.5 准确率从 100% 降至 44.6%，Qwen3-14B 降至 87.8%，所有 50 个有害配对翻转均被归类为过时。32K 历史复现同样显示 BM25-64 优于 BM25-512（69% 对 58%）。结果表明，检索到正确事实并不足够，过时的竞争陈述可能比同等大小的无关证据更具危害性。

关键要点

01现有基准难以区分检索失败与未利用已检索更新导致的错误。
02增大检索证据预算反而降低准确率，因引入过时的竞争陈述。
03过时陈述对模型性能的危害大于同等大小的无关证据。
04在 6K 和 32K 历史长度下，较小的检索预算表现更优。

解读

尚无解读。

原始英文摘要

arXiv:2610.36308v1 Announce Type: new Abstract: Memory-agent benchmarks often collapse retrieval and answer generation into one score, making it hard to tell whether a failure comes from missing an update or from not using an update that was already retrieved. We examine this distinction on the single-hop FactConsolidation task of MemoryAgentBench. With Qwen2.5-7B-Instruct on 6K-history BM25 accuracy drops from 45% to 32% when the retrieved evidence budget grows from 64 to 512 tokens, even though a gold-assisted post-hoc audit finds the newest statement in every retrieved set. Most of the errors occur on questions that also contain an obsolete statement version. Under a frozen prompt and equal four-statement evidence size, replacing one unrelated statement with the obsolete competing statement reduces accuracy from 100% to 44.6% for Qwen2.5 and from 100% to 87.8% for Qwen3-14B, with all 50 harmful paired flips classified as stale. A 32K-history replication likewise favors BM25-64 over BM25-512 (69% versus 58%). These results suggest that retrieving the correct fact is not enough. An obsolete competing statement can be more harmful than an unrelated piece of evidence of the same size.