arXiv cs.CL· Zedong Peng·· 13 小时前AI 评分48
MemoryTax：智能体记忆的苦涩教训
MemoryTax: The Bitter Lesson of Agent Memory
AI 导读
一项实证研究在 LoCoMo 单段对话（419 轮、152 个问题）上对比 full context、BM25、Mem0、Zep 四种记忆方法与 Pi、Codex、Claude Code 三种编码智能体，分别搭配 GPT-4o-mini、GPT-6 Luna 和 Claude Opus 5.5。
正文 · 原文
Abstract:Agent memory systems such as Mem0 and Zep decide what to keep before any question arrives; coding agents keep the raw history and let the model search it with its own tools. We ask whether designed memory still helps once the model is strong enough to search for itself. On one LoCoMo conversation (419 turns, 152 questions), we run four memory methods (full context, BM25, Mem0, Zep) and three coding agents (Pi, Codex, Claude Code), each with GPT-4o-mini, GPT-6 Luna and Claude Opus 5.5, and record what every rollout costs. This is a short empirical study, with one conversation and one run per cell. Agent memory is a costly patch for a weak model: Mem0 beats full context only with GPT-4o-mini, and its build alone costs more than a whole full-context rollout. Agent memory also has a ceiling, set by the model that writes it: a stronger reader does not recover what a weaker writer left out. With Claude Opus 5.5, Pi tops the cost-accuracy frontier, scoring 99% for less than half the cost of full context; its one miss is a wrong answer key. Part of that lead comes from photo captions that only the agents could read: on the 88 questions whose evidence has no photo, Pi and full context differ by one answer.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.29823 [cs.CL] |
| (or arXiv:2610.29823v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.29823 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zedong Peng [view email]
[v1]
Wed, 21 Oct 2026 14:46:27 UTC (854 KB)
来源：arXiv cs.CL · arxiv.org