arXiv cs.CL· Uiji Hwang, Richeng Xuan, Taeuk Kim·· 2 天前AI 评分32
重新审视工具图记忆在多轮工具调用中的作用
Rethinking the Role of Tool Graph Memory in Multi-Turn Tool Use
AI 导读
研究在九个模型和两个基准上评估工具图记忆，发现其相较基线很少带来显著提升，且即使有效也需付出超过两倍的 LLM 调用成本。其引导常不正确，随机或无引导仍保留大部分增益，说明收益来自额外计算而非记忆本身；将相同经验存为文本洞察则以几乎零额外调用超越工具图记忆。
正文 · 原文
Abstract:Large language model (LLM) agents solve complex tasks by calling external tools over multiple turns. Recent work has proposed tool graph memory, which stores past experience as tool transitions to guide subsequent selection. However, it remains unclear whether this component consistently improves agents' tool use or undermines it. We evaluate tool graph memory on nine models and two benchmarks, finding that it seldom improves significantly over baselines, and that its gains, where present, cost over twice the LLM calls. The memory's guidance is often incorrect, and random or no guidance retains most of the gain, implying that the benefit stems from extra computation rather than the memory itself. Moreover, even with the correct next tool suggested, many failures remain, such as giving up without a tool call. Storing the same experience instead as textual insights outperforms tool graph memory with almost no extra LLM calls. This suggests that memory serves better as a soft prior than as a prescription of the next tool, leading to a more effective and generalizable solution.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.21571 [cs.CL] |
| (or arXiv:2610.21571v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21571 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Uiji Hwang [view email]
[v1]
Thu, 15 Oct 2026 09:24:30 UTC (201 KB)
来源：arXiv cs.CL · arxiv.org