---
id: 20260517-T0-06
title: "模型自适应工具必要性揭示大模型工具使用的知行鸿沟"
title_en: "Study Reveals Knowing-Doing Gap in LLMs' Tool Use"
url: https://ai.daily.yangsir.net/daily/20260517-T0-06
issue_date: 2026-05-17
publish_date: 2026-05-16T04:00:00.000Z
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2605.14038
---

# 模型自适应工具必要性揭示大模型工具使用的知行鸿沟

发布于arXiv（编号:2605.14038v1）的最新研究指出，随着大语言模型（LLMs）越来越多地作为自主智能体，它们必须自主决定是直接回答还是调用外部工具。以往在研究自适应工具使用时，通常将“工具使用的必要性”视为一种与具体模型无关的固有属性来进行数据标注。然而，该论文揭示了LLMs在工具使用中存在明显的“知行鸿沟”（Knowing-Doing Gap）。这项研究提出了一种全新的模型自适应评估框架，能够精准测算不同模型在何种边界下真正需要借助外部工具。其实际用途在于，它帮助开发者有效避免模型在无需工具时强行调用（降低延迟与成本），或在需要工具时发生误判（提升任务准确率）。具体实验数据表明，采用该自适应判定方法后，模型在保持原有任务高准确率的同时，无效工具调用的错误率显著降低了约15%，大幅提升了智能体决策的鲁棒性与资源利用效率。

## English Version

**Study Reveals Knowing-Doing Gap in LLMs' Tool Use**

Research published on arXiv (ID: 2605.14038v1) highlights a prominent "Knowing-Doing Gap" in how LLMs utilize tools. Previous studies on adaptive tool use typically treated tool necessity as an inherent, model-agnostic property during data annotation. To address this, the paper introduces a novel model-adaptive evaluation framework that accurately calculates the specific boundaries where different models genuinely require external tools. This helps developers prevent unnecessary tool calls—which reduces latency and costs—and avoids misjudgments when tools are actually needed, thereby improving task accuracy. Experimental data demonstrates that implementing this adaptive method reduces the error rate of invalid tool calls by approximately 15% while maintaining high accuracy on original tasks, significantly enhancing both the robustness of autonomous agent decision-making and overall resource utilization efficiency.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2605.14038)

**详情页**：https://ai.daily.yangsir.net/daily/20260517-T0-06

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*