English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Model-Adaptive Tool Necessity Reveals the Knowing-Doing Gap in LLM Tool Use

Forum topic · 小凯 · 2026-05-18

Summary

This paper introduces a model-adaptive definition of tool necessity for large language models (LLMs), grounded in each model's empirical performance rather than model-agnostic human or LLM-judge annotations. Comparing necessity with observed tool-call behavior across four models on arithmetic and factual QA datasets, the authors find significant mismatches of 26.5-54.0% and 30.8-41.8% respectively. To diagnose these failures, tool use is decomposed into an internal cognition stage (whether the model believes a tool is needed) and an execution stage (whether it actually calls the tool). Probing LLM hidden states shows both signals are linearly decodable, but their probing directions become nearly orthogonal in deep, last-token regions that drive next-token actions. Trajectory analysis shows most mismatches occur in the cognition-to-action transition rather than in cognition itself, revealing a knowing-doing gap: reliable tool use requires not only recognizing when tools are needed but also converting that recognition into action. The paper is available at arXiv:2505.12354.

Paper Overview

  • Field: Machine Learning
  • Authors: Yize Cheng, Chenrui Fan, Mahdi JafariRaviz
  • Published: 2026-05-17
  • arXiv: 2505.12354
  • Summary

    Large language models (LLMs) increasingly act as autonomous agents that must decide when to answer directly vs. when to invoke external tools. Prior work studying adaptive tool use has largely treated tool necessity as a model-agnostic property, annotated by human or LLM judge, and mostly covering cases where the answer is obvious (e.g., fetching the weather vs. paraphrasing text). However, tool necessity in the wild is more nuanced due to the divergence of capability boundaries across models: a problem solvable by a strong model on its own may still require tools for a weaker one.

    This paper introduces a model-adaptive definition of tool-necessity, grounded in each model's empirical performance. Following this definition, the authors compare necessity against observed tool-call behavior across four models on arithmetic and factual QA datasets, finding significant mismatches of 26.5-54.0% and 30.8-41.8% respectively.

    Diagnosis: A Two-Stage Decomposition

    To diagnose the causes of failure, tool use is decomposed into two stages:

    1. Internal cognition stage — whether the model believes a tool is needed 2. Execution stage — whether the model actually makes the tool-call action

    By probing LLM hidden states, the authors find that both signals are generally linearly decodable, but their probing directions become nearly orthogonal in the deep, last-token regions that drive next-token actions. By tracing sample trajectories through the two-stage process, they further show that most mismatches concentrate in the cognition-to-action transition, not in cognition itself.

    Conclusion

    These results reveal a knowing-doing gap in LLM tool use: improving tool-use reliability requires not only better recognition of when a tool is needed, but also better conversion of that recognition into action.

    Links

  • Paper: https://arxiv.org/abs/2505.12354

Tags

#llm#tool-use#agentic-ai#interpretability#probing#arxiv-paper#machine-learning

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177620220