让大模型学会正确使用工具，而非只是决定何时用。
Learning How to Use Tools, Not Just When: Pattern-Aware Tool-Integrated Reasoning
- 区分计算与算法两类工具使用模式，构建代码能力基础
- 在MATH500上代码准确率提升至70.5%，AIME24达50.0%
- 适合需要精准工具调用的数学推理任务
工具集成推理（TIR）已成为提升大模型解决复杂问题能力的关键方法。以往研究多关注何时调用工具，却忽视了如何正确使用。我们识别出两种常见模式：直接计算的计算器模式和将问题编码为程序的算法模式。若模式选择不当，即使推理过程正确也会失败。为此提出两阶段框架：先从两类模式中建立代码能力，再对齐模式选择与教师偏好。在多个挑战性数学数据集上，该模式感知方法显著提升代码使用率与准确率，例如在MATH500上Code@1从64.0%提升至70.5%，在AIME24上从26.7%升至50.0%。结果表明模式感知策略对工具集成推理极为有效。
原文摘要 · Abstract (English)
Tool-integrated reasoning (TIR) has become a key approach for improving large reasoning models (LRMs) on complex problems. Prior work has mainly studied when to invoke tools, while overlooking how tools are applied. We identify two common patterns: a calculator pattern that uses code for direct computation, and an algorithmic pattern that encodes problems as programs. Misaligned choices often cause failures even when reasoning is sound. We propose a two-stage framework that first builds code competence from both patterns and then aligns pattern selection with teacher preferences. Across challenging math datasets, our pattern-aware method substantially improves both code usage and accuracy, for instance raising Code@1 on MATH500 from 64.0% to 70.5% and on AIME24 from 26.7% to 50.0%. These gains highlight the effectiveness of a pattern-aware approach for tool-integrated reasoning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。