Learning How to Use Tools, Not Just When: Pattern-Aware Tool-Integrated Reasoning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917191203422208 |
|---|---|
| author | Xu, Ningning Jiang, Yuxuan Dipta, Shubhashis Roy Zhang, Hengyuan |
| author_facet | Xu, Ningning Jiang, Yuxuan Dipta, Shubhashis Roy Zhang, Hengyuan |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_23292 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Learning How to Use Tools, Not Just When: Pattern-Aware Tool-Integrated Reasoning Xu, Ningning Jiang, Yuxuan Dipta, Shubhashis Roy Zhang, Hengyuan Artificial Intelligence Computation and Language 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. |
| title | Learning How to Use Tools, Not Just When: Pattern-Aware Tool-Integrated Reasoning |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2509.23292 |