Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2505.20016 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913859617423360 |
|---|---|
| author | Huang, Chengrui Gao, Shen Shi, Zhengliang Wang, Dongsheng Shang, Shuo |
| author_facet | Huang, Chengrui Gao, Shen Shi, Zhengliang Wang, Dongsheng Shang, Shuo |
| contents | Existing tool-learning methods usually rely on supervised fine-tuning, they often overlook fine-grained optimization of internal tool call details, leading to limitations in preference alignment and error discrimination. To overcome these challenges, we propose Token-level Tool-use Preference Alignment Training Framework (TTPA), a training paradigm for constructing token-level tool-use preference datasets that align LLMs with fine-grained preferences using a novel error-oriented scoring mechanism. TTPA first introduces reversed dataset construction, a method for creating high-quality, multi-turn tool-use datasets by reversing the generation flow. Additionally, we propose Token-level Preference Sampling (TPS) to capture fine-grained preferences by modeling token-level differences during generation. To address biases in scoring, we introduce the Error-oriented Scoring Mechanism (ESM), which quantifies tool-call errors and can be used as a training signal. Extensive experiments on three diverse benchmark datasets demonstrate that TTPA significantly improves tool-using performance while showing strong generalization ability across models and datasets. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_20016 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | TTPA: Token-level Tool-use Preference Alignment Training Framework with Fine-grained Evaluation Huang, Chengrui Gao, Shen Shi, Zhengliang Wang, Dongsheng Shang, Shuo Computation and Language Existing tool-learning methods usually rely on supervised fine-tuning, they often overlook fine-grained optimization of internal tool call details, leading to limitations in preference alignment and error discrimination. To overcome these challenges, we propose Token-level Tool-use Preference Alignment Training Framework (TTPA), a training paradigm for constructing token-level tool-use preference datasets that align LLMs with fine-grained preferences using a novel error-oriented scoring mechanism. TTPA first introduces reversed dataset construction, a method for creating high-quality, multi-turn tool-use datasets by reversing the generation flow. Additionally, we propose Token-level Preference Sampling (TPS) to capture fine-grained preferences by modeling token-level differences during generation. To address biases in scoring, we introduce the Error-oriented Scoring Mechanism (ESM), which quantifies tool-call errors and can be used as a training signal. Extensive experiments on three diverse benchmark datasets demonstrate that TTPA significantly improves tool-using performance while showing strong generalization ability across models and datasets. |
| title | TTPA: Token-level Tool-use Preference Alignment Training Framework with Fine-grained Evaluation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.20016 |