CodeTool: Enhancing Programmatic Tool Invocation of LLMs via Process Supervision
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918115913236480 |
|---|---|
| author | Lu, Yifei Ye, Fanghua Li, Jian Gao, Qiang Liu, Cheng Luo, Haibo Du, Nan Li, Xiaolong Ren, Feiliang |
| author_facet | Lu, Yifei Ye, Fanghua Li, Jian Gao, Qiang Liu, Cheng Luo, Haibo Du, Nan Li, Xiaolong Ren, Feiliang |
| contents | Tool invocation significantly enhances the capabilities of Large Language Models (LLMs), yet challenges persist, particularly in complex task scenarios. Current methods, such as instruction-enhanced reasoning and supervised fine-tuning, often result in unnecessarily long reasoning paths and face difficulties in verifying the correctness of intermediate steps. In this paper, we propose CodeTool, a novel framework for stepwise code generation that improves LLM tool invocation by leveraging the concise and easily verifiable nature of code. CodeTool incorporates two distinct process rewards: the On-the-spot Reward, which provides immediate feedback on the accuracy of each tool invocation, and the Latent Reward, which assesses the contribution of each step toward overall task completion. By maximizing the cumulative reward of the On-the-spot and Latend Rewards at each step, LLMs are guided to follow efficient and accurate reasoning paths. Extensive experiments on StableToolBench and RestBench-TMDB demonstrate the superiority of CodeTool over existing approaches. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_20840 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | CodeTool: Enhancing Programmatic Tool Invocation of LLMs via Process Supervision Lu, Yifei Ye, Fanghua Li, Jian Gao, Qiang Liu, Cheng Luo, Haibo Du, Nan Li, Xiaolong Ren, Feiliang Artificial Intelligence Computation and Language Machine Learning Software Engineering Tool invocation significantly enhances the capabilities of Large Language Models (LLMs), yet challenges persist, particularly in complex task scenarios. Current methods, such as instruction-enhanced reasoning and supervised fine-tuning, often result in unnecessarily long reasoning paths and face difficulties in verifying the correctness of intermediate steps. In this paper, we propose CodeTool, a novel framework for stepwise code generation that improves LLM tool invocation by leveraging the concise and easily verifiable nature of code. CodeTool incorporates two distinct process rewards: the On-the-spot Reward, which provides immediate feedback on the accuracy of each tool invocation, and the Latent Reward, which assesses the contribution of each step toward overall task completion. By maximizing the cumulative reward of the On-the-spot and Latend Rewards at each step, LLMs are guided to follow efficient and accurate reasoning paths. Extensive experiments on StableToolBench and RestBench-TMDB demonstrate the superiority of CodeTool over existing approaches. |
| title | CodeTool: Enhancing Programmatic Tool Invocation of LLMs via Process Supervision |
| topic | Artificial Intelligence Computation and Language Machine Learning Software Engineering |
| url | https://arxiv.org/abs/2503.20840 |