Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908954195394560 |
|---|---|
| author | Xu, Siyuan Li, Shiyang Liu, Xin Liu, Tianyi Li, Yixiao Shi, Zhan Zhang, Zixuan Wang, Zilong Yin, Qingyu Chen, Jianshu Zhao, Tuo Yin, Bing |
| author_facet | Xu, Siyuan Li, Shiyang Liu, Xin Liu, Tianyi Li, Yixiao Shi, Zhan Zhang, Zixuan Wang, Zilong Yin, Qingyu Chen, Jianshu Zhao, Tuo Yin, Bing |
| contents | Existing synthetic tool-use corpora are primarily designed for offline supervised fine-tuning, yet reinforcement learning (RL) requires executable environments that support reward-checkable online rollouts. We propose COVERT, a two-stage pipeline that first generates reliable base tool-use trajectories through self-evolving synthesis with multi-level validation, and then applies oracle-preserving augmentations that systematically increase environmental complexity. These augmentations introduce distractor tools, indirect or ambiguous user queries, and noisy, multi-format, or erroneous tool outputs, while strictly preserving oracle tool calls and final answers as ground truth. This design enables automatic reward computation via reference matching for standard cases and lightweight judge-assisted verification for special behaviors such as error detection, supporting RL optimization of tool-calling policies. On Qwen2.5-Instruct-14B, COVERT-RL improves overall accuracy on BFCL v3 from 56.5 to 59.9 and on ACEBench from 53.0 to 59.3, with minimal regressions on general-ability benchmarks; when stacked on SFT, it further reaches 62.1 and 61.8, confirming additive gains. These results suggest that oracle-preserving synthetic environments offer a practical RL refinement stage, complementary to SFT, for improving tool-use robustness under ambiguity and unreliable tool feedback. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_09813 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement Learning Xu, Siyuan Li, Shiyang Liu, Xin Liu, Tianyi Li, Yixiao Shi, Zhan Zhang, Zixuan Wang, Zilong Yin, Qingyu Chen, Jianshu Zhao, Tuo Yin, Bing Artificial Intelligence Existing synthetic tool-use corpora are primarily designed for offline supervised fine-tuning, yet reinforcement learning (RL) requires executable environments that support reward-checkable online rollouts. We propose COVERT, a two-stage pipeline that first generates reliable base tool-use trajectories through self-evolving synthesis with multi-level validation, and then applies oracle-preserving augmentations that systematically increase environmental complexity. These augmentations introduce distractor tools, indirect or ambiguous user queries, and noisy, multi-format, or erroneous tool outputs, while strictly preserving oracle tool calls and final answers as ground truth. This design enables automatic reward computation via reference matching for standard cases and lightweight judge-assisted verification for special behaviors such as error detection, supporting RL optimization of tool-calling policies. On Qwen2.5-Instruct-14B, COVERT-RL improves overall accuracy on BFCL v3 from 56.5 to 59.9 and on ACEBench from 53.0 to 59.3, with minimal regressions on general-ability benchmarks; when stacked on SFT, it further reaches 62.1 and 61.8, confirming additive gains. These results suggest that oracle-preserving synthetic environments offer a practical RL refinement stage, complementary to SFT, for improving tool-use robustness under ambiguity and unreliable tool feedback. |
| title | Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement Learning |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2604.09813 |