EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917508425973760 |
|---|---|
| author | Xu, Minrui Wang, Zilin DENG, Mengyi Li, Zhiwei Yang, Zhicheng Zhu, Xiao Liu, Yinhong Zhu, Boyu Huang, Baiyu Chen, Chao Deng, Heyuan Mi, Fei Shang, Lifeng Zeng, Xingshan Guo, Zhijiang |
| author_facet | Xu, Minrui Wang, Zilin DENG, Mengyi Li, Zhiwei Yang, Zhicheng Zhu, Xiao Liu, Yinhong Zhu, Boyu Huang, Baiyu Chen, Chao Deng, Heyuan Mi, Fei Shang, Lifeng Zeng, Xingshan Guo, Zhijiang |
| contents | Equipping LLMs with tool-use capabilities via Agentic Reinforcement Learning (Agentic RL) is bottlenecked by two challenges: the lack of scalable, robust execution environments and the scarcity of realistic training data that captures implicit human reasoning. Existing approaches depend on costly real-world APIs, hallucination-prone LLM simulators, or synthetic environments that are often single-turn or depend on pre-collected documents. Moreover, synthetic trajectories are frequently over-specified, resembling instruction sequences rather than natural human intents, reducing their effectiveness for RL training. We introduce EnvFactory, a fully automated framework that addresses both challenges. EnvFactory autonomously explores and verifies stateful, executable tool environments from authentic resources, and synthesizes natural multi-turn trajectories through topology-aware sampling and calibrated refinement, producing grounded queries with implicit intents. Using only 85 verified environments across 7 domains, EnvFactory generates 2,575 SFT and RL trajectories. Despite using significantly fewer environments than prior work, which are often 5 times more, EnvFactory achieves superior training efficiency and downstream performance, improving Qwen3-series models by up to +15% on BFCLv3, +8.6% on MCP-Atlas, and +6% on conversational benchmarks including $τ^2$-Bench and VitaBench. By fully automating both environment construction and trajectory synthesis, EnvFactory provides a scalable, extensible, and robust foundation for Agentic RL. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_18703 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL Xu, Minrui Wang, Zilin DENG, Mengyi Li, Zhiwei Yang, Zhicheng Zhu, Xiao Liu, Yinhong Zhu, Boyu Huang, Baiyu Chen, Chao Deng, Heyuan Mi, Fei Shang, Lifeng Zeng, Xingshan Guo, Zhijiang Computation and Language Machine Learning Equipping LLMs with tool-use capabilities via Agentic Reinforcement Learning (Agentic RL) is bottlenecked by two challenges: the lack of scalable, robust execution environments and the scarcity of realistic training data that captures implicit human reasoning. Existing approaches depend on costly real-world APIs, hallucination-prone LLM simulators, or synthetic environments that are often single-turn or depend on pre-collected documents. Moreover, synthetic trajectories are frequently over-specified, resembling instruction sequences rather than natural human intents, reducing their effectiveness for RL training. We introduce EnvFactory, a fully automated framework that addresses both challenges. EnvFactory autonomously explores and verifies stateful, executable tool environments from authentic resources, and synthesizes natural multi-turn trajectories through topology-aware sampling and calibrated refinement, producing grounded queries with implicit intents. Using only 85 verified environments across 7 domains, EnvFactory generates 2,575 SFT and RL trajectories. Despite using significantly fewer environments than prior work, which are often 5 times more, EnvFactory achieves superior training efficiency and downstream performance, improving Qwen3-series models by up to +15% on BFCLv3, +8.6% on MCP-Atlas, and +6% on conversational benchmarks including $τ^2$-Bench and VitaBench. By fully automating both environment construction and trajectory synthesis, EnvFactory provides a scalable, extensible, and robust foundation for Agentic RL. |
| title | EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2605.18703 |