Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Chai, Jiajun, Yin, Guojun, Xu, Zekun, Yue, Chuhuai, Jia, Yi, Xia, Siyu, Wang, Xiaohan, Jiang, Jiwen, Li, Xiaoguang, Dong, Chengqi, He, Hang, Lin, Wei
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2509.06980
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915484432072704
author Chai, Jiajun
Yin, Guojun
Xu, Zekun
Yue, Chuhuai
Jia, Yi
Xia, Siyu
Wang, Xiaohan
Jiang, Jiwen
Li, Xiaoguang
Dong, Chengqi
He, Hang
Lin, Wei
author_facet Chai, Jiajun
Yin, Guojun
Xu, Zekun
Yue, Chuhuai
Jia, Yi
Xia, Siyu
Wang, Xiaohan
Jiang, Jiwen
Li, Xiaoguang
Dong, Chengqi
He, Hang
Lin, Wei
contents Large language models excel at basic reasoning but struggle with tasks that require interaction with external tools. We present RLFactory, a plug-and-play reinforcement learning post-training framework for multi-round tool use. RLFactory tackles (i) tool-call stability and adaptability amid tool heterogeneity and interface issues via an asyncio-based asynchronous caller and a decoupled tool/training architecture, and (ii) diverse evaluation needs via a reward layer supporting rule-based, model-judgment, and tool-verification signals. It reconstructs the MDP by introducing observation markers from tool feedback, closing the loop among model, tools, and environment, and implements a generate-parse-invoke-update workflow for dynamic policy optimization. On Search-R1 with Qwen3-4B, RLFactory achieves a 0.486 test score on the Natural Questions (NQ) dataset, surpassing larger models trained with similar techniques (e.g., Qwen2.5-7B-Instruct-GRPO at 0.473), and increases training throughput by 6.8x. RLFactory provides a low-barrier, highly adaptable framework for strengthening multi-round tool use of LLMs in real-world scenarios. Code: https://github.com/Simple-Efficient/RL-Factory.
format Preprint
id arxiv_https___arxiv_org_abs_2509_06980
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RLFactory: A Plug-and-Play Reinforcement Learning Post-Training Framework for LLM Multi-Turn Tool-Use
Chai, Jiajun
Yin, Guojun
Xu, Zekun
Yue, Chuhuai
Jia, Yi
Xia, Siyu
Wang, Xiaohan
Jiang, Jiwen
Li, Xiaoguang
Dong, Chengqi
He, Hang
Lin, Wei
Machine Learning
Artificial Intelligence
Large language models excel at basic reasoning but struggle with tasks that require interaction with external tools. We present RLFactory, a plug-and-play reinforcement learning post-training framework for multi-round tool use. RLFactory tackles (i) tool-call stability and adaptability amid tool heterogeneity and interface issues via an asyncio-based asynchronous caller and a decoupled tool/training architecture, and (ii) diverse evaluation needs via a reward layer supporting rule-based, model-judgment, and tool-verification signals. It reconstructs the MDP by introducing observation markers from tool feedback, closing the loop among model, tools, and environment, and implements a generate-parse-invoke-update workflow for dynamic policy optimization. On Search-R1 with Qwen3-4B, RLFactory achieves a 0.486 test score on the Natural Questions (NQ) dataset, surpassing larger models trained with similar techniques (e.g., Qwen2.5-7B-Instruct-GRPO at 0.473), and increases training throughput by 6.8x. RLFactory provides a low-barrier, highly adaptable framework for strengthening multi-round tool use of LLMs in real-world scenarios. Code: https://github.com/Simple-Efficient/RL-Factory.
title RLFactory: A Plug-and-Play Reinforcement Learning Post-Training Framework for LLM Multi-Turn Tool-Use
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2509.06980