arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2510.24645cs.AI
收藏

FunReason-MT技术报告:面向真实世界多轮工具使用的先进数据合成方案

FunReason-MT Technical Report: Advanced Data Synthesis Solution for Real-world Multi-Turn Tool-use

Zhejiang University(浙江大学);City University of Hong Kong(香港城市大学)
Zengzhuang Xu, Bingguang Hao, Zechuan Wang, Yuntao Wen, Xinyi Xu, Yang Liu, Long Chen, Dong Wang, Maolin Wang, Tong Zhao, Yicheng Chen, Cunyin Peng, Jinjie Gu, … 展开作者

Zengzhuang Xu, Bingguang Hao, Zechuan Wang, Yuntao Wen, Xinyi Xu, Yang Liu, Long Chen, Dong Wang, Maolin Wang, Tong Zhao, Yicheng Chen, Cunyin Peng, Jinjie Gu, Leilei Gan, Xiangyu Zhao, Chenyi Zhuang, Shi Gu

AI总结:

针对现有多轮工具调用数据合成方法存在的定向生成难、难查询构建难、多轮逻辑依赖难等问题,提出FunReason-MT数据合成框架,训练的4B模型在BFCL基准同规模模型中达SOTA。

摘要:

函数调用(FC)使大语言模型(LLM)和自主智能体能够对接外部工具,这是解决复杂真实世界问题的关键能力。随着该能力在先进AI系统中的地位愈发核心,开发和优化该能力所需的高质量多轮训练数据的重要性不言而喻。现有数据合成方法(如随机环境采样、多智能体角色扮演)不足以在真实环境中生成高质量数据,实际挑战集中在三方面:定向数据合成、难查询构建、多轮逻辑依赖。为解决这些结构性缺陷,我们提出FunReason-MT,这一面向真实世界多轮工具使用的新型数据合成框架。FunReason-MT通过三项技术突破多轮FC数据的复杂度壁垒:1. 环境-API图交互,采集带定向工具的多样化高质量轨迹;2. 先进工具查询合成,简化难查询构建流程;3. 引导迭代链,生成精细的思维链(CoT)。在Berkeley Function-Calling Leaderboard(BFCLv3)上的评估验证了框架的效果:基于FunReason-MT生成数据训练的4B模型在同参数量级模型中达到当前最优性能。模型在BFCLv4上的性能进一步提升,证实FunReason-MT可为智能体学习提供可靠、鲁棒的数据来源。

英文摘要:

Function calling (FC) empowers large language models (LLMs) and autonomous agents to interface with external tools, a critical capability for solving complex, real-world problems. As this ability becomes increasingly central to advanced AI systems, the need for high-quality, multi-turn training data to develop and refine it cannot be overstated. Existing data synthesis methods, such as random environment sampling or multi-agent role-playing, are not powerful enough to generate high-quality data in real-world environments. Practical challenges come in three folds: targeted data synthesis, hard query construction, and multi-turn logical dependency. To address these structural deficiencies, we present FunReason-MT, a novel data synthesis framework for real-world multi-turn tool use. FunReason-MT resolves the complexity barrier in multi-turn FC data by employing 1) Environment-API Graph Interactions to gather varied high-quality trajectories with targeted tool, 2) Advanced Tool-Query Synthesis to simplify hard query construction, and 3) Guided Iterative Chain for sophisticated CoT generation. Evaluations on Berkeley Function-Calling Leaderboard (BFCLv3) demonstrate the power of our framework: a 4B model built upon FunReason-MT generated data achieves state-of-the-art performance among comparable-sized models. Further performance improvements on BFCLv4 confirm that FunReason-MT provides a reliable and robust source for agentic learning.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑