arXiv cs.AI· Bingxiang He, Pengqi Sun, Shihao Peng, Shaoxuan Li, Huiyu Xiao, Hongjian Shi, Zixuan Fu, Ziqing Qiao, Yuxin Zuo, Cheng Qian, Xinqiang Yu, Jiayi Li, Xiusi Chen, Yizhen Song, Xiangyang Wang, Iofai Lei, Yudong Wang, Ning Ding, Zhiyuan Liu, Xiaozhi Wang, Chaojun Xiao·· 2 天前AI 评分34
AutoSFT：面向监督微调自动化数据研究的多智能体系统
AutoSFT: An Agentic System for Automated Data Research in Supervised Fine-Tuning
AI 导读
AutoSFT 是一个多智能体系统，用于自动化监督微调（SFT）中的数据研究。它通过元智能体将执行与分析分离，从独立角度收集提案，并调用合成子智能体补充数据池无法覆盖的能力。相比单智能体仅能调整类别比例，其分析子智能体将最佳 IFEval 提升 +6.5，合成再增 +3.4 IFEval 和 +2.9 IFBench，代码任务上分析带来 +3.7 MBPP+。
正文 · 原文
Authors:Bingxiang He, Pengqi Sun, Shihao Peng, Shaoxuan Li, Huiyu Xiao, Hongjian Shi, Zixuan Fu, Ziqing Qiao, Yuxin Zuo, Cheng Qian, Xinqiang Yu, Jiayi Li, Xiusi Chen, Yizhen Song, Xiangyang Wang, Iofai Lei, Yudong Wang, Ning Ding, Zhiyuan Liu, Xiaozhi Wang, Chaojun Xiao
Abstract:Language model agents increasingly automate post-training, but when given the full training stack they converge on supervised fine-tuning and spend most effort revising training data. Existing systems give the agent one kind of control, selecting rows from a fixed pool or calling a generator, whereas a human data engineer adjusts proportions, prunes, repairs, replaces and synthesizes the data. We design a data action space for these changes and ask whether an automated system can make the full range rather than one of them, and what structure that requires. We first run this loop with a single agent, and it reaches only one action, adjusting category proportions, while its context fills with execution output and is discarded mid-search. We therefore introduce \textbf{AutoSFT}, a multi-agent system where a meta agent separates execution from analysis, gathers proposals from independent angles, and calls a synthesis subagent for capabilities the pool cannot cover, with the model, the trainer, and the budget fixed within each comparison. Adding each component in turn helps: the executor halves context loss, the analysis subagents lift the best IFEval by $+6.5$, and synthesis adds $+3.4$ IFEval and $+2.9$ IFBench. The same loop transfers to code, where analysis adds $+3.7$ MBPP+ over a single agent that does not improve. Every result comes from runs that retrain from the base model each round, so each component's effect is individually attributable. Our findings suggest that automating SFT data research depends more on covering diverse data actions than on a stronger single action.
| Comments: | 27 pages, 16 figures |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.22194 [cs.AI] |
| (or arXiv:2610.22194v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22194 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Bingxiang He [view email]
[v1]
Thu, 15 Oct 2026 16:44:34 UTC (2,543 KB)
来源：arXiv cs.AI · arxiv.org