AutoSFT:用於監督式微調自動化數據研究的智能體系統
AutoSFT: An Agentic System for Automated Data Research in Supervised Fine-Tuning
AutoSFT 以多智能體協作自動執行監督式微調數據研究中的多種數據操作,並透過分離執行、分析和合成改善評測表現。分析子智能體令最佳 IFEval 提升 +6.5,合成子智能體再令 IFEval 提升 +3.4、IFBench 提升 +2.9。同一流程應用於程式碼任務時,分析令 MBPP+ 較未能提升的單智能體高 +3.7。
Published on Oct 15
Authors:
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
Abstract
AutoSFT uses multi-agent collaboration to automate diverse supervised fine-tuning data actions, improving evaluation scores over single-agent systems by separating execution, analysis, and synthesis.
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 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.
View arXiv page View PDF GitHub Add to collection
Get this paper in your agent:
hf papers read 2610.22194
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash
Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2610.22194 in a model README.md to link it from this page.
Datasets citing this paper 0
No dataset linking this paper
Cite arxiv.org/abs/2610.22194 in a dataset README.md to link it from this page.
Spaces citing this paper 0
No Space linking this paper
Cite arxiv.org/abs/2610.22194 in a Space README.md to link it from this page.
Collections including this paper 0
No Collection including this paper
Add this paper to a collection to link it from this page.
來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co