# AutoSFT automates data research with specialized agent roles for SFT

> Original title: 🤖 What happens when you hand an agent the full SFT data-research task?

- Source: OpenBMB
- Published: 2026-11-05T14:01:04.000Z
- Bad Influence: https://badinfluence.news/items/ynuk1hjfg9oyto28t36879bgc
- Original: https://x.com/OpenBMB/status/2118342839398715877

## Summary

THUNLP, BUPT, ModelBest, UIUC and CUHK present AutoSFT, a multi-agent system that splits SFT data research into an Executor, four Analysis Subagents, a Synthesis writer with a verifier, and a meta agent choosing the next recipe. In 30 rounds, AutoSFT uses all six action families and reaches IFEval 63.33 in round 6, above the single-agent baseline's best over 50 rounds, and 67.10 by round 30. The authors say automating data research may depend more on covering diverse data actions than on a stronger single action.

## Post

🤖 What happens when you hand an agent the full SFT data-research task?

It only mixes ratios. No repaired samples, no new responses, no synthetic data — a word cloud full of "weights" and "proportions", and almost no score gain over dozens of rounds.

AutoSFT (THUNLP × BUPT × ModelBest × UIUC × CUHK) splits the data-research lifecycle into roles: a dedicated Executor, 4 Analysis Subagents (training data / failure cases / learning dynamics / research history), a Synthesis writer with a verifier, and a meta agent that picks the next recipe.

1️⃣ Single-agent baseline: 33 rounds, only Mixture; context hit 94.7%, compressed 10 times; Executor cuts compression 10→5 and shell events 58%→29%.

2️⃣ AutoSFT: all 6 action families appear across 30 rounds; IFEval round 6 = 63.33, beating the baseline's best over 50 rounds; +6.5 at the same budget; round 30 = 67.10.

3️⃣ Synthesis alone: IFEval +3.4, IFBench +2.9; MBPP+ coding +3.7.
The takeaway: automating data research may depend more on covering diverse data actions than on a stronger single action.

📄 Paper: https://arxiv.org/abs/2610.22194
💻 Code: https://github.com/Thinking-Space/AutoSFT

#AIAgent #SFT #LLM #DataCentricAI #PostTraining #RSI #AgenticAI #THUNLP #AutoSFT #FineTuning

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWXdlhGbEAAcAbR.jpg&mode=thumb&exp=1794096000&sig=0da8d8f690d3f18d)