| title | Can Generalist Agents Automate Data Curation? |
|---|---|
| authors | Feiyang Kang, Hanze Li, Adam Nguyen, Mahavir Dabas, Jiaqi W. Ma, Frederic Sala, Dawn Song, Ruoxi Jia |
| date | 2026-06-02 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2606.04261 |
| upvotes | 3 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Curating training data is among the most consequential yet labor-intensive parts of modern AI development: practitioners iteratively propose, implement, evaluate, and revise data policies against noisy benchmark feedback. We ask whether generalist coding agents can automate this data-curation loop. We introduce *Curation-Bench*, an agent-centric benchmark that fixes the model, training recipe, and evaluation suite while giving agents command-line access to inspect data, implement policies, submit them to a fixed training/evaluation pipeline, and revise. In a vision-language instruction-tuning instantiation, out-of-the-box agents reach strong published data-selection baselines within ten iterations. However, trajectory analysis reveals a persistent *execution-research gap*: agents mainly tune local policy variants rather than explore new policy families, even when given strategy guides and paper references. Scaffolds requiring each iteration to cite, instantiate, and adapt a prior method shift agents toward method-guided exploration. The scaffolded agent autonomously composes -- without human design input -- a data-selection policy that outperforms strong published baselines at one-tenth their data budget. Overall, current agents can run the curation loop, but reliable data research requires scaffolded method adaptation, not open-ended prompting alone. Code and benchmark are open-sourced. |
cite
@misc{kang2026can,
title={Can Generalist Agents Automate Data Curation?},
author={Feiyang Kang and Hanze Li and Adam Nguyen and Mahavir Dabas and Jiaqi W. Ma and Frederic Sala and Dawn Song and Ruoxi Jia},
year={2026},
eprint={2606.04261},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2606.04261}
}