Datasets:

DSGym
/

DSGym-Tasks

license: mit
task_categories:
  - question-answering
  - text-generation
tags:
  - data-science
  - data-analysis
  - bioinformatics
  - statistical-analysis
  - machine-learning
language:
  - en
pretty_name: DSGym-Tasks

DSGym

DSGym is a unified benchmark and execution framework for evaluating and training data science agents. It provides standardized, executable tasks that require agents to plan, implement, and validate data analyses through interaction with real data files in isolated environments.

Attribution

If you find our work useful, please cite our paper:

@misc{nie2026dsgym,
      title={DSGym: A Holistic Framework for Evaluating and Training Data Science Agents}, 
      author={Fan Nie and Junlin Wang and Harper Hua and Federico Bianchi and Yongchan Kwon and Zhenting Qi and Owen Queen and Shang Zhu and James Zou},
      year={2026},
      eprint={2601.16344},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2601.16344}, 
}