| --- |
| task_categories: |
| - question-answering |
| language: |
| - en |
| tags: |
| - biology |
| - bioinformatics |
| - data-analysis |
| - data-science |
| - agent |
| license: mit |
| pretty_name: DSBio |
| --- |
| ## DSBio: Scientific Analysis Tasks |
|
|
| DSBio is a suite of **90 expert-derived bioinformatics tasks** constructed from |
| peer-reviewed academic publications and public scientific datasets. |
|
|
| These tasks are designed to evaluate whether agents can perform |
| domain-grounded scientific analysis, including: |
|
|
| - Interpreting high-dimensional biological data (e.g., single-cell and spatial omics) |
| - Understanding domain-specific terminology and conventions |
| - Executing multi-step analytical workflows with specialized libraries |
|
|
| ## Attribution |
|
|
| If you use DSBio in academic work, please cite our paper: |
| ``` |
| @misc{nie2026dsgymholisticframeworkevaluating, |
| 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}, |
| } |
| ``` |
