OpenMedText Dataset

A comprehensive biomedical text corpus consisting of MDPI journal articles and open-source medical textbooks for language model training and research.

Dataset Summary

OpenMedText is a large-scale biomedical text dataset that includes:

• Med-MDPI: 121,489 biomedical journal articles across 37 categories from MDPI journals
• Med-Textbooks: 29 open-source medical textbooks covering various medical disciplines

Folder Structure

OpenMedText/
├── Med-MDPI/               # All under CC BY 4.0 license
│   ├── Allergies/
│   ├── Antibiotics/
│   ├── Antibodies/
│   └── ...                 # More journal categories
│
└── Med-Textbooks/          # Organized by license type
    ├── CC-BY/              # CC BY 4.0 and CC BY licensed books
    ├── CC-BY-SA/           # CC BY-SA 4.0, 3.0, 2.0 books
    ├── CC-BY-NC/           # CC BY-NC 4.0, 3.0 books
    └── CC-BY-NC-SA/        # CC BY-NC-SA 4.0, 3.0, 2.0 books

Licenses

This dataset includes materials with different licenses:

1. Med-MDPI: All journal articles are under CC BY 4.0 license.

2. Med-Textbooks: Textbooks have various licenses:

   • CC BY: Creative Commons Attribution
   • CC BY-SA: Creative Commons Attribution-ShareAlike
   • CC BY-NC: Creative Commons Attribution-NonCommercial
   • CC BY-NC-SA: Creative Commons Attribution-NonCommercial-ShareAlike

Please respect the license of each component when using this dataset.

Intended Uses

The OpenMedText dataset is designed for:

• Training and fine-tuning language models for biomedical applications

Data Selection Criteria

The dataset was curated to provide broad coverage of biomedical subjects while ensuring all content is openly accessible. Textbooks were specifically selected to provide structured medical knowledge organized for educational purposes.

Citation

If you use this dataset in your research, please cite:

@inproceedings{
    choi2025teaching,
    title={Teaching {LLM}s How To Learn with Contextual Fine-Tuning},
    author={Younwoo Choi and Muhammad Adil Asif and Ziwen Han and John Willes and Rahul Krishnan},
    booktitle={The Thirteenth International Conference on Learning Representations},
    year={2025},
    url={https://arxiv.org/abs/2503.09032}
}