CancerGUIDE Synthetic Patient is a synthetic oncology patient dataset generated based on GPT-4.1, released by Microsoft Research in 2025. The related paper is「CancerGUIDE: Cancer Guideline Understanding via Internal Disagreement Estimation, which aims to provide privacy-preserving, controllable benchmark data for evaluating and training large language models on clinical guideline adherence and reasoning tasks.

This dataset contains synthetic oncology patient records and corresponding treatment recommendations, generated in both structured and unstructured formats. The structured data includes 165 records, while the unstructured data includes 151 records. These data were generated by GPT-4.1 to support research on clinical reasoning and recommendation tasks for large models, while avoiding the privacy risks associated with using real patient data.

Dataset Composition

The dataset contains two configuration files (configs), corresponding to different generation methods:

• synthetic_structured: Contains 165 records, generated using a structured (tabular prompt template) method.
• synthetic_unstructured: Contains 151 records, generated using an unstructured (free-text narrative) method.

Each record (in JSON format) contains the following fields:

• patient_id: Unique patient identifier
• patient_note: Textual description of the synthetic patient's medical history
• label: Treatment regimen recommended by the model

Citation

@article{CancerGUIDE,
  title = {CancerGUIDE: Cancer Guideline Understanding via Internal Disagreement Estimation},
  author = {Unell,  Alyssa and Codella,  Noel C. F. and Preston,  Sam and Argaw,  Peniel and Yim,  Wen-wai and Gero,  Zelalem and Wong,  Cliff and Jena,  Rajesh and Horvitz,  Eric and Hall,  Amanda K. and Zhong,  Ruican Rachel and Li,  Jiachen and Jain,  Shrey and Wei,  Mu and Lungren,  Matthew and Poon,  Hoifung},
  url = {https://doi.org/10.48550/arxiv.2509.07325},
  publisher = {arXiv},
  year = {2025},
}