zanellac/mediqal-partages

MediQAl mcqu (PARTAGES evaluation mirror)

Used for

Used in the Evaluation Benchmark for LLMs for the PARTAGES project, to evaluate the PARTAGES language models against general-purpose Qwen3 baselines on French medical multiple-choice QA.

Citation for the PARTAGES models:

@article{mannion2026biomedical,
  title={Is Biomedical Specialization Still Worth It? Insights from Domain-Adaptive Language Modelling with a New French Health Corpus},
  author={Mannion, Aidan and Macaire, C\'ecile and Violle, Armand and Ohayon, St\'ephane and Tannier, Xavier and Schwab, Didier and Goeuriot, Lorraine and Portet, Fran\c{c}ois},
  year={2026},
  eprint={2604.06903},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2604.06903}
}

Description

MediQAl is a French medical question-answering dataset built from French medical board examination questions (ECNi/EDN), each grounded in a clinical case vignette. This mirror is the mcqu (multiple-choice, unique answer) subset: questions with exactly one correct answer among five choices (A-E).

Source

ANR-MALADES/MediQAl on the Hugging Face Hub, CC-BY-4.0.

Split

Official train/validation/test split, copied verbatim from the source dataset.

Split	Size
train	10,113
validation	2,561
test	4,343

Citation

Please also cite the original dataset:

@article{Bazoge2026,
  author = {Bazoge, Adrien},
  title = {MediQAl: A French Medical Question Answering Dataset for Knowledge and Reasoning Evaluation},
  journal = {Scientific Data},
  year = {2026},
  volume = {13},
  pages = {356},
  doi = {10.1038/s41597-026-06680-y}
}