pedsQA is a re-partitioned and retrieval-augmented version of the
PediatricsMQA dataset,
prepared for fine-tuning and controlled evaluation of vision-language models on
pediatric medical visual question answering (VQA).
The original dataset is a multiple-choice VQA benchmark spanning radiology,
dermatology, ophthalmology, orthopaedics, and other pediatric subspecialties.
This release introduces stratified splits by imaging modality and anatomical region,
and offline RAG… See the full description on the dataset page:
https://huggingface.co/datasets/mahmoudg0/pedsQA.
