This repository contains the Med4-VQA dataset, a multimodal medical visual question answering dataset designed for joint reasoning and spatial grounding.
Med4-Grounded VQA is constructed to support research in grounded medical VQA, where models are required to not only answer clinical questions but also localize relevant visual evidence.
The dataset covers four imaging modalities:
Medical… See the full description on the dataset page:
https://huggingface.co/datasets/neuripsqgr2026/Med4VQA.
