HEAL-MedVQA

HEAL-MedVQA: Hallucination Evaluation and Analysis for Medical Visual Question Answering This dataset is part of the MM-Hallu benchmark collection for evaluating multimodal hallucination in vision-language models. Dataset Description HEAL-MedVQA evaluates medical VQA hallucination on chest X-ray images. It includes both closed-ended (yes/no) and open-ended questions about radiological findings, with annotated anatomical regions. Statistics Split Samples… See the full description on the dataset page:

Source: Hugging Face Hub (MM-Hallu/HEAL-MedVQA). Metadata imported from the dataset’s Hub tags.