document-vqa
DocumentVQADocumentVQA
Neural Metrics · Asking documents questions, and grading the answers.
Document visual question answering: given a page image and a natural-language question, produce the answer. It measures whether a model genuinely read the layout or merely pattern-matched the text.
We use it for: evaluating question-answering over extracted documents - catching models that read text but misread structure.
Attribution
This is an unmodified fork of… See the full description on the dataset page: https://huggingface.co/datasets/NeuralMetrics/DocumentVQA.DocumentVQADocumentVQADocumentVQA-smallDocumentVQA_subset