vidore/

tatdqa_test

This is the test set taken from the TAT-DQA dataset. TAT-DQA is a large-scale Document VQA dataset that was constructed from publicly available real-world financial reports. It focuses on rich tabular and textual content requiring numerical reasoning. Questions and answers were manually annotated by human experts in finance. Example of data (see viewer)

Unlike other 'academic' datasets, we kept the full test set as this dataset… See the full description on the dataset page: https://huggingface.co/datasets/vidore/tatdqa_test.