The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.

This is the cleaned version of Deepset's COVID-QA dataset which is described in section 3.2 of Leveraging External Knowledge Resources to Enable Domain-Specific Comprehension. While you can use either version of the dataset, we recommend using this version since we have corrected many mistakes in the original question set & also, it yields better scores, even zero-shot, as reported in our paper.

Downloads last month
17