Original dataset by Abacus AI.
This benchmark is a modified wikiQA benchmark. The dataset is composed of Wikipedia articles (of 2-16 thousand tokens) and corresponding questions. We modify the… See the full description on the dataset page:
https://huggingface.co/datasets/normalcomputing/wikiqa-counterfactual.