This is Belebele dataset naively adapted for a QA task. project. Of all the lines in the original dataset, only those where selected where the correct answer is exactly the same span in context. Nevertheless, the questions themselves remain formulated for the original tasks of the dataset.
