This dataset evaluates candidate answers for various question-answering (QA) tasks across multiple datasets such as Jeopardy!, hotpotQA, nq-open, narrativeQA, and BIOMRC, etc. See details in paper. It contains questions, reference answers (ground truth), model-generated candidate answers, and human judgments indicating whether the candidate answers are correct.

	Dataset Details

Column Type Description

question string The question asked… See the full description on the dataset page: https://huggingface.co/datasets/zli12321/pedants_qa_evaluation_bench.