Ai2Arc

Dataset Card for "ai2_arc" Dataset Summary A new dataset of 7,787 genuine grade-school level, multiple-choice science questions, assembled to encourage research in advanced question-answering. The dataset is partitioned into a Challenge Set and an Easy Set, where the former contains only questions answered incorrectly by both a retrieval-based algorithm and a word co-occurrence algorithm. We are also including a corpus of over 14 million science sentences relevant to… See the full description on the dataset page: https://huggingface.co/datasets/allenai/ai2_arc.