OpenBookQA is a multiple-choice question answering dataset modeled after open-book exams to probe deeper understanding and multi-step reasoning. The dataset provides an “open book” of elementary-level science facts (≤1.3k facts) plus roughly 6k multiple-choice questions that require combining a provided core science fact with broad common-sense or world knowledge to answer. Each example contains a question stem, four answer choices, an answer key, and an associated core fact (the ‘‘open book’’ fact). The data is split into train (~4.96k questions), validation (500) and test (500). It was created to encourage research on reasoning and knowledge-combination beyond surface-level reading comprehension.
No results indexed yet — be the first to submit a score.
Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.