HOTPOTQA is a new dataset designed to challenge and advance the capabilities of question-answering (QA) systems. The dataset consists of 113,000 Wikipedia-based question-answer pairs, featuring four key features: (1) questions require reasoning over multiple supporting documents, (2) questions are diverse and not constrained by pre-existing knowledge bases, (3) sentence-level supporting facts are provided for reasoning, and (4) a new type of factoid comparison questions is introduced. The dataset aims to provide strong supervision and enable explainable reasoning by allowing models to understand the underlying reasoning process. The authors collected the data through crowdsourcing, ensuring that the questions are natural and not constrained by any specific knowledge base schema. They also designed a data collection pipeline to ensure high-quality multi-hop questions. The dataset is publicly available and can be used to evaluate and improve the performance of QA systems, particularly in multi-hop reasoning and explainability.HOTPOTQA is a new dataset designed to challenge and advance the capabilities of question-answering (QA) systems. The dataset consists of 113,000 Wikipedia-based question-answer pairs, featuring four key features: (1) questions require reasoning over multiple supporting documents, (2) questions are diverse and not constrained by pre-existing knowledge bases, (3) sentence-level supporting facts are provided for reasoning, and (4) a new type of factoid comparison questions is introduced. The dataset aims to provide strong supervision and enable explainable reasoning by allowing models to understand the underlying reasoning process. The authors collected the data through crowdsourcing, ensuring that the questions are natural and not constrained by any specific knowledge base schema. They also designed a data collection pipeline to ensure high-quality multi-hop questions. The dataset is publicly available and can be used to evaluate and improve the performance of QA systems, particularly in multi-hop reasoning and explainability.