RACE is a new dataset designed for evaluating machine reading comprehension, specifically for middle and high school students in China aged 12 to 18. The dataset consists of 27,933 passages and 97,687 questions, with a significant focus on reasoning tasks. Unlike other datasets, RACE includes a variety of question types, such as passage summarization and attitude analysis, and covers a broad range of topics and writing styles. The questions are generated by human experts and are not restricted to text spans in the original passage, making the task more challenging. The dataset is freely available and aims to provide a valuable resource for researchers and developers in the field of machine comprehension. The performance gap between state-of-the-art models (43%) and human performance (95%) highlights the need for more advanced models to match human-level comprehension.RACE is a new dataset designed for evaluating machine reading comprehension, specifically for middle and high school students in China aged 12 to 18. The dataset consists of 27,933 passages and 97,687 questions, with a significant focus on reasoning tasks. Unlike other datasets, RACE includes a variety of question types, such as passage summarization and attitude analysis, and covers a broad range of topics and writing styles. The questions are generated by human experts and are not restricted to text spans in the original passage, making the task more challenging. The dataset is freely available and aims to provide a valuable resource for researchers and developers in the field of machine comprehension. The performance gap between state-of-the-art models (43%) and human performance (95%) highlights the need for more advanced models to match human-level comprehension.