amu-cai/reVISION-dataset

reVISION数据集是一个基于图像的问题集合，来源于波兰国家考试。每个问题以图像形式呈现，并且只有一个正确答案。数据集涵盖了不同类型和学科的考试，包括8年级考试、中学考试、高中考试和专业考试，学科范围从波兰语、数学到自然科学等。数据集中共有超过10万但少于100万个样本。需要注意的是，在需要图像上下文的问题元数据中，有122个图像被错误分类，其中大多数是假阳性。这种错误水平对于探索性分析和粗略分组是可以接受的，但不适合用于训练下游模型。

The reVISION dataset is a collection of image-based questions sourced from Polish National Exams. Each question is represented in the form of an image with only one correct answer. The dataset covers various types of exams across different disciplines, including 8th-grade exams, middle school exams, high school exams, and professional exams, ranging from Polish Language, Mathematics to Natural Sciences, etc. The dataset contains between 100,000 and 1 million samples. It is worth noting that out of the images requiring image context in the question metadata, 122 were misclassified, with the majority being false positives. This level of error is acceptable for exploratory analysis and broad grouping but is not sufficiently accurate for training downstream models.