FlagEval/SAT

这是一个用于空间能力训练的多模态语言模型的数据集，包含150个图像问答对。数据集从array/SAT数据集改编而来，转换成了更易于访问和使用的格式。每个样本包含问题ID、问题文本、答案选项、问题类型、正确答案和对应的图像数据。This is a dataset for spatial aptitude training of multimodal language models, containing 150 image QA pairs. The dataset is adapted from the array/SAT dataset and converted into a more accessible and user-friendly format. Each sample includes a question ID, question text, answer options, question type, correct answer, and corresponding image data.