M4U-Benchmark/M4U-mini

M4U是一个多语言多模态数据集，用于评估大型模型在理解和推理方面的多语言能力。数据集涵盖了科学、工程和医疗保健三个学科领域，并包含英语、中文、德语、阿拉伯语、泰语和日语等多种语言的数据。每个样本包括图像、问题、选项、答案等详细信息。M4U-mini是M4U的一个小型版本，包含了三种额外的语言，是测试数据的5%随机样本。

M4U is a multilingual multimodal dataset designed for evaluating the multilingual understanding and reasoning capabilities of large models. It covers three disciplinary areas: science, engineering, and healthcare, and includes data in multiple languages such as English, Chinese, German, Arabic, Thai, and Japanese. Each sample contains detailed information including images, questions, options, answers, and more. M4U-mini is a smaller version of M4U, which includes three additional languages and is constructed from 5% of the random test data.