JudSacr/MMLU-ProX

这是一个多语言选择题数据集，包含多种语言（如南非荷兰语、阿拉伯语、孟加拉语、捷克语、德语、英语、西班牙语、法语、印地语、匈牙利语、印度尼西亚语、意大利语、日语、韩语、马拉地语、尼泊尔语、葡萄牙语、俄语、塞尔维亚语、斯瓦希里语、泰卢固语、泰语、乌克兰语、乌尔都语、越南语、沃洛夫语、约鲁巴语、中文、祖鲁语）的配置。每个配置都有验证集和测试集，数据集特征包括问题ID、问题文本、最多10个选项（option_0到option_9）、答案文本、答案索引、思维链内容（cot_content）、类别（category）、来源（src）和原始问题ID（question_id_src）。数据集可能用于多语言问答或推理任务评估，但具体目的和来源未在README中说明。数据规模较大，每个语言的测试集约有11759个示例，验证集有70个示例。

This is a multilingual multiple-choice question dataset that includes configurations for various languages (e.g., Afrikaans, Arabic, Bengali, Czech, German, English, Spanish, French, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Marathi, Nepali, Portuguese, Russian, Serbian, Swahili, Telugu, Thai, Ukrainian, Urdu, Vietnamese, Wolof, Yoruba, Chinese, Zulu). Each configuration has validation and test splits. The dataset features include question ID, question text, up to 10 options (option_0 to option_9), answer text, answer index, chain-of-thought content (cot_content), category, source (src), and original question ID (question_id_src). It is likely designed for multilingual question-answering or reasoning tasks evaluation, but the specific purpose and origin are not detailed in the README. The dataset is large-scale, with approximately 11,759 examples in the test set and 70 examples in the validation set for each language.