medmcqa
收藏资源简介:
该数据集是一个大规模选择题问答数据集,包含159,075条样本,其中训练集154,921条,验证集4,077条,测试集4,077条。每条数据代表一个完整的选择题,包括以下字段:唯一标识符(id)、问题文本(question)、四个选项(opa, opb, opc, opd)、正确答案(cop,取值为a/b/c/d)、选择题类型(choice_type)、题目解析(exp)、所属学科名称(subject_name)和具体主题名称(topic_name)。数据集采用文本格式,适用于教育评估、自动问答系统、学科知识测试模型训练等任务,特别适合用于选择题理解和答案预测的研究与应用。
This dataset is a large-scale multiple-choice question answering dataset, comprising a total of 159,075 samples, with 154,921 in the training split, 4,077 in the validation split, and 4,077 in the test split. Each data instance represents a complete multiple-choice question, including the following fields: unique identifier (id), question text (question), four options (opa, opb, opc, opd), correct answer (cop, with values of a/b/c/d), multiple-choice question type (choice_type), question explanation (exp), subject name (subject_name), and specific topic name (topic_name). The dataset is provided in text format, and is applicable to tasks such as educational assessment, automatic question answering systems, and training of subject knowledge test models, and is particularly well-suited for research and applications focused on multiple-choice question understanding and answer prediction.
数据集概述:MedMCQA
数据集名称:MedMCQA
来源页面:https://huggingface.co/datasets/leandrodevai/medmcqa
数据集结构
MedMCQA 是一个用于医学领域多选题问答的数据集,包含以下字段:
- id (string):样本唯一标识符
- question (string):问题文本
- opa, opb, opc, opd (string):四个选项(A、B、C、D)
- cop (class_label):正确答案标签,映射为:0 → A, 1 → B, 2 → C, 3 → D
- choice_type (string):选项类型(如单/多选)
- exp (string):答案解释
- subject_name (string):所属学科名称
- topic_name (string):所属主题名称
数据划分与规模
数据集分为三个子集,总计约 16.3 万 条样本:
| 划分 | 样本数 | 数据量(字节) |
|---|---|---|
| 训练集 | 154,921 | 111,113,267 |
| 验证集 | 4,077 | 2,924,127 |
| 测试集 | 4,077 | 2,924,127 |
- 总下载大小:77,561,378 字节
- 总数据集大小:116,961,521 字节
数据格式与配置
- 默认配置名:
default - 数据文件路径:
- 训练集:
data/train-* - 验证集:
data/validation-* - 测试集:
data/test-*
- 训练集:
每个样本的选项和正确答案以类别标签形式存储,便于直接用于分类任务。




