RandomFab/p14-medical-sft-dpo

该数据集是一个用于医学领域自然语言处理任务的数据集，包含两个配置：dpo和sft。dpo配置用于直接偏好优化，包含问题、选择的回答、拒绝的回答、语言、问题类型、置信度、数据集名称以及各部分的token计数等特征，共有3499个训练样本、1000个验证样本和501个测试样本。sft配置用于监督微调，包含问题、回答、医学主题、是否包含临床案例、语言、问题类型、置信度、数据集名称、token计数以及原始回答等特征，共有3499个训练样本、1000个验证样本和500个测试样本。数据集支持多语言，涵盖多种医学主题和问题类型，适用于训练和评估医学AI模型。This dataset is designed for natural language processing tasks in the medical domain, featuring two configurations: dpo and sft. The dpo configuration is for direct preference optimization, containing features such as question, chosen answer, rejected answer, language, question type, confidence level, dataset name, and token counts for each part, with 3499 training samples, 1000 validation samples, and 501 test samples. The sft configuration is for supervised fine-tuning, containing features such as question, answer, medical subject, presence of clinical case, language, question type, confidence level, dataset name, token counts, and original answer, with 3499 training samples, 1000 validation samples, and 500 test samples. The dataset supports multiple languages, covers various medical subjects and question types, and is suitable for training and evaluating medical AI models.