SKIML-ICL/med_retrieved

该数据集包含两个配置：冲突(conflict)配置和默认(default)配置。每个配置都包含问题ID、标准化问题、标准化答案、问题文本、答案序列、上下文信息、是否有答案、可回答性、实体类型、相似实体及其分数、随机实体及其分数、冲突句子、冲突生成提示、冲突段落、是否为有效GPT冲突段落等特征。测试集分别包含485和18701个示例。

The dataset consists of two configurations: conflict and default. Each configuration includes features such as question ID, normalized question, normalized answers, question text, answer sequence, context information, whether there is an answer, answerability, entity type, similar entity and its score, random entity and its score, conflict sentence, prompt for conflict generation, conflict passage, whether it is a valid GPT conflict passage, etc. The test sets contain 485 and 18701 examples respectively.