99s42m/QAWithRAGLLMPrompts

这是一个用于检测大型语言模型（LLM）幻觉现象的清理和整合的问题（带有上下文）和答案数据集。每个问题-答案对并非作者原创，而是从Narrative QA、DROP、MS_MARCO、Beer QA和COSMOS QA等数据集中选取的。数据集包含passage、question和answers三个特征，适用于文本分类、文本生成、文本到文本生成和问答等任务。使用此数据集时，需引用上述来源及Shreyan Mitra和Leilani Gilpin的相关论文。

A cleaned and consolidated set of questions (with context) and answers for LLM hallucination detection. Each question-answer pair is not the work of the author, but was selected from one of the following datasets: Narrative QA, DROP, MS_MARCO, Beer QA, COSMOS QA. The dataset features include: passage, question, and answers. The dataset is divided into a training set containing 1,327,990 samples with a total size of 3,925,787,429 bytes. The main use of the dataset is to detect hallucinations by feeding the questions and corresponding passages into the chosen LLM and calculating the similarity of the LLM-generated response with the correct response given in the answers column.