allenai/compred

ComPRed是一个基于Reddit社区偏好的问答数据集，旨在研究不同社区在回答问题时的偏好多样性。该数据集由Allen Institute for AI、The Ohio State University和Carnegie Mellon University共同策划，包含五个子集，分别基于不同的用户偏好驱动因素，如意识形态、人口统计、社区规范和专业知识水平。数据集的语言为英语，采用ODC-BY许可证。

ComPRed is a question-answering dataset with community-level preferences collected from Reddit. The dataset aims to study diversity in preferences without privacy concerns associated with individual feedback. It is divided into multiple subsets based on factors driving diverging user preferences, such as politics, gender_and_sexuality, finance, history, and science. Each subset has specific features and splits for training, validation, and testing. The dataset is curated by the Allen Institute for AI, The Ohio State University, and Carnegie Mellon University, and is available under the ODC-BY license.