Gen-Review
收藏资源简介:
Gen-Review是一个包含超过81,000条AI生成评论的数据集,针对2018年至2025年提交给ICLR的论文,每个论文有三个独立的提示:负面、正面和中立。该数据集包括32,000篇论文及其原始人类编写的评论,允许对AI生成内容的同行评审进行全面的评估。Gen-Review旨在促进对AI生成评论中潜在偏见的、质量的、与人类编写评论的匹配性以及AI检测工具的评价的研究。
Gen-Review is a dataset comprising over 81,000 AI-generated peer reviews for papers submitted to ICLR between 2018 and 2025. Each paper is associated with three distinct prompts for generating reviews: negative, positive, and neutral. The dataset also includes 32,000 papers along with their original human-written reviews, enabling comprehensive evaluations of AI-generated peer reviews. Gen-Review aims to facilitate research on evaluating potential biases, quality, alignment with human-written reviews, and the efficacy of AI detection tools for AI-generated peer reviews.




