ReviewEval
收藏IEEE2026-04-17 收录
下载链接：
https://ieee-dataport.org/documents/revieweval
下载链接
链接失效反馈官方服务：
资源简介：
ReviewEval is a high-quality benchmark constructed through human annotations, consisting of 224 tuples of paper content, high-quality reference peer reviews, high-quality chosen peer reviews, and low-quality rejected peer reviews. It mainly serves for two important purposes: (1) to assess the reliability of automatic metrics for evaluating peer reviews by manually annotating the preferences between chosen and rejected peer reviews; and (2) to provide a more reliable benchmark for evaluating the ability of peer review generation models. The reference reviews in ReviewEval are crafted by integrating valuable opinions from multiple reviewers based on meta reviews, making them more holistic and objective.
提供机构：
Lan, Tian; Zhang, Yun-He; Liu, Ran; Mao, Xian-Ling; Liu, Shu-Hang; Huang, Heyan; Xu, Chen
