PyrXSum

该数据集是一个新收集的英文元评估数据集，旨在评估摘要评价指标。它包含了来自XSum数据集的示例，每个示例都有手动标记的摘要内容单元（SCUs）以及SCU存在标签。该数据集为每份摘要提供了4个回应，这些回应经过筛选以确保质量，其标签通过多数投票确定。规模上，数据集包含了来自10个摘要系统的100个示例，其任务是进行摘要评估。

This is a newly collected English meta-evaluation dataset designed for assessing summarization evaluation metrics. It includes samples from the XSum dataset, where each sample is annotated with manually marked Summary Content Units (SCUs) and SCU presence labels. The dataset provides four responses for each summary; these responses have been screened to ensure quality, and their labels are determined via majority voting. In terms of scale, the dataset contains 100 samples from 10 summarization systems, with the task focused on summarization evaluation.