SLING

该数据集是一个由中文树库派生出的3.8万最小句子对集合，旨在通过应用句法和词汇转换，改进早期数据集的性能。此外，该数据集还致力于解决早期数据集中词汇多样性的局限，并用于句法评估任务。

This dataset is a collection of 38,000 minimal sentence pairs derived from Chinese treebanks, which aims to improve the performance of earlier datasets via syntactic and lexical transformations. Furthermore, this dataset addresses the limitation of lexical diversity in prior datasets and is intended for syntactic evaluation tasks.