JBLIMP

该数据集是一个专门用于针对性评估日语语法的最小对立体数据集，它包含了从语法文章中精心挑选的331组语法可接受性判断的最小对立体。该数据集采用SLOR评分方法来计算准确性，以减轻词汇频率的影响，其任务是进行语法评估。

This dataset is a minimal pair dataset specifically designed for targeted evaluation of Japanese grammar. It includes 331 sets of minimal pairs for grammatical acceptability judgment, carefully curated from grammatical texts. The dataset uses the SLOR scoring method to calculate accuracy, so as to mitigate the confounding effect of lexical frequency, and its core task is grammatical acceptability evaluation.