LAMA

该数据集名为LAMA，是一个广泛使用的事实探查数据集，旨在评估预训练语言模型的事实知识掌握情况。LAMA测试集进一步被划分为LAMA-Easy和LAMA-Hard两个子集，分别对应不同难度级别的探查任务。其主要任务是对事实知识进行探查。

The dataset named LAMA is a widely adopted factual probing benchmark designed to evaluate the factual knowledge mastery of pre-trained language models. The LAMA test set is further divided into two subsets: LAMA-Easy and LAMA-Hard, which correspond to probing tasks at different difficulty levels. Its core task is to probe factual knowledge.