FewGlue

该数据集名为FewGLUE，是SuperGLUE基准测试的少样本版本，其中每个数据集在SuperGLUE上包含32个训练样本。它被用于检验所提出算法的少样本学习能力。每个任务的规模为32个训练样本，涉及的自然语言理解任务旨在评估模型的性能。

Named FewGLUE, this dataset is a few-shot variant of the SuperGLUE benchmark, with each of its constituent datasets containing 32 training samples. It is utilized to evaluate the few-shot learning capabilities of proposed algorithms. Each target task has a fixed training set size of 32 samples, and the involved natural language understanding tasks are designed to assess model performance.