SCAN

该数据集是一个广泛使用的基准，用于评估模型在组合泛化任务上的表现。它为评估模型的组合泛化能力提供了一个重要的衡量标准。该数据集的核心任务在于促进对组合泛化能力的深入研究和理解。

This dataset is a widely used benchmark for evaluating model performance on compositional generalization tasks. It provides a critical evaluation metric for assessing the compositional generalization capabilities of models. The core purpose of this dataset is to facilitate in-depth research and understanding of compositional generalization abilities.