Wilds

该数据集是一套旨在评估机器学习模型在分布偏移情况下泛化能力的基准测试集。它包含了WILDS系列数据集，这些数据集专注于不同类型的分布偏移，以应对各种任务中的分布外泛化问题。

This benchmark suite is designed to evaluate the generalization performance of machine learning models under distribution shifts. It encompasses the WILDS series of datasets, which focus on diverse types of distribution shifts to address out-of-distribution (OOD) generalization challenges across various tasks.