CALVIN

该数据集名为CLIPORT，是一个专为评估语言条件策略适应中的复合泛化能力而设计的基准测试。它被用于评估任务成功率以及复合泛化能力。该数据集包含了10个不同场景下的100个评估实例，旨在衡量复合泛化以及在分布外泛化的任务表现。

This dataset, named CLIPORT, is a benchmark specifically developed to evaluate compositional generalization abilities in language-conditioned policy adaptation. It is used to assess both task success rates and compositional generalization capabilities. This dataset contains 100 evaluation instances across 10 distinct scenarios, aiming to measure task performance for both compositional generalization and out-of-distribution generalization.