ARC (The Abstraction and Reasoning Corpus)

该数据集旨在通过抽象推理任务来评估深度学习模型的一般化和推理能力。基于知识先验对数据集进行了标注，并用于评估在需要几何先验的任务上的性能表现。具体任务包括从输入-输出对中进行几何变换推断。

This dataset is designed to evaluate the generalization and reasoning capabilities of deep learning models through abstract reasoning tasks. It is annotated based on knowledge priors, and is utilized to assess model performance on tasks that require geometric priors. Specific tasks include inferring geometric transformations from input-output pairs.