CoSIm

该数据集名为CoSIm，旨在评估人工智能系统对场景变化想象力的推理能力。它包含了图像与常识性问题-回答对的配对，对想象中的反事实变化的描述，针对新问题的回答，以及干扰项回答。数据集涵盖了各种复杂的场景变化类型，如物体的增加/移除/状态变化、事件描述和环境变化。规模上，该数据集包含3.5千个实例（3.5千张图像及其相关文本）。其任务是对反事实场景想象力的常识推理。

This dataset, named CoSIm, is designed to evaluate the reasoning capabilities of artificial intelligence systems for imagination-based reasoning about scene changes. It consists of paired images and commonsense question-answer pairs, along with descriptions of imagined counterfactual changes, answers to novel questions, and distractor answers. The dataset covers a wide range of complex scene variation types, including object addition, removal, state changes, event descriptions, and environmental changes. In terms of scale, the dataset contains 3.5 thousand instances (3.5 thousand images and their associated text). The core task of this dataset is commonsense reasoning for counterfactual scene imagination.