AVoE

AVoE是一个合成3D数据集，由新加坡国家大学和A*STAR的研究所联合创建，专注于物理推理中的期望违反现象。该数据集包含5个事件类别，每个类别进一步细分为多个子类别，总计3125个视频。数据集通过Blender软件生成，提供了视觉数据以及抽象特征和规则的地面实况标签，旨在帮助模型进行高级符号预测。AVoE的应用领域包括机器人辅助、自动驾驶和安全AI工具，旨在解决物理推理任务中的期望违反问题。

AVoE is a synthetic 3D dataset co-created by the National University of Singapore and A*STAR's research institutes, focusing on expectation violation phenomena in physical reasoning. This dataset includes 5 event categories, each further subdivided into multiple subcategories, totaling 3,125 videos. Generated via Blender software, the dataset provides visual data along with ground-truth labels for abstract features and rules, aiming to assist models in conducting advanced symbolic prediction. Its application areas cover robotics assistance, autonomous driving and safe AI tools, with the purpose of addressing expectation violation problems in physical reasoning tasks.