Simulated Environment for Concept Learning

该数据集是一个全新的模拟环境，包含一个由位置、颜色和形状属性特征化的二维场景实体。该环境用于评估在生成和识别背景下进行概念学习的情况。此外，该环境使模型能够从生成和识别两种背景中学习，并具备在不同场景下复用已学概念的能力。任务包括从演示事件中学习视觉、定量、关系和时间概念。

This dataset is a novel simulated environment containing 2D scene entities characterized by positional, color, and shape attributes. This environment is developed to evaluate concept learning in the context of generation and recognition. Furthermore, it enables models to learn from both generative and recognition contexts, and supports the reuse of learned concepts across diverse scenarios. The tasks included focus on learning visual, quantitative, relational, and temporal concepts from demonstration events.