Gameworld

该数据集包含10款多样化设计的小游戏，旨在让人类学习者在几分钟内能够解决。这些游戏采用了轻量级和结构化的设计。此外，Gameworld 10k数据集还支持进行受控干预，例如改变对象的颜色或形状，以测试智能体在表面领域变化中的泛化能力。任务旨在评估基于像素的深度强化学习方法在学习玩游戏时，Axiom算法的性能表现。

This dataset contains 10 mini-games with diverse designs, all intended to be solvable by human learners within minutes. These games feature lightweight and structured designs. Furthermore, the Gameworld 10k dataset supports controlled interventions, such as modifying the color or shape of in-game objects, to test an AI Agent's generalization ability against superficial domain shifts. The tasks are designed to evaluate the performance of the Axiom algorithm when learning to play games via pixel-based deep reinforcement learning approaches.