Concept-Guided Spatial Regularization for World Models in Atari Pong· 概念引导的空间正则化在 Atari Pong 世界模型中的应用
World models are usually evaluated as components of model-based reinforcement learning (MBRL) systems, while the world models themselves are rarely studied in isolation. We examine five representative visual world-model agents in Atari Pong: DreamerV3, DIAMOND, TWISTER, Simulus, and STORM. After reproducing their training pipelines and matching the reported agent performance, we freeze the learned world models and evaluate them with a closed-loop rollout diagnostic: a policy trained separately from the corresponding MBRL agent interacts with each frozen model, and the generated video trajectories are inspected for visual and dynamical errors. Across all five models, the rollouts contain clear failures, including ball disappearance, incorrect ball motion, and invalid ball-paddle interaction
研究五种Atari Pong世界模型的视觉和动态错误。
- 核心方法
- 通过冻结训练好的世界模型，并使用独立训练的策略与这些模型互动，检测生成的视频轨迹中的错误。
- 适合谁读
- 研究者
- 要解决的问题
- 世界模型在MBRL系统中的性能问题，特别是在视觉和动态表现上的缺陷。
- 关键实验
- 对DreamerV3, DIAMOND, TWISTER, Simulus, 和 STORM五种模型进行了闭合回路评估，检测了包括球消失、球运动不正确和球-拍互动失效等问题。
- 主要贡献
- 揭示了五种代表性世界模型在Atari Pong任务中的具体缺陷，为改进模型提供了方向。
- 意义与局限
- 为世界模型的研究提供了一个新的视角，强调了模型本身性能评估的重要性。局限在于实验仅限于Atari Pong环境，可能不适用于其他更复杂的游戏。