AdaWorld: Learning Adaptable World Models with Latent Actions
This addresses the problem of costly adaptation for intelligent agents in heterogeneous environments, representing a novel method for a known bottleneck.
The paper tackles the challenge of adapting world models to novel environments with limited action-labeled data by proposing AdaWorld, which learns latent actions from videos in a self-supervised manner and conditions an autoregressive world model on them, achieving superior performance in simulation quality and visual planning across multiple environments.
World models aim to learn action-controlled future prediction and have proven essential for the development of intelligent agents. However, most existing world models rely heavily on substantial action-labeled data and costly training, making it challenging to adapt to novel environments with heterogeneous actions through limited interactions. This limitation can hinder their applicability across broader domains. To overcome this limitation, we propose AdaWorld, an innovative world model learning approach that enables efficient adaptation. The key idea is to incorporate action information during the pretraining of world models. This is achieved by extracting latent actions from videos in a self-supervised manner, capturing the most critical transitions between frames. We then develop an autoregressive world model that conditions on these latent actions. This learning paradigm enables highly adaptable world models, facilitating efficient transfer and learning of new actions even with limited interactions and finetuning. Our comprehensive experiments across multiple environments demonstrate that AdaWorld achieves superior performance in both simulation quality and visual planning.