WorldVLA: A Unified Vision-Language-Action and World Model

We present WorldVLA, an autoregressive action world model that unifies action and image understanding and generation. Our WorldVLA integrates Vision-Language-Action (VLA) model and world model in one single framework. The world model predicts future images by leveraging both action and image understanding, with the purpose of learning the underlying physics of the environment to improve action generation. Meanwhile, the VLA model generates the subsequent actions based on image observations, aiding in visual understanding and in turn helps visual generation of the world model. We demonstrate that WorldVLA outperforms standalone VLA and world models, highlighting the mutual enhancement between the world model and the VLA model. We evaluate WorldVLA in both simulation and real-world robot tasks. WorldVLA achieves 97.4% success rate on the LIBERO simulation benchmark without pretraining, while in real-world LeRobot experiments, its integrated world model boosts the overall success rate by 50%.