Learning Unsupervised World Models for Autonomous Driving via Discrete Diffusion