BYD Unveils Hybrid World Model HyWorldVLA

On July 29, BYD's artificial intelligence team published a research paper titled "HyWorldVLA," introducing HyWorldVLA, a hybrid world model for autonomous driving. The model integrates pixel-level and latent-space world modeling within a vision-language-action (VLA) architecture. On the public NAVSIM v1 benchmark, HyWorldVLA achieved a PDMS score of 90.59, which evaluates driving quality based on metrics including collision safety, compliance with drivable areas, safe following distance, task completion, and motion comfort.