DriveLaW:Unifying Planning and Video Generation in a Latent Driving World

arXiv:2512.23421v3 Announce Type: replace Abstract: World models have become crucial for autonomous driving, as they learn how scenarios evolve over time to address the long-tail challenges of the real world. However, current approaches relegate world models to limited roles: they operate within ostensibly unified architectures that still keep world prediction and motion planning as decoupled processes. To bridge this gap, we propose DriveLaW, a novel paradigm that unifies video generation and motion planning. By directly injecting the latent representation from its video generator into the planner, DriveLaW ensures inherent consistency between high-fidelity future generation and reliable trajectory planning. Specifically, DriveLaW consists of two core components: DriveLaW-Video, our powerful world model that generates high-fidelity forecasting with expressive latent representations, and DriveLaW-Act, a diffusion planner that generates consistent and reliable trajectories from the latent of DriveLaW-Video, with both components optimized by a three-stage progressive training strategy. The power of our unified paradigm is demonstrated by new state-of-the-art results across both tasks. DriveLaW not only advances video prediction significantly, surpassing best-performing work by 33.3% in FID and 1.8% in FVD, but also achieves a new record on the NAVSIM planning benchmark.

Related

• Latent Chain-of-Thought World Modeling for End-to-End Driving
• Target-Bench: Can Video World Models Achieve Mapless Path Planning with Semantic Targets?
• SimScale: Learning to Drive via Real-World Simulation at Scale
• INSPATIO-WORLD: A Real-Time 4D World Simulator via Spatiotemporal Autoregressive Modeling
• VADv2: End-to-End Vectorized Autonomous Driving via Probabilistic Planning

Source: arXiv cs.CV | 2026-04-20