PhysicsDrivenWorld (PDW) fine-tunes CogVideoX-2b using Low-Rank Adaptation (LoRA) supervised by an NVIDIA Warp rigid-body physics simulator.
Modern video diffusion models generate visually plausible but physically inconsistent results — objects float, bounce unrealistically, or violate Newton's laws. PDW injects a physics prior into the model's denoising weights by training on Warp-simulated ground-truth trajectories.
The training objective is standard diffusion denoising MSE, but applied exclusively to frames that are physically correct by construction from the Warp simulator — so the model learns to denoise physics-consistent content better than physics-inconsistent content.






Recommended for:

Physics-grounded text-to-video generation for educational and scientific visualisation
Research baseline for physics-consistent video generation methods
Demonstrating that small LoRA adapters (<0.2% of parameters) can inject physical priors into large video diffusion models
Foundation for extension to more complex physics scenarios