Dyn-VPP: Video Prediction Policy Optimization for Improved Visual Dynamics

Video action models are a promising foundation for Vision-Language-Action (VLA) because they can learn rich visual dynamics directly from video. However, likelihood-oriented training of diffusion predictors emphasizes globally plausible futures and does not guarantee precision-critical visual dynamics needed for manipulation, so small prediction errors can be amplified by downstream policies. We propose Dyn-VPP, a posttraining framework that casts multi-step denoising as policy optimization and aligns predicted future latents with expert visual dynamics via a verifiable terminal reward, without modifying any architecture. This enables explicit optimization of dynamics signals that are not captured by likelihood-only training. As a result, Dyn-VPP yields more accurate visual dynamics and improves downstream task execution. Experiments across diverse simulated and real-world manipulation settings show improved dynamics consistency and consistently higher task success.