Dream Diffusion Policy enables robots to survive severe OOD disturbances by detecting reality-imagination discrepancies and switching to an internal world model.
March 24, 2026
Original Paper
Dreaming the Unseen: World Model-regularized Diffusion Policy for Out-of-Distribution Robustness
arXiv · 2603.21017
The Takeaway
Standard diffusion policies fail when objects are moved or vision is corrupted; this framework allows a robot to 'dream' its way through a disturbance and realign later. The jump from 3.3% to 83.3% success rate under spatial shifts represents a major milestone for real-world robotic robustness.
From the abstract
Diffusion policies excel at visuomotor control but often fail catastrophically under severe out-of-distribution (OOD) disturbances, such as unexpected object displacements or visual corruptions. To address this vulnerability, we introduce the Dream Diffusion Policy (DDP), a framework that deeply integrates a diffusion world model into the policy's training objective via a shared 3D visual encoder. This co-optimization endows the policy with robust state-prediction capabilities. When encountering