3D in the Era of World Models

Research in 3D vision and graphics has built explicit 3D-aware pipelines that provide geometric consistency, interpretability, and controllability. At the same time, recent progress in generative video models suggests that large models may implicitly acquire spatial, geometric, and physical knowledge from scale.

This workshop asks how the role of 3D vision research should evolve in the era of large-scale foundation models for spatially aware world models, and how explicit 3D representations can work together with video-based implicit 3D priors.