MV-Forcing: Long Multi-View Video Generation via 4D-Grounded Spatio-Temporal Self-Forcing | Gal Fiebelman et al. | ResearchPod