World-R1
Dataset Analysis
World-R1 uses reinforcement learning to align text-to-video generation with 3D constraints, improving geometric consistency without architectural changes.
Provenance
Collected from papers.
Derived from paper: World-R1: Reinforcing 3D Constraints for Text-to-Video Generation