Yao Yao

MLLMs can generate 3D models, but they often miss the mark on precise geometry and coherent assemblies, revealing significant limitations in their structural reasoning abilities.

JenBridge outperforms existing soundtracking methods by intelligently adapting to scene transitions, ensuring narrative coherence in long-form videos like never before.

Real-time 3D scene reconstruction from streaming video is now possible with a feed-forward transformer that outperforms traditional SLAM methods.