Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priors | Read Paper on Bytez