From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D | Read Paper on Bytez