M$^3$GPT: An Advanced Multimodal, Multitask Framework for Motion Comprehension and Generation | Read Paper on Bytez