MMRL: Multi-Modal Representation Learning for Vision-Language Models | Read Paper on Bytez