VaMP: Variational Multi-Modal Prompt Learning for Vision-Language Models | Read Paper on Bytez