HoVLE: Unleashing the Power of Monolithic Vision-Language Models with Holistic Vision-Language Embedding | Read Paper on Bytez