Enhancing Visual Document Understanding with Contrastive Learning in Large Visual-Language Models | Read Paper on Bytez