WMarkGPT: Watermarked Image Understanding via Multimodal Large Language Models | Read Paper on Bytez