Demystifying the Visual Quality Paradox in Multimodal Large Language Models | Best AI papers explained | Podwise