Are We on the Right Way for Evaluating Large Vision-Language Models? | Read Paper on Bytez