Evaluating Vision-Language Models on Bistable Images | Read Paper on Bytez