Sherlock: Self-Correcting Reasoning in Vision-Language Models | Read Paper on Bytez