Grounding Everything: Emerging Localization Properties in Vision-Language Transformers | Read Paper on Bytez