Image2Struct: Benchmarking Structure Extraction for Vision-Language Models | Read Paper on Bytez