Hi, I want to ask about the https://huggingface.co/datasets/csfufu/mmrl dataset. I want to know, how to get extracted_answers? Do you use some vlm to predict these answer? Best wishes,