Vision-Language models, VQA, DocumentAI