ScreenAI: A Vision-Language Model for UI and Infographics Understanding

Benchmark Model Rank Results
chart-question-answering-on-chartqaScreenAI 5B (4.62 B params, w/ OCR)#41:1 Accuracy: 76.7
visual-question-answering-on-docvqa-testScreenAI 5B (4.62 B params, w/OCR)#4ANLS: 0.8988
visual-question-answering-vqa-on-infographicvqaScreenAI 5B (4.62 B params, w/ OCR)#2ANLS: 65.90