PaLI-3 Vision Language Models: Smaller, Faster, Stronger

Benchmark Model Rank Results
temporal-casual-qa-on-next-qaPaLI-3#2WUPS: 37.7
visual-question-answering-vqa-on-infographicvqaPaLI-3 (w/ OCR)#15ANLS: 62.4
visual-question-answering-vqa-on-infographicvqaPaLI-3#17ANLS: 57.8
visual-question-answering-on-docvqa-testPaLI-3 (w/ OCR)#47ANLS: 0.886
visual-question-answering-on-docvqa-testPaLI-3#54ANLS: 0.876
chart-question-answering-on-chartqaPaLI-3#961:1 Accuracy: 70
chart-question-answering-on-chartqaPaLI-3 (w/ OCR)#981:1 Accuracy: 69.5