mPLUG-DocOwl 1.5: Unified Structure Learning for OCR-free Document Understanding

Benchmark Model Rank Results
multimodal-table-fact-verification-on-tabfact-image-inputDocOwl 1.5 (extra data)#1Accuracy: 80.4
infographic-visual-question-answering-on-infovqaDocOwl 1.5 Chat (extra data)#9ANLS: 50.7
visual-question-answering-on-docvqa-testDocOwl 1.5 Chat (extra data)#67ANLS: 0.822
chart-question-answering-on-chartqaDocOwl 1.5 (extra data)#921:1 Accuracy: 70.5