Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

OCR-VQA

Benchmarks

Task NameDataset NameSOTA ResultTrend
Visual Question AnsweringOCR-VQA (test)
Accuracy77.8
77
OCR-based Visual Question AnsweringOCR-VQA
Accuracy65.6
61
Image question answeringOCR-VQA
Accuracy75.8
27
Image Question AnsweringOCR-VQA
ROUGE-L70.5
20
Visual Question AnsweringOCR-VQA (val)
Accuracy71.1
17
Visual Question AnsweringOCR-VQA
Exact Match (EM)77.8
9
Visual Question AnsweringOCR-VQA Non-IID
Accuracy76.39
5
Visual Question AnsweringOCR-VQA IID
Accuracy (ACC)75.86
5
QA over IllustrationsOCR-VQA (test)
F1 Score74
5
Showing 9 of 9 rows