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

OKVQA

Benchmarks

Task NameDataset NameSOTA ResultTrend
Visual Question AnsweringOKVQA
Top-1 Accuracy75.29
283
Visual Question AnsweringOKVQA (val)
VQA Score66.1
101
Knowledge-based Visual Question AnsweringOKVQA
Accuracy0.661
79
Visual Question AnsweringOKVQA
ASR100
42
Vision Question AnsweringOKVQA
ASR (Success Rate)73.07
30
Knowledge-based Visual Question AnsweringOKVQA (val)
Accuracy66.7
27
Knowledge-based Visual RetrievalOKVQA Google Search (test)
PR@584.66
16
Cognition and ReasoningOKVQA
Score61.92
16
Visual Question AnsweringOKVQA
VQA Accuracy (Clean)60.3
14
Visual Question AnsweringOKVQA
Accuracy60.13
14
Visual Question AnsweringOKVQA (test)
Accuracy65.7
11
Visual Question AnsweringOKVQA (I) (test)
VQA Accuracy57.8
11
Visual Question AnsweringOKVQA N=200
Score D61.7
11
Visual Question AnsweringOKVQA (N=100)
CD Score56.6
11
Visual Question AnsweringOKVQA N=40
CD48.8
11
Object Hallucination ProbingOKVQA POPE Popular
Accuracy85
11
Visual Question AnsweringOKVQA
AUROC0.788
9
Knowledge-based Question AnsweringOKVQA
Score64.56
9
End-to-end RetrievalOKVQA
R@10041
6
Visual Question AnsweringOKVQA (val-lite)
Accuracy48.68
6
Knowledge-based Visual RetrievalOKVQA WK11M (test)
MRR@551.15
6
Knowledge-based Visual Question AnsweringOKVQA M2KR
VQA Score0.661
6
Object Hallucination ProbingOKVQA POPE Adversarial
POPE Score (Zh)79.97
6
Object Hallucination ProbingOKVQA POPE Random
Accuracy (Zh)86.03
6
RetrievalOKVQA (test)
PR@590.9
5
Showing 25 of 27 rows