VQA-RAD
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
VQA-RAD
80.4Accuracy
251
VQA-RAD (Closed)
1.3ECE
96
VQA-RAD (Open)
0.819AUROC
96
VQA-RAD Closed
70.2AUROC
96
VQA-RAD
85.5Overall Accuracy
67
VQA-RAD (test)
87.9Closed Accuracy
65
VQA-RAD (All)
78.23AUC
57
VQA-RAD Open-Ended
83.13AUC
57
VQA-RAD
80.45Accuracy (%)
48
VQA-RAD (test)
90.4Overall Accuracy
48
VQA-RAD closed-end
84.86Accuracy
45
VQA-RAD Open
61.5Accuracy
26
VQA-RAD open-ended
29Exact Match (EM)
25
VQA-RAD Open
73.7Token Recall
20
vqa-rad
16.85ECE
16
VQA-RAD
92.16PR-AUC
16
VQA-RAD closed-ended
82.5Accuracy
16
VQA-RAD 10
78.8Closed Accuracy
12
VQA-RAD
87.1Accuracy
12
VQA-RAD Closed
88Exact Match
11
VQA-RAD cross-domain
0.789Accuracy
10
VQA-RAD (in-domain)
83.3Accuracy
10
VQA-RAD n=350 (test)
77.7LLM-Judge Accuracy
7
VQA-RAD (test)
60.6Risk
7
VQA-RAD Closed
88.2Accuracy
7