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

Perturbation-based evaluation on VQA v2

1.13Positive Perturbation AUC

DEX-AR

0.74520.84510.9451.0449Mar 6, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
1.130.81
2026.03
1.090.86
2026.03
1.060.86
2026.03
1.050.84
2026.03
1.030.84
2026.03
1.020.86
2026.03
1.010.86
2026.03
0.980.86
2026.03
0.970.97
2026.03
0.970.95
2026.03
0.960.75
2026.03
0.950.9
2026.03
0.950.99
2026.03
0.950.99
2026.03
0.950.79
2026.03
0.950.75
2026.03
0.941
2026.03
0.940.98
2026.03
0.940.75
2026.03
0.940.74
2026.03
0.930.78
2026.03
0.930.77
2026.03
0.930.76
2026.03
0.920.78
2026.03
0.920.77
2026.03
0.910.77
2026.03
0.910.8
2026.03
0.90.8
2026.03
0.90.8
2026.03
0.890.89
2026.03
0.880.89
2026.03
0.860.86
2026.03
0.850.87
2026.03
0.760.74