CIFAR-10
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
CIFAR-10 (250 labels)
2.42Error Rate
66
CIFAR-10 Symmetric Noise (test)
93.6Test Accuracy (Overall)
64
CIFAR-10 BigGAN
8.01FID
62
CIFAR-10 SNGAN
10.6FID
62
CIFAR-10-C (test)
91.57Accuracy (Clean)
61
CIFAR-10 Split (test)
98.31Average Accuracy
59
CIFAR-10 Long Tailed Imbalance Ratio 50 (test)
89.2Top-1 Accuracy
57
CIFAR-10 vs CIFAR-10.1 1.0 (test)
0.993Test Power
54
CIFAR-10 20% uniform noise (test)
91.94Accuracy
52
CIFAR-10 40% uniform noise (test)
90.2Accuracy
51
CIFAR-10 (train)
0Error Rate
51
CIFAR-10 (test)
91Attack AUC
49
CIFAR-10 3 classes
95.37Retain Accuracy
48
CIFAR-10 (test)
94.4Accuracy
44
CIFAR-10 32x32 EDM (test)
1.98FID
44
CIFAR-10 uniform label noise, η=0.3 (test)
57.12Accuracy
43
CIFAR-10 uniform label noise, η=0.1 (test)
65.08Accuracy
43
CIFAR-10N (test)
98.8Accuracy (Random 1)
43
CIFAR-10 Standard data augmentation (test)
3.38Test Error Rate
43
CIFAR-10 10% label
91.32Accuracy
42
CIFAR-10 32x32 unconditional (test)
1.87FID
42
CIFAR-10 (test)
37.2FID
42
CIFAR-10 (test)
92.86Clean Accuracy
40
CIFAR-10
94.1Robust Accuracy
40
CIFAR-10 (test)
86.39Clean Accuracy
40