CIFAR
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Split CIFAR-10 (5 disjoint subsets)
53.07ACC
22
CIFAR-100 Split 1.0 (test)
76.14ACC
22
CIFAR-100
75.31Accuracy
22
CIFAR-10
96.37Accuracy
22
CIFAR-100 (test)
0.852Accuracy (All)
22
CIFAR-100 50% symmetric noise
79.19Accuracy
22
CIFAR-100 20% symmetric noise
79.43Accuracy
22
CIFAR-10 (test)
96.1Accuracy (Sym, 20%)
22
CIFAR-10
64.9ASR
22
CIFAR-100 (test)
55.05Top-1 Error
22
CIFAR-10 ε = 36/255 (test)
87.5Clean Accuracy
22
CIFAR-100 Standard data augmentation (test)
17.18Test Error
22
CIFAR-10 400 labels per class (test)
88.44Accuracy
22
CIFAR-10 (train)
0Average Gap
22
CIFAR-10N Random 1 noise
82.9Test Accuracy
21
CIFAR-10
100Attack Success Rate (ASR)
21
CIFAR-10 LT-10
90.65Clean Accuracy
21
CIFAR-10
99.5MDSE
21
CIFAR-100 50% random data removal
99.98Retain Accuracy
21
CIFAR-100 (10% random data removal)
99.98Retain Accuracy (RA)
21
CIFAR-10 (val)
91.54Top-1 Accuracy
21
CIFAR100
1ASR
21
CIFAR-10 class-independent noise (test)
91.96Accuracy
21
CIFAR-100
74.06Accuracy Retention
21
CIFAR-100
74.34Accuracy Retention
21