CIFAR
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
CIFAR-10
99.99RA (Retention Accuracy)
10
CIFAR-10
91.2Accuracy
10
CIFAR-10 cross-silo (test)
86.54Aθ
10
CIFAR-10
88.6Accuracy
10
CIFAR-10 (test train forget)
16.71Test Error
10
CIFAR-100
62.16Accuracy on Dr
10
CIFAR-10
86.68Accuracy (Dr)
10
CIFAR100 (test)
81.15Top-1 Accuracy
10
CIFAR100
81.54Lower Level Accuracy
10
Cifar100 Normal Fatigue (test)
76.93AUACC
10
CIFAR-10
62TPR @ 0.01% FPR
10
CIFAR-10-C OOD corruptions
63.37Accuracy
10
CIFAR-10
93.67Overall Accuracy (OA)
10
CIFAR-100 Instance-dependent Noise (1k clean samples)
67.42Accuracy (eta=0.2)
10
CIFAR-100 with Asymmetric Noise (1k clean samples)
71.53Accuracy (eta=0.2)
10
CIFAR-100 with Symmetric Noise (1k clean samples)
70.36Accuracy (eta=0.2)
10
CIFAR-100 Uniformly Mild scenario
94.1Average Accuracy
10
CIFAR-10 MEST (test)
70Epochs to 90% Acc
10
CIFAR-100 Forget 50%
75.69Test Accuracy
10
CIFAR-10 Forget 50%
94.18Test Accuracy
10
CIFAR-10 Forget 10%
94.29Test Accuracy
10
CIFAR-10 Forget 1%
94.34Test Accuracy
10
CIFAR-100 Pathological heterogeneous (K = 20, IR = 10) (test)
54.57Accuracy
10
CIFAR-10 Pathological heterogeneous (K = 20, IR = 10) (test)
90.36Accuracy
10
CIFAR-100 Practical heterogeneous (alpha = 0.1, K = 20) (test)
54.01Accuracy
10