CIFAR
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
CIFAR10 Symmetric Noise, η=0.8
62.08Accuracy
10
CIFAR10 Symmetric Noise, η=0.6
81.82Accuracy
10
CIFAR10 Symmetric Noise, η=0.4
88.17Accuracy
10
CIFAR-10 (test)
0.237Error Bound
10
CIFAR100C Severity 5 (test)
66.1Accuracy (Gaussian Corruption)
10
CIFAR-10 homogeneous client partitions
80.57Accuracy
10
CIFAR-10
90.12Accuracy (ε=0.5)
10
CIFAR-10 (val)
92.15Accuracy
10
CIFAR-10 (Static)
96.18Accuracy
10
CIFAR-10 1% random data forgetting
39.2Utility Retention (UA)
10
CIFAR-10
100Unlearning Accuracy (UA)
10
CIFAR-10 Subclass airplane dinov2-mlp1 (test)
0KL Divergence (t)
10
CIFAR-100
81.55Clean Accuracy
10
CIFAR-10
0Activation Distance (Dr)
10
CIFAR-10
0Activation Distance (Dr)
10
CIFAR-10
93.71Accuracy
10
CIFAR-10
35.9FID
10
CIFAR-100 50 tasks
89.12Final Accuracy
10
CIFAR-10 (test)
6.11Error Rate
10
CIFAR-100
60.74Clean Accuracy
10
CIFAR-10 Dirichlet α = 0.01
0.97Average Pearson Correlation
10
CIFAR-10 (retained set D\Du)
100Macro F1 Score
10
CIFAR-10
99.8Retention Accuracy (RA)
10
CIFAR-10
99.74Retention Rate (RA)
10
CIFAR-100
99.98Retention Accuracy (RA)
10