CIFAR-10
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
CIFAR-10 most-memorized (forget top 5%)
93.5AUC
5
CIFAR-10 Split (val)
87Accuracy (Val)
5
CIFAR-10 IR=100 long-tail (test)
68.07Clean Accuracy
5
CIFAR-10 long-tail (IR=20) (test)
78.49Accuracy (Clean)
5
CIFAR-10 (test)
86Concept Accuracy
5
CIFAR-10 (test)
46.53Accuracy (Exit 1)
5
CIFAR-10 (test)
71.84AP
5
CIFAR-10 (test)
95.9Accuracy (10% Noise)
5
CIFAR-10-C severity 4
67.8Accuracy
5
CIFAR-10-C (severity 3)
75.2Accuracy
5
CIFAR-10-C level 3 (test)
7.9Error Rate (Original)
5
CIFAR-10 SVHN open-set noise
80.62Average Accuracy
5
CIFAR-10 40% noise fraction (val)
89Accuracy (Val)
5
CIFAR-10 Clean original (test)
6.02AURC
4
CIFAR-10 N=20,000
0.06MAE CWCE (Equal Binning 15)
4
CIFAR-10 Order 2: imbalanced class increments (test)
44.45AP
4
long-tailed CIFAR-10 with asymmetric noise (test)
26.47Recall
4
CIFAR-10 PUE (test)
90.09Accuracy
4
CIFAR-10 EntF (test)
87.27Accuracy (Test)
4
CIFAR-10 Error-minimizing (test)
85.26Accuracy
4
CIFAR-10 NTGA (test)
87.38Accuracy
4
CIFAR-10 gray
34.6Clean Accuracy
4
CIFAR-10 (test)
90.1Robust Accuracy
4
CIFAR-10 (val)
81.96Average Validation Accuracy
4
CIFAR-10 L2 RobustBench (test)
0.897Spearman Rank Correlation
4