CIFAR-100
Benchmarks
TaskDatasetSOTA resultTrendResults
CIFAR-100 (train)
0.004Training Loss
30
CIFAR-100 Step-Imbalance
59.54Accuracy
29
CIFAR-100-C
58.2Accuracy (Noise)
28
CIFAR-100 ID (test)
81.66Accuracy
28
CIFAR-100-LT Imbalance ratio 50
57.75Top-1 Acc
28
CIFAR-100
0.355MAE
27
CIFAR-100
0.799ACC (Seen)
27
CIFAR-100
84.2Top-1 Accuracy
26
CIFAR-100N
74.5Accuracy
26
CIFAR-100 T=20 (test)
87.62Final Accuracy
25
CIFAR-100 Noisy (test)
75.2Top-1 Accuracy
25
CIFAR-100 From Half LT (test)
38.5Accuracy
24
CIFAR-100 ID Far-OOD Average OpenOOD v1.5
96.9AUROC
24
CIFAR-100N ResNet-18 (test)
79.3Clean Accuracy
24
CIFAR-100 Split 10 sequential tasks (test)
0Final Forgetting (FF)
24
CIFAR-100 pre-trained on CIFAR-10 (test)
47.92AA
24
CIFAR-100 Split
0.73Forgetting Measure
24
CIFAR-100 2 Types
24Failure Rate
24
CIFAR-100 2500 labels (test)
23.93Error Rate
24
CIFAR-100 80% symmetric noise (test)
73.8Accuracy
24
CIFAR-100 synthetic label noise (test)
74.67Test Acc
24
CIFAR-100 Noisy
75.98Accuracy
24
CIFAR-100 Seq
29.94Average Forgetting
23
CIFAR-100 Cross-Silo, alpha=0.5
39.61Top-1 Accuracy
22
CIFAR-100 50% symmetric noise (test)
0.984AUROC
22