Benchmarks

Task NameDataset NameSOTA ResultTrendResults

Class Incremental Learning

CIFAR-100 B5 Inc5

91.32Avg Performance (A-bar)

Image Generation

CIFAR-10 32x32 with ReFlow

30.4FID

Image Classification

CIFAR-10

0.01Latency (ms)

Risky Sample Transferability

CIFAR-100

41.1Error Rate

Poisoning Attack Control

CIFAR10 Conv8 (test)

0.06Control Error

Deep Clustering

CIFAR-10

0.863SC

Out-of-Distribution Detection

CIFAR-10-C

76.63AUROC

Image Synthesis

CIFAR-100 (train test)

11.37IS

Image Synthesis

CIFAR-10 (train test)

9.52Inception Score

Certified Robustness

CIFAR-10 l2 (test)

768,560Average Sample Count

OOD Detection

CIFAR100 ID Corrupted OOD

73.4AUROC

Image Classification

CIFAR-10 (test)

71Overall Accuracy

Image Classification

CIFAR10

16.52Speedup

Image Classification

CIFAR-100 1x10^6 EDM-generated images-augmented (train)

90.17Cleantr

Image Classification

CIFAR-100 1x10^6 EDM-generated images-augmented (test)

66.62Cleante Accuracy

Image Classification

CIFAR-10 DM-AT (train)

96.58Clean Accuracy

Image Classification

CIFAR-100 1.0 (val)

79.42Top-1 Acc

Image Classification

CIFAR-100 (test)

1.731Log Likelihood

Image Classification

CIFAR-10 (test)

0.329Log Likelihood

Open-set recognition

CIFAR+50 1.0 (4 closed CIFAR10 classes, 50 open CIFAR100 classes)

98.5AUROC

Class Unlearning

CIFAR-10

13.98Test Error

Out-of-Distribution Detection

CIFAR10 (In-distribution) vs SVHN (OOD) (test)

91.61AUPR

Image Classification

CIFAR-10 (test)

79.44Accuracy

Image Classification

CIFAR-100

51.84Standard Accuracy

Image Classification

CIFAR-10 (test)

88.9Clean Accuracy