Misclassification Detection

Benchmarks

Dataset Name	SOTA Method	Metric
CIFAR-10	OpenMix	AUROC94.81	28	4d ago
SMS Spam class		AUROC0.982	27	4d ago
CIFAR-100	ODIN	AURC167.53	27	4d ago
SMS Non-Spam class		AUROC0.9369	26	4d ago
MNIST (test)	Softmax Baseline	AUROC0.97	13	4d ago
Fashion-MNIST (test)	MC-Dropout	ROC-AUC0.818	12	4d ago
CIFAR-100 (test)	D_alpha	AUROC88.2	11	4d ago
CIFAR-10 (test)	D_alpha	AUROC95.2	11	4d ago
SVHN (test)	D_alpha	AUROC (%)93	10	4d ago
Tiny ImageNet (test)	D_alpha	AUROC86.1	10	4d ago
CIFAR-100-C 1.0 (test)	OpenMix	AUROC84.05	9	4d ago
CIFAR-10-C 1.0 (test)	OpenMix	AUROC90.38	9	4d ago
DMNIST	F-EDL	AUPR (Conf)96.17	8	4d ago
ImageNet-1K (val)	sample difficulty-aware entropy regularization	FPR@95% (MSP)45.69	7	4d ago
CIFAR-10-LT (ρ = 0.1)	F-EDL	AUPR (Confidence)97.6	6	4d ago
CIFAR-10-LT (ρ = 0.01)	F-EDL	AUPR (Confidence)85.99	6	4d ago
IMDB (test)	D_alpha	AUROC0.844	3	4d ago
AMAZON SOFTWARE (test)	D_alpha	AUROC0.688	3	4d ago
AMAZON FASHION (test)	D_alpha	AUROC (%)89.7	3	4d ago
CIFAR100	D_alpha	AUROC88.2	2	4d ago
CIFAR10	D_alpha	AUROC95.2	2	4d ago
Twitter POS-annotated tweets (test)	Softmax Baseline	AUROC89	2	4d ago
WSJ (Penn Treebank) (test)	Softmax Baseline	AUROC0.96	2	4d ago
Reuters 40 subset of 52 (test)	Softmax Baseline	AUROC91	1	4d ago
Reuters 6 subset of Reuters 8 (test)	Softmax Baseline	AUROC89	1	4d ago

Showing 25 of 26 rows