List your company
ResearchFormerly WizWand

OOD Detection on CIFAR-10 (IND) vs SVHN (OOD)

0.9998AUROCMedix
0.931680.9493650.967050.984735Oct 18, 2021Jul 25, 2022May 1, 2023Feb 5, 2024Nov 11, 2024Aug 18, 2025May 25, 2026
Updated 3mo ago

Evaluation results

MethodLinks
MedixTraining Data=Pin and Pwild2025.100.9998——0.06
WOODSTraining Data=Pin and Pwild2025.100.9991——0.17
GradPCA-BatchBackbone=ResNetV2-50 (BiT-M)2025.050.9973——0.52
PALMMethod type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.050.997——124
Energy + pNMLBackbone=WideResNet-402021.100.99698.997.6—
Hopfield BoostingMethod type=OE, Augmentations=Weak, Auxiliary outlier data=true, Backbone=ResNet-182024.050.9957——23
OETraining Data=Pin and Pwild2025.100.9953——1.13
KNNBackbone=ResNetV2-50 (BiT-M)2025.050.9942——1.99
SSD+Method type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.050.9941——305
KNN+Training Data=Pin only2025.100.9941——2.99
EnergyBackbone=WideResNet-402021.100.99398.396.9—
SPEM-noiseDensity Estimator=Glow2026.020.9916———
GradPCA (block 4)Backbone=ResNetV2-50 (BiT-M)2025.050.9901——5.11
GAIA-ZBackbone=ResNet-34 (TIMM)2025.050.9879——4.39
GradPCA-VecBackbone=ResNetV2-50 (BiT-M)2025.050.9877——6.08
GradOrthBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9872——5.84
Energy (w/ OE)Training Data=Pin and Pwild2025.100.9872——5.24
Kernel PCA (CoRP)Backbone=ResNetV2-50 (BiT-M)2025.050.987——6.17
GradPCABackbone=ResNetV2-50 (BiT-M)2025.050.9867——6.92
ASH-SBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9865——6.51
ASHTraining Data=Pin only2025.100.9865——6.51
NPOSMethod type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.050.9837——904
MahalanobisBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9831——6.42
MahalanobisBackbone=ResNetV2-50 (BiT-M)2025.050.9819——9.07
GradPCA+DICEBackbone=ResNetV2-50 (BiT-M)2025.050.9796——10.98
Bootstrap DistillUncertainty Measure=Differential Entropy (Dent)2024.020.978———
SPEMDensity Estimator=Glow2026.020.9768———
MahalanobisTraining Data=Pin only2025.100.9762——12.89
BMUncertainty Measure=Mutual Information (MI)2024.020.975———
CreDROEnsemble Number M=202026.020.974———
CreDRO2026.020.974———
CSITraining Data=Pin only2025.100.974——17.3
Bootstrap DistillUncertainty Measure=Mutual Information (MI)2024.020.972———
CreDROEnsemble Number M=152026.020.971———
GradPCA (block 3)Backbone=ResNet-34 (TIMM)2025.050.9701——15.07
BMUncertainty Measure=Differential Entropy (Dent)2024.020.97———
CreDROEnsemble Number M=102026.020.97———
END2Uncertainty Measure=Differential Entropy (Dent)2024.020.969———
ASH-BBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9686——17.92
VRA-PBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9668——18.75
CreDROEnsemble Number M=52026.020.966———
EDLUncertainty Measure=Differential Entropy (Dent)2024.020.965———
EDLUncertainty Measure=Mutual Information (MI)2024.020.962———
CreWraEnsemble Number M=152026.020.961———
KNNTraining Data=Pin only2025.100.9596——24.53
GradPCA-BatchBackbone=ResNet-34 (TIMM)2025.050.9595——24.9
GradPCA-VecBackbone=ResNet-34 (TIMM)2025.050.9591——25.34
DICEBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.959——25.99
END2Uncertainty Measure=Mutual Information (MI)2024.020.958———
EN-DROEnsemble Number M=202026.020.957———
CreWraEnsemble Number M=202026.020.957———
EN-DRO2026.020.957———
CreWra2026.020.957———
PostNetUncertainty Measure=Differential Entropy (Dent)2024.020.955———
CreEns0.0Ensemble Number M=202026.020.955———
CreEnsparameter=0.02026.020.955———
CreEns0.0Ensemble Number M=152026.020.954———
DICEBackbone=ResNet-34 (TIMM)2025.050.9538——16.01
EN-DROEnsemble Number M=52026.020.953———
DEEnsemble Number M=152026.020.953———
ASH-PBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9529——30.14
NCIBackbone=ResNet-34 (TIMM)2025.050.9522——23.65
EN-DROEnsemble Number M=152026.020.952———
EnergyBackbone=ResNet-34 (TIMM)2025.050.9516——19.15
GAIA-ABackbone=ResNet-34 (TIMM)2025.050.9512——25.07
EN-DROEnsemble Number M=102026.020.951———
Max logitsBackbone=ResNet-34 (TIMM)2025.050.9506——20.07
ODINBackbone=ResNet-34 (TIMM)2025.050.9506——20.07
CreEns0.0Ensemble Number M=102026.020.949———
ASHMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.050.9486——2,517
Revisited PCABackbone=ResNet-34 (TIMM)2025.050.9485——24.08
DEEnsemble Number M=202026.020.948———
CreRL1.0Ensemble Number M=202026.020.948———
DE2026.020.948———
CreRLparameter=1.02026.020.948———
CreWraEnsemble Number M=102026.020.946———
CreRL1.0Ensemble Number M=152026.020.946———
ODINBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9457——25.29
PostNetUncertainty Measure=Mutual Information (MI)2024.020.944———
CreRL1.0Ensemble Number M=52026.020.943———
CreDEEnsemble Number M=202026.020.943———
CreDE2026.020.943———
KNNBackbone=ResNet-34 (TIMM)2025.050.943——37.25
NCIBackbone=ResNetV2-50 (BiT-M)2025.050.9425——23.15
GradNormBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9411——18.63
CreDEEnsemble Number M=152026.020.941———
CreEns0.0Ensemble Number M=52026.020.94———
Energy scoreBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9399——40.61
MSPBackbone=ResNet-34 (TIMM)2025.050.9399——37.47
MoDEXsetting=standard, uncertainty_estimate=epistemic2026.050.9397———
ReActBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9387——41.64
CreWraEnsemble Number M=52026.020.938———
CreDEEnsemble Number M=102026.020.938———
F-EDLArchitecture=VGG-16, Uncertainty Type=epistemic, Setting=classical2025.100.9374———
F-EDLsetting=standard, uncertainty_estimate=epistemic2026.050.9374———
Fisher-EDLUncertainty Measure=Differential Entropy (Dent)2024.020.937———
CreRL1.0Ensemble Number M=102026.020.936———
GENMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.050.9353——3,326
Softmax scoreBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.080.9348——47.24
EBOMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.050.9343——3,210