List your company
ResearchFormerly WizWand

Out-of-Distribution Detection on CIFAR100 (ID) SVHN (OOD)

99.87AUROCMedix
19.311640.225861.1482.0542May 27, 2024Sep 15, 2024Jan 4, 2025Apr 25, 2025Aug 14, 2025Dec 3, 2025Mar 24, 2026
Updated 3mo ago

Evaluation results

MethodLinks
MedixBackbone=ResNet-34, Performance averaged over five runs=true2025.1099.870.48——
Prototype FusionBackbone=DenseNet-1002026.0396.33——17.91
GramBackbone=ResNet182024.0595.5520.06——
Prototype FusionBackbone=ResNet-182026.0395.52——22.31
CIDERBackbone=ResNet-34, Performance averaged over five runs=true2025.1095.1623.09——
ESOODBackbone=DenseNet-1002026.0394.56——30.32
SSD+Backbone=ResNet-34, Performance averaged over five runs=true2025.1094.1931.19——
DRLBackbone=ResNet-34, Performance averaged over five runs=true2025.1094.0720.15——
ViMBackbone=DenseNet-1002026.0393.66——28.37
LaRExBackbone=DenseNet-1002026.0393.49——29.98
KNN+Backbone=ResNet-34, Performance averaged over five runs=true2025.1092.7839.23——
CSIBackbone=ResNet-34, Performance averaged over five runs=true2025.1092.6544.53——
NACBackbone=ResNet182024.0592.424.39——
LaRExBackbone=ResNet-182026.0390.45——36.09
CONJBackbone=ResNet-34, Performance averaged over five runs=true2025.1090.4446.19——
KNNBackbone=ResNet-34, Performance averaged over five runs=true2025.1090.3946.25——
ESOODBackbone=ResNet-182026.0390.19——35.45
ASHBackbone=ResNet-34, Performance averaged over five runs=true2025.1090.1952.96——
MahalanobisBackbone=DenseNet-1002026.0389.23——43.25
ReActBackbone=ResNet-34, Performance averaged over five runs=true2025.1088.7550.93——
ReActBackbone=DenseNet-1002026.0385.63——41.7
ASHBackbone=ResNet182024.0585.646——
EnergyBackbone=ResNet-34, Performance averaged over five runs=true2025.1085.2566.91——
ODINBackbone=ResNet-34, Performance averaged over five runs=true2025.1084.8870.16——
VimBackbone=ResNet-34, Performance averaged over five runs=true2025.1084.6273.42——
NECOBackbone=DenseNet-1002026.0383.85——47.19
MaxLogitBackbone=DenseNet-1002026.0383.45——48.09
EnergyBackbone=DenseNet-1002026.0383.44——48.1
NNGuideBackbone=DenseNet-1002026.0383.04——41.86
GradNormBackbone=DenseNet-1002026.0382.86——64.1
VOSBackbone=ResNet-34, Performance averaged over five runs=true2025.1082.843.24——
ProxyAnchorBackbone=ResNet-34, Performance averaged over five runs=true2025.1082.4387.21——
Entropy-MCMC2025.1081.14—87.18—
WeiPer+MSPBackbone=ResNet182024.0580.959.31——
MahalanobisBackbone=ResNet-34, Performance averaged over five runs=true2025.1080.6287.09——
fSGLD2025.1080.52—88.01—
ASAM2025.1079.86—87.93—
MSPBackbone=DenseNet-1002026.0379.82——52.74
MSPBackbone=ResNet-34, Performance averaged over five runs=true2025.1079.878.89——
Entropy-SGD2025.1079.15—86.92—
MahalanobisBackbone=ResNet-182026.0377.92——61.76
ViMBackbone=ResNet-182026.0376.72——68.15
NNGuideBackbone=ResNet-182026.0376.51——68.25
SAM2025.1074.56—84.61—
MSPBackbone=ResNet-182026.0372.92——83.01
SGLD2025.1072.51—83.35—
SGD2025.1071.96—84.08—
Entropy-SGLD2025.1071.83—82.89—
MaxLogitBackbone=ResNet-182026.0367.27——90.25
EnergyBackbone=ResNet-182026.0358.26——91.68
NECOBackbone=ResNet-182026.0345.93——91.06
ReActBackbone=ResNet-182026.0336.6——92.76
GradNormBackbone=ResNet-182026.0322.41——98.05
DAEDL2025.10——72.07—
Dropout2025.10——71.83—
EDL2025.10——56.21—
F-EDL2025.10——75.35—
I-EDL2025.10——67.51—
R-EDL2025.10——61.8—