Out-of-Distribution Detection on CIFAR-10 (ID) vs SVHN (OOD) (test)
100AUROCWAIC
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| WAICapproach_type=Generative-based2019.09 | 100 | — | — | — | |
| GramEvaluation Setting=Other settings, Oracle=false2020.12 | 100 | 98 | — | — | |
| nnPUEvaluation Setting=SSND, Oracle=true2020.12 | 100 | 100 | — | — | |
| MCDEvaluation Setting=SSND, Oracle=true2020.12 | 100 | 98 | — | — | |
| Binary ClassificationEvaluation Setting=SSND, Oracle=false2020.12 | 100 | 100 | — | — | |
| ERDEvaluation Setting=SSND, Oracle=false2020.12 | 100 | 100 | — | — | |
| CSI2021.03 | 99.8 | — | — | — | |
| SSDkk=52021.03 | 99.7 | — | — | — | |
| SSD2021.03 | 99.6 | — | — | — | |
| OursBackbone=ResNet2019.12 | 99.5 | 98.14 | 96.71 | — | |
| OursModel Architecture=ResNet2019.12 | 99.5 | 97.6 | 96.7 | — | |
| MahalanobisBackbone=ResNet2019.12 | 99.1 | 87.8 | 95.8 | — | |
| MahalanobisModel Architecture=ResNet2019.12 | 99.1 | 87.8 | 95.8 | — | |
| OursModel Architecture=DenseNet2019.12 | 99.1 | 96.1 | 95.9 | — | |
| Mahalanobisapproach_type=Classifier-based2019.09 | 99.1 | — | — | — | |
| GSDBackbone=ResNet-18, Variant=alpha-regularized, Scoring Criterion=Norm2021.10 | 99.05 | — | — | — | |
| MahalanobisEvaluation Setting=SSND, Oracle=true2020.12 | 99 | 96 | — | — | |
| Mahalanobis-UEvaluation Setting=SSND, Oracle=false2020.12 | 99 | 96 | — | — | |
| Outlier exposureapproach_type=Classifier-based2019.09 | 98.4 | — | — | — | |
| TEBackbone=ViT-B/162025.12 | 98.35 | — | — | — | |
| PS-KDBackbone=ViT-B/162025.12 | 98.28 | — | — | — | |
| BaselineBackbone=ViT-B/162025.12 | 98.21 | — | — | — | |
| GSDBackbone=ResNet-18, Variant=alpha-unregularized, Scoring Criterion=Norm2021.10 | 98.2 | — | — | — | |
| DLBBackbone=ViT-B/162025.12 | 98.15 | — | — | — | |
| MahalanobisModel Architecture=DenseNet2019.12 | 98.1 | 90.8 | 93.9 | — | |
| DPSBackbone=ViT-B/162025.12 | 98.03 | — | — | — | |
| Rotation-loss2021.03 | 97.9 | — | — | — | |
| GSDBackbone=ResNet-18, Variant=alpha-regularized, Scoring Criterion=Similarity2021.10 | 97.09 | — | — | — | |
| OEEvaluation Setting=Other settings, Oracle=false2020.12 | 97 | 89 | — | — | |
| ODINBackbone=ResNet2019.12 | 96.7 | 70.3 | 91.1 | — | |
| ODINModel Architecture=ResNet2019.12 | 96.7 | 70.3 | 91.1 | — | |
| ODINapproach_type=Classifier-based2019.09 | 96.6 | — | — | — | |
| DPNBackbone=ResNet2019.12 | 95.9 | 79.9 | 87.3 | — | |
| ODINModel Architecture=DenseNet2019.12 | 95.5 | 86.2 | 91.4 | — | |
| S scoregenerative_model=Glow, compression_algorithm=FLIF2019.09 | 95 | — | — | — | |
| DPNEvaluation Setting=Other settings, Oracle=false2020.12 | 95 | 85 | — | — | |
| DPSBackbone=ResNet-182025.12 | 94.84 | — | — | — | |
| GSDBackbone=ResNet-18, Variant=alpha-unregularized, Scoring Criterion=Similarity2021.10 | 94.72 | — | — | — | |
| SITN2026.05 | 94.6 | — | — | — | |
| FedRoD + FOOGDalpha=0.12024.10 | 94.56 | — | — | 17.61 | |
| TEBackbone=DenseNet-1212025.12 | 94.05 | — | — | — | |
| VanillaBackbone=ResNet-18, Scoring Criterion=Similarity2021.10 | 93.87 | — | — | — | |
| DPSBackbone=DenseNet-1212025.12 | 93.65 | — | — | — | |
| VDBackbone=ResNet2019.12 | 93.2 | 71.3 | 86.4 | — | |
| GSDBackbone=ResNet-18, Variant=alpha-regularized, Scoring Criterion=Norm2021.10 | 93.16 | — | — | — | |
| S scoregenerative_model=PixelCNN++, compression_algorithm=FLIF2019.09 | 92.9 | — | — | — | |
| TEBackbone=ResNet-182025.12 | 92.9 | — | — | — | |
| PS-KDBackbone=ResNet-182025.12 | 92.36 | — | — | — | |
| Vanilla EnsemblesEvaluation Setting=Other settings, Oracle=false2020.12 | 92 | 78 | — | — | |
| DLBBackbone=ResNet-182025.12 | 91.25 | — | — | — | |
| Likelihood-ratioapproach_type=Generative-based2019.09 | 91.2 | — | — | — | |
| DLBBackbone=DenseNet-1212025.12 | 91.19 | — | — | — | |
| PS-KDBackbone=DenseNet-1212025.12 | 91.11 | — | — | — | |
| BaselineBackbone=DenseNet-1212025.12 | 90.96 | — | — | — | |
| BaselineBackbone=ResNet-182025.12 | 90.67 | — | — | — | |
| VanillaBackbone=ResNet-18, Scoring Criterion=Norm2021.10 | 90.48 | — | — | — | |
| BaselineBackbone=ResNet2019.12 | 89.9 | 50.5 | 85.1 | — | |
| BaselineModel Architecture=ResNet2019.12 | 89.9 | 50.5 | 85.1 | — | |
| BaselineModel Architecture=DenseNet2019.12 | 89.9 | 40.2 | 83.2 | — | |
| GSDBackbone=ResNet-18, Variant=alpha-unregularized, Scoring Criterion=Norm2021.10 | 88.29 | — | — | — | |
| FedSDWCalpha=0.12025.11 | 88.08 | — | — | 35.46 | |
| FedAvg + FOOGDalpha=0.12024.10 | 87.63 | — | — | 44.59 | |
| FedTHEalpha=0.12024.10 | 85.95 | — | — | 39.22 | |
| FedTHEalpha=0.12025.11 | 85.95 | — | — | 39.22 | |
| FedICONalpha=0.12024.10 | 85.56 | — | — | 34.94 | |
| FedICONalpha=0.12025.11 | 85.56 | — | — | 34.94 | |
| FedRoDalpha=0.12024.10 | 83.55 | — | — | 40.72 | |
| FOSTERalpha=0.12024.10 | 83.07 | — | — | 39.55 | |
| FOSTERalpha=0.12025.11 | 83.07 | — | — | 39.55 | |
| FedATOLalpha=0.12024.10 | 82.17 | — | — | 85.39 | |
| FedATOLalpha=0.12025.11 | 82.17 | — | — | 85.39 | |
| DoSE2026.05 | 80.2 | — | — | — | |
| Typicality2026.05 | 79 | — | — | — | |
| PerAdaalpha=0.12025.11 | 78.34 | — | — | 47.31 | |
| FedLNalpha=0.12024.10 | 76.82 | — | — | 70.95 | |
| FedLNalpha=0.12025.11 | 76.82 | — | — | 70.95 | |
| Outlier exposureapproach_type=Generative-based2019.09 | 75.8 | — | — | — | |
| GSDBackbone=ResNet-18, Variant=alpha-regularized, Scoring Criterion=Similarity2021.10 | 74.82 | — | — | — | |
| VanillaBackbone=ResNet-18, Scoring Criterion=Norm2021.10 | 67.23 | — | — | — | |
| FedIIRalpha=0.12024.10 | 64.04 | — | — | 83.68 | |
| FedIIRalpha=0.12025.11 | 64.04 | — | — | 83.68 | |
| FedAvgalpha=0.12024.10 | 62.14 | — | — | 80.02 | |
| FedAvgalpha=0.12025.11 | 62.14 | — | — | 80.02 | |
| GSDBackbone=ResNet-18, Variant=alpha-unregularized, Scoring Criterion=Similarity2021.10 | 60.63 | — | — | — | |
| VanillaBackbone=ResNet-18, Scoring Criterion=Similarity2021.10 | 56.98 | — | — | — | |
| VIBapproach_type=Classifier-based2019.09 | 52.8 | — | — | — | |
| Typicality testapproach_type=Generative-based, batch_size=22019.09 | 42 | — | — | — | |
| PixelCNN++2021.03 | 15.8 | — | — | — | |
| Deep-SVDD2021.03 | 14.5 | — | — | — | |
| Log-likelihood2026.05 | 10.5 | — | — | — | |
| Autoencoder2021.03 | 2.5 | — | — | — | |
| VAE2021.03 | 2.4 | — | — | — |