Out-of-Distribution Detection on CIFAR-10 (ID) vs Texture (OOD)
99.9AUROCDynProto
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DynProtoBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 99.9 | — | 0.16 | — | |
| ZODE-KNN2022.12 | 99.88 | — | — | 0.16 | |
| AdaNDBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 99.63 | — | 0.55 | — | |
| CSPBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 99.21 | — | 3.3 | — | |
| ZODE-Mahalanobis2022.12 | 99.12 | — | — | 3.88 | |
| NegRefineBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 98.68 | — | 6.38 | — | |
| ResNet18*Backbone=ResNet18*2022.12 | 98.57 | — | — | 8.09 | |
| KNN+2022.12 | 98.56 | — | — | 8.09 | |
| SSD+Contrastive Learning=Yes2026.01 | 98.35 | — | 9.27 | — | |
| SSD+2022.12 | 97.7 | — | — | 12.98 | |
| KNN+Contrastive Learning=Yes2026.01 | 97.43 | — | 10.11 | — | |
| NegLabelBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 97.38 | — | 13.51 | — | |
| GLMCMBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 97.18 | — | 11.26 | — | |
| ResNet101Backbone=ResNet1012022.12 | 96.89 | — | — | 18.42 | |
| CIDERContrastive Learning=Yes2026.01 | 96.85 | — | 12.33 | — | |
| CADRefBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 96.71 | — | 17.54 | — | |
| ResNet152Backbone=ResNet1522022.12 | 96.6 | — | — | 20.76 | |
| ResNet50Backbone=ResNet502022.12 | 96.59 | — | — | 20.85 | |
| CE + SimCLRContrastive Learning=Yes2026.01 | 96.56 | — | 16.77 | — | |
| Reg. MahalanobisBackbone=ResNet182026.02 | 96.41 | 97.58 | — | — | |
| DenseNetBackbone=DenseNet2022.12 | 96.25 | — | — | 20.78 | |
| VIMBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 96.19 | — | 20.37 | — | |
| COMBOODBackbone=ResNet182026.02 | 95.98 | 97.66 | — | — | |
| MCMBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 95.61 | — | 20.14 | — | |
| GramBackbone=ResNet182026.02 | 95.42 | 96.69 | — | — | |
| ASH-SBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 95.29 | — | 23.6 | — | |
| ZODE-Energy2022.12 | 95.14 | — | — | 37.34 | |
| ResNet18Backbone=ResNet182022.12 | 94.97 | — | — | 26.74 | |
| DynProtoBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 94.9 | — | 21.05 | — | |
| CSIContrastive Learning=Yes2026.01 | 94.87 | — | 28.85 | — | |
| CSI2022.12 | 94.87 | — | — | 28.85 | |
| KNN2022.12 | 94.71 | — | — | 27.57 | |
| ZODE-MSP2022.12 | 94.68 | — | — | 43.16 | |
| ResNet34Backbone=ResNet342022.12 | 94.53 | — | — | 31.65 | |
| OptFSBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 94.25 | — | 33.46 | — | |
| KNNBackbone=ResNet182026.02 | 93.16 | 87.26 | — | — | |
| ProxyAnchorContrastive Learning=Yes2026.01 | 93.16 | — | 42.7 | — | |
| MahalanobisContrastive Learning=No2026.01 | 92.91 | — | 23.21 | — | |
| Mahalanobis2022.12 | 92.91 | — | — | 23.21 | |
| GODINContrastive Learning=No2026.01 | 92.2 | — | 33.58 | — | |
| KNNContrastive Learning=No2026.01 | 89.93 | — | 47.84 | — | |
| GODIN2022.12 | 89.69 | — | — | 46.91 | |
| ODINContrastive Learning=No2026.01 | 89.47 | — | 55.59 | — | |
| EnergyContrastive Learning=No2026.01 | 89.37 | — | 55.23 | — | |
| Energy2022.12 | 89.37 | — | — | 55.23 | |
| MSPContrastive Learning=No2026.01 | 88.5 | — | 66.45 | — | |
| MSP2022.12 | 88.5 | — | — | 66.45 | |
| MSPBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 88.24 | — | 63.88 | — | |
| DICEBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 87.33 | — | 45.34 | — | |
| Comp.VAEContrastive Learning=No2026.01 | 86.9 | — | 32.2 | — | |
| EnergyBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 86.56 | — | 56.76 | — | |
| ODIN2022.12 | 86.21 | — | — | 56.4 | |
| ODINBackbone=ResNet182026.02 | 80.7 | 82.25 | — | — | |
| Comp.VAE (vMF)Contrastive Learning=No2026.01 | 76.1 | — | 72.3 | — | |
| ReActBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 68.08 | — | 90.87 | — | |
| MDSBackbone=ResNet182026.02 | 57.72 | 62.75 | — | — | |
| KNN*Contrastive Learning=No2026.01 | 55.7 | — | 88.9 | — |