OOD Detection on ImageNet OOD Average (iNaturalist, SUN, Places, Textures)
6.72Mean FPR95 (OOD Avg)TINS
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| TINSBackbone=CLIP ViT-B/16, Category=Test-time Adaptation Methods2026.05 | 6.72 | — | — | 10.09 | 97.83 | 3.84 | 99.14 | 12.73 | 97.14 | 0.21 | 99.93 | 98.51 | |
| InterNegBackbone=CLIP ViT-B/16, Category=Test-time Adaptation Methods2026.05 | 14.04 | — | — | 21.85 | 96.26 | 6.78 | 98.68 | 27.11 | 95.01 | 0.4 | 99.79 | 97.43 | |
| CSPBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=no training on ID or extra data2026.05 | 17.51 | — | — | 25.52 | 93.86 | 13.66 | 96.66 | 29.32 | 92.9 | 1.54 | 99.6 | 95.76 | |
| OODDBackbone=CLIP ViT-B/16, Category=Test-time Adaptation Methods2026.05 | 18.79 | — | — | 30.67 | 94.51 | 12.94 | 97.17 | 30.68 | 92.51 | 0.85 | 99.79 | 96 | |
| AdaNegBackbone=CLIP ViT-B/16, Category=Test-time Adaptation Methods2026.05 | 18.92 | — | — | 31.27 | 94.93 | 9.5 | 97.44 | 34.34 | 94.55 | 0.59 | 99.71 | 96.66 | |
| CLIPScopeBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=no training on ID or extra data2026.05 | 20.88 | — | — | 38.37 | 91.41 | 15.56 | 96.77 | 28.45 | 93.54 | 1.29 | 99.61 | 95.3 | |
| NegLabelBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=no training on ID or extra data2026.05 | 25.4 | — | — | 43.56 | 90.22 | 20.53 | 95.49 | 35.59 | 91.64 | 1.91 | 99.49 | 94.21 | |
| LoCoOpBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=training on ID or extra data2026.05 | 28.66 | — | — | 42.28 | 90.19 | 23.44 | 95.07 | 32.87 | 91.98 | 16.05 | 96.86 | 93.52 | |
| ID-LikeBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=training on ID or extra data2026.05 | 30.07 | — | — | 25.27 | 94.32 | 42.03 | 91.64 | 44 | 90.57 | 8.98 | 98.19 | 93.68 | |
| EOEBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=no training on ID or extra data2026.05 | 30.09 | — | — | 57.53 | 85.64 | 20.4 | 95.73 | 30.16 | 92.95 | 12.29 | 97.52 | 92.96 | |
| LSNBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=training on ID or extra data2026.05 | 30.22 | — | — | 38.54 | 90.42 | 26.32 | 94.35 | 34.48 | 91.25 | 21.56 | 95.83 | 92.96 | |
| ReActBackbone=RegNet2026.03 | 31.08 | — | — | — | — | — | — | — | — | — | — | 92.92 | |
| Hopfield BoostingBackbone=ResNet-502024.05 | 36.6 | — | — | 44.59 | 88.01 | 37.37 | 91.24 | 53.31 | 87.1 | 11.11 | 97.65 | 91 | |
| Prototype FusionBackbone=RegNet2026.03 | 37.43 | — | — | — | — | — | — | — | — | — | — | 89.4 | |
| EnergyBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 39.89 | — | — | 57.61 | 86.76 | 35.97 | 92.66 | 39.87 | 91.41 | 26.12 | 95.33 | 91.54 | |
| VOSBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 41.32 | — | — | 61.02 | 86.33 | 36.88 | 92.57 | 38.39 | 91.23 | 28.99 | 94.62 | 91.19 | |
| KNNBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 42.19 | — | — | 64.35 | 85.67 | 35.62 | 92.67 | 39.61 | 91.02 | 29.17 | 94.52 | 90.97 | |
| LareXBackbone=RegNet2026.03 | 42.35 | — | — | — | — | — | — | — | — | — | — | 90.49 | |
| Prototype FusionBackbone=ResNet-502026.03 | 43.02 | — | — | — | — | — | — | — | — | — | — | 87 | |
| MCMBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=no training on ID or extra data2026.05 | 43.93 | — | — | 58.5 | 86.12 | 38.8 | 92.25 | 46.2 | 90.31 | 32.2 | 94.59 | 90.82 | |
| LareXBackbone=ResNet-502026.03 | 45.62 | — | — | — | — | — | — | — | — | — | — | 88.99 | |
| ODINBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 47.75 | — | — | 51.67 | 87.85 | 54.04 | 87.17 | 55.06 | 85.54 | 30.22 | 94.65 | 88.8 | |
| ESOODBackbone=RegNet2026.03 | 50.13 | — | — | — | — | — | — | — | — | — | — | 83.55 | |
| ViMBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 50.2 | — | — | 53.94 | 87.18 | 54.01 | 87.19 | 60.67 | 83.75 | 32.19 | 93.16 | 87.82 | |
| POEMBackbone=ResNet-502024.05 | 50.74 | — | — | 31.26 | 92.22 | 57.46 | 85.38 | 74.58 | 78.89 | 45.37 | 92.01 | 87.85 | |
| MahalanobisBackbone=ResNet-502026.03 | 50.9 | — | — | — | — | — | — | — | — | — | — | 83.62 | |
| NNGuideBackbone=RegNet2026.03 | 51.01 | — | — | — | — | — | — | — | — | — | — | 84.99 | |
| DivOEBackbone=ResNet-502024.05 | 51.35 | — | — | 42.8 | 88.18 | 61 | 83.64 | 57.69 | 85.83 | 30.51 | 93.81 | 86.49 | |
| EBO-OEBackbone=ResNet-502024.05 | 51.6 | — | — | 29.67 | 92.4 | 57.69 | 85.83 | 71.09 | 80.35 | 49.02 | 91.44 | 87.75 | |
| NECOBackbone=RegNet2026.03 | 53.69 | — | — | — | — | — | — | — | — | — | — | 84.67 | |
| DOSBackbone=ResNet-502024.05 | 54.71 | — | — | 40.29 | 89.88 | 59.29 | 84.3 | 69.72 | 81.62 | 49.55 | 90.49 | 86.57 | |
| MSPBackbone=RegNet2026.03 | 55.17 | — | — | — | — | — | — | — | — | — | — | 83.86 | |
| EnergyBackbone=RegNet2026.03 | 55.18 | — | — | — | — | — | — | — | — | — | — | 84.57 | |
| MixOEBackbone=ResNet-502024.05 | 56.2 | — | — | 41.05 | 88.51 | 65.14 | 82.2 | 70.03 | 81.35 | 47.28 | 90.19 | 85.3 | |
| NNGuideBackbone=ResNet-502026.03 | 57.53 | — | — | — | — | — | — | — | — | — | — | 82.53 | |
| DALBackbone=ResNet-502024.05 | 58.9 | — | — | 43.88 | 87.39 | 65.31 | 81.47 | 66.01 | 81.45 | 51.92 | 88.33 | 83.98 | |
| MaxLogitBackbone=RegNet2026.03 | 59.81 | — | — | — | — | — | — | — | — | — | — | 78.5 | |
| MSP-OEBackbone=ResNet-502024.05 | 60.17 | — | — | 48.38 | 86.25 | 66.01 | 81.79 | 74.46 | 78.72 | 51.73 | 88.51 | 83.78 | |
| ESOODBackbone=ResNet-502026.03 | 68.95 | — | — | — | — | — | — | — | — | — | — | 76.94 | |
| GradNormBackbone=RegNet2026.03 | 69.41 | — | — | — | — | — | — | — | — | — | — | 79.78 | |
| MSPBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 69.61 | — | — | 71.93 | 79.69 | 73.72 | 79.73 | 74.41 | 79.67 | 58.36 | 87.44 | 81.63 | |
| ViMBackbone=RegNet2026.03 | 72.25 | — | — | — | — | — | — | — | — | — | — | 64.58 | |
| MahalanobisBackbone=RegNet2026.03 | 76.03 | — | — | — | — | — | — | — | — | — | — | 61.48 | |
| MSPBackbone=ResNet-502026.03 | 77.13 | — | — | — | — | — | — | — | — | — | — | 79.04 | |
| GradNormBackbone=CLIP ViT-B/16, Category=Visual-based, Training Requirement=training on ID or extra data2026.05 | 80.82 | — | — | 79.36 | 70.26 | 82 | 72.86 | 80.41 | 73.7 | 81.5 | 72.56 | 72.35 | |
| DOEBackbone=ResNet-502024.05 | 81.17 | — | — | 83.83 | 64.22 | 83.73 | 72.95 | 86.3 | 70.37 | 70.82 | 83.82 | 72.84 | |
| EnergyBackbone=CLIP ViT-B/16, Category=VLM-based, Training Requirement=no training on ID or extra data2026.05 | 82.21 | — | — | 93.65 | 65.56 | 79.02 | 84.24 | 75.08 | 83.38 | 81.08 | 85.09 | 79.57 | |
| ViMBackbone=ResNet-502026.03 | 82.72 | — | — | — | — | — | — | — | — | — | — | 58.73 | |
| MaxLogitBackbone=ResNet-502026.03 | 84.98 | — | — | — | — | — | — | — | — | — | — | 74.23 | |
| EnergyBackbone=ResNet-502026.03 | 90.44 | — | — | — | — | — | — | — | — | — | — | 50.3 | |
| GradNormBackbone=ResNet-502026.03 | 92.84 | — | — | — | — | — | — | — | — | — | — | 32.38 | |
| NECOBackbone=ResNet-502026.03 | 95.37 | — | — | — | — | — | — | — | — | — | — | 31 | |
| ReActBackbone=ResNet-502026.03 | 95.9 | — | — | — | — | — | — | — | — | — | — | 26.55 | |
| DICEBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 34.75 | 90.78 | — | — | — | — | — | — | — | — | — | |
| EnergyBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 58.4 | 86.17 | — | — | — | — | — | — | — | — | — | |
| EnergyBackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 71.03 | 82.74 | — | — | — | — | — | — | — | — | — | |
| KNNBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 54.32 | 84.59 | — | — | — | — | — | — | — | — | — | |
| MahalanobisBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 87.43 | 55.47 | — | — | — | — | — | — | — | — | — | |
| MahalanobisBackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 81.69 | 62.02 | — | — | — | — | — | — | — | — | — | |
| MOSBackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 39.97 | 90.11 | — | — | — | — | — | — | — | — | — | |
| MSPBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 66.95 | 81.99 | — | — | — | — | — | — | — | — | — | |
| MSPBackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 76.96 | 79.54 | — | — | — | — | — | — | — | — | — | |
| ODINBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 56.48 | 85.41 | — | — | — | — | — | — | — | — | — | |
| ODINBackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 72.99 | 82.56 | — | — | — | — | — | — | — | — | — | |
| ReActBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 31.43 | 92.95 | — | — | — | — | — | — | — | — | — | |
| ReActBackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 67.3 | 83.27 | — | — | — | — | — | — | — | — | — | |
| SHEBackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 44.71 | 88.24 | — | — | — | — | — | — | — | — | — | |
| VRABackbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 25.49 | 94.57 | — | — | — | — | — | — | — | — | — | |
| VRABackbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 34.95 | 93.34 | — | — | — | — | — | — | — | — | — | |
| VRA+Backbone=ResNet-50, Pre-trained=ImageNet2023.02 | — | 23.31 | 94.97 | — | — | — | — | — | — | — | — | — | |
| VRA+Backbone=ResNetv2-101, Pre-trained=ImageNet2023.02 | — | 30.85 | 93.71 | — | — | — | — | — | — | — | — | — |