OOD Detection on CIFAR-10 standard (test) vs Multiple OOD Sets
28.39Avg FPR95GradNorm
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GradNormBackbone=ResNet-202021.10 | 28.39 | 17.76 | 96.66 | 0.23 | 99.87 | 57.85 | 85.2 | 37.71 | 90.76 | 93.12 | — | — | |
| MahalanobisBackbone=ResNet-202021.10 | 37.16 | 20.91 | 95.99 | 9.66 | 97.9 | 89.24 | 61.15 | 28.83 | 92.31 | 86.84 | — | — | |
| GradOrthBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 38.22 | — | — | — | — | — | — | — | — | 91.64 | — | — | |
| ODINBackbone=ResNet-202021.10 | 39.12 | 55.52 | 89.63 | 2.32 | 99.39 | 45.86 | 90.81 | 52.78 | 89.99 | 92.46 | — | — | |
| Energy scoreBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 39.4 | — | — | — | — | — | — | — | — | 91.64 | — | — | |
| EnergyBackbone=ResNet-202021.10 | 39.7 | 49.8 | 91.97 | 3.86 | 99.03 | 46.48 | 90.55 | 58.67 | 88.79 | 92.59 | — | — | |
| VRA-PBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 39.98 | — | — | — | — | — | — | — | — | 91.69 | — | — | |
| ASH-PBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 40.46 | — | — | — | — | — | — | — | — | 91.76 | — | — | |
| ReActBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 43.31 | — | — | — | — | — | — | — | — | 91.03 | — | — | |
| ASH-SBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 48.45 | — | — | — | — | — | — | — | — | 88.34 | — | — | |
| ASH-BBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 48.47 | — | — | — | — | — | — | — | — | 89.93 | — | — | |
| DICEBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 48.59 | — | — | — | — | — | — | — | — | 89.13 | — | — | |
| GradNormBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 50.43 | — | — | — | — | — | — | — | — | 84.29 | — | — | |
| ODINBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 52.85 | — | — | — | — | — | — | — | — | 88.55 | — | — | |
| MSPBackbone=ResNet-202021.10 | 60.53 | 66.09 | 89.86 | 37.73 | 94.87 | 70.05 | 85.99 | 68.23 | 87.62 | 89.59 | — | — | |
| Softmax scoreBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 63.02 | — | — | — | — | — | — | — | — | 88.57 | — | — | |
| MahalanobisBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.08 | 85.14 | — | — | — | — | — | — | — | — | 63.15 | — | — | |
| Energy2026.02 | — | — | — | — | — | — | — | — | — | — | 61.1 | 91.13 | |
| GSC2026.02 | — | — | — | — | — | — | — | — | — | — | 59.59 | 86.35 | |
| kNN+2026.02 | — | — | — | — | — | — | — | — | — | — | 59.56 | 92.23 | |
| Mahalonobis2026.02 | — | — | — | — | — | — | — | — | — | — | 56.2 | 94.35 | |
| MSP2026.02 | — | — | — | — | — | — | — | — | — | — | 59.78 | 92.1 | |
| ODIN2026.02 | — | — | — | — | — | — | — | — | — | — | 54.78 | 92.22 | |
| TASTE2026.02 | — | — | — | — | — | — | — | — | — | — | 59.54 | 86.11 |