Out-of-distribution Detection on OOD Suite Mean
43.74FPR@95Clipped HNN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Clipped HNNTraining Dataset=CIFAR10, Scoring Function=softmax2021.07 | 43.74 | 91.38 | 97.87 | |
| ENNTraining Dataset=CIFAR10, Scoring Function=softmax2021.07 | 49.56 | 89.53 | 97.36 | |
| PT KNNBackbone=CE DinoV22025.12 | 90 | 99.6 | — | |
| PT KNNBackbone=CE Resnet2025.12 | 90 | 99.6 | — | |
| PT KNNBackbone=SC Resnet2025.12 | 90 | 99.6 | — | |
| DF + KNNBackbone=CE Resnet2025.12 | 250 | 99.2 | — | |
| DF + MahalanobisBackbone=CE DinoV22025.12 | 290 | 99 | — | |
| DF + KNNBackbone=CE DinoV22025.12 | 320 | 99 | — | |
| DF + KNNBackbone=SC Resnet2025.12 | 320 | 99 | — | |
| DF + ReActBackbone=CE DinoV22025.12 | 370 | 98.8 | — | |
| DF + ReActBackbone=CE Resnet2025.12 | 410 | 99.1 | — | |
| DF + MahalanobisBackbone=SC Resnet2025.12 | 1,160 | 95.2 | — | |
| DF + MahalanobisBackbone=CE Resnet2025.12 | 1,180 | 96.3 | — | |
| MahalanobisBackbone=CE Resnet2025.12 | 1,620 | 94.4 | — | |
| MahalanobisBackbone=CE DinoV22025.12 | 1,850 | 93.4 | — | |
| KNNBackbone=CE DinoV22025.12 | 2,540 | 91 | — | |
| KNNBackbone=CE Resnet2025.12 | 2,580 | 91.1 | — | |
| KNNBackbone=SC Resnet2025.12 | 3,290 | 87.8 | — | |
| MahalanobisBackbone=SC Resnet2025.12 | 3,470 | 87.6 | — | |
| NCIBackbone=CE DinoV22025.12 | 3,530 | 86.6 | — | |
| NCIBackbone=CE Resnet2025.12 | 3,610 | 88.5 | — | |
| ReActBackbone=CE DinoV22025.12 | 3,640 | 86.3 | — | |
| EnergyBackbone=CE DinoV22025.12 | 3,730 | 85.6 | — | |
| MSPBackbone=CE Resnet2025.12 | 3,890 | 87.4 | — | |
| EnergyBackbone=CE Resnet2025.12 | 4,140 | 87.6 | — | |
| MSPBackbone=CE DinoV22025.12 | 4,300 | 82 | — | |
| ReActBackbone=CE Resnet2025.12 | 4,770 | 84.9 | — |