Out-of-Distribution Detection on FMNIST vs. MNIST
99.6AUROC (%)Maha
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MahaArchitecture=LeNet2020.11 | 99.6 | 99.7 | 0 | — | |
| OEArchitecture=LeNet2020.11 | 99.6 | 99.6 | 0 | — | |
| FSSDArchitecture=LeNet2020.11 | 99.6 | 99.7 | 0 | — | |
| DiffPath-6DCategory=Post-Hoc Statistics2026.02 | 99.1 | — | — | — | |
| Diverging FlowsCategory=Active Training, Variant=Transformations2026.02 | 98.7 | — | — | — | |
| DiffPath-1DCategory=Post-Hoc Statistics2026.02 | 98.1 | — | — | — | |
| DIMSBackbone=LeNet, Hessian Approximation=low-rank, Parameter samples=100, friction coefficient=0.52026.05 | 97.1 | — | — | — | |
| ODINArchitecture=LeNet2020.11 | 96.9 | 93 | 2.5 | — | |
| Diverging FlowsCategory=Active Training, Variant=PGD2026.02 | 95.5 | — | — | — | |
| RLABackbone=LeNet, Hessian Approximation=low-rank, Parameter samples=1002026.05 | 93.9 | — | — | — | |
| DEArchitecture=LeNet2020.11 | 83.9 | 83.3 | 27.5 | — | |
| LINLABackbone=LeNet, Hessian Approximation=low-rank, Parameter samples=1002026.05 | 82.9 | — | — | — | |
| MCDArchitecture=LeNet2020.11 | 81.7 | 85.3 | 36.8 | — | |
| BaseArchitecture=LeNet2020.11 | 77.3 | 79.2 | 43.5 | — | |
| LABackbone=LeNet, Hessian Approximation=low-rank, Parameter samples=1002026.05 | 74.5 | — | — | — | |
| RNFs-MLvariant=exact2021.06 | — | — | — | 92 | |
| RNFs-MLK=12021.06 | — | — | — | 97 | |
| RNFs-MLK=42021.06 | — | — | — | 77 | |
| RNFs-MLvariant=exact2021.06 | — | — | — | 91 | |
| RNFs-MLK=12021.06 | — | — | — | 78 | |
| RNFs-MLK=42021.06 | — | — | — | 89 | |
| RNFs-TS2021.06 | — | — | — | 98 | |
| RNFs-TS2021.06 | — | — | — | 96 |