OOD Detection on CIFAR-10 (ID) vs LSUN-R (OOD)
0.52FPR95DynProto
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DynProtoBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 0.52 | 99.81 | |
| AdaNDBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 0.6 | 99.72 | |
| DynProtoBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 1.33 | 99.76 | |
| VIMBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 3.18 | 99.29 | |
| CADRefBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 3.64 | 99.23 | |
| DICEBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 4.22 | 99.06 | |
| ASH-SBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 4.85 | 98.91 | |
| CIDERContrastive Learning=Yes2026.01 | 5.45 | 99.01 | |
| NegLabelBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 6.77 | 98.01 | |
| CSPBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 7.25 | 98.21 | |
| KNN+Contrastive Learning=Yes2026.01 | 7.89 | 98.01 | |
| MCMBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 8.47 | 97.84 | |
| EnergyBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 8.77 | 98.19 | |
| OptFSBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 9.4 | 98.09 | |
| EnergyContrastive Learning=No2026.01 | 10.19 | 98.05 | |
| SSD+Contrastive Learning=Yes2026.01 | 10.56 | 97.83 | |
| CSIContrastive Learning=Yes2026.01 | 10.63 | 97.93 | |
| ODINContrastive Learning=No2026.01 | 10.93 | 97.93 | |
| GODINContrastive Learning=No2026.01 | 11.52 | 97.12 | |
| GLMCMBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 14.51 | 96.78 | |
| Comp.VAEContrastive Learning=No2026.01 | 15.6 | 93.9 | |
| ProxyAnchorContrastive Learning=Yes2026.01 | 21.04 | 97.02 | |
| NegRefineBackbone=CLIP-B/16, Method Category=Vision Language Model based Methods2026.04 | 23.45 | 95.91 | |
| KNNContrastive Learning=No2026.01 | 24.68 | 95.52 | |
| FedTHEalpha=0.12024.10 | 34.94 | 88.69 | |
| FedSDWCalpha=0.12025.11 | 36.69 | 87.7 | |
| FedRoD+FOOGDalpha=0.12024.10 | 41.46 | 93.39 | |
| MSPBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 42.07 | 94.55 | |
| FedTHEalpha=0.12025.11 | 42.95 | 83.46 | |
| FedRoDalpha=0.12024.10 | 43.72 | 83.5 | |
| FedAvg+FOOGDalpha=0.12024.10 | 44.16 | 91.72 | |
| MSPContrastive Learning=No2026.01 | 45.21 | 93.8 | |
| Comp.VAE (vMF)Contrastive Learning=No2026.01 | 46.8 | 81.5 | |
| FOSTERalpha=0.12025.11 | 48.09 | 76.24 | |
| FedICONalpha=0.12024.10 | 49.05 | 83.3 | |
| FedICONalpha=0.12025.11 | 49.05 | 83.3 | |
| FOSTERalpha=0.12024.10 | 54.23 | 82.92 | |
| PerAdaalpha=0.12025.11 | 55.77 | 73.64 | |
| FedIIRalpha=0.12025.11 | 58.44 | 78.69 | |
| FedAvgalpha=0.12024.10 | 61.31 | 80.53 | |
| FedLNalpha=0.12025.11 | 61.31 | 78.34 | |
| FedAvgalpha=0.12025.11 | 62.01 | 77.02 | |
| FedATOLalpha=0.12025.11 | 64.01 | 79.89 | |
| CE + SimCLRContrastive Learning=Yes2026.01 | 64.53 | 85.6 | |
| FedLNalpha=0.12024.10 | 66.33 | 78.69 | |
| MahalanobisContrastive Learning=No2026.01 | 67.73 | 73.61 | |
| ReActBackbone=ResNet50, Method Category=Vision Model based Methods2026.04 | 71.12 | 84.87 | |
| FedIIRalpha=0.12024.10 | 78.77 | 43.4 | |
| FedATOLalpha=0.12024.10 | 87.63 | 62.32 | |
| KNN*Contrastive Learning=No2026.01 | 90 | 59.9 |