Model Calibration on CIFAR10 (test)
0.74ECEBSCE-GRA
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BSCE-GRABackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 0.74 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=ResNet50, Temperature Scaling=Post T2025.03 | 0.74 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=DenseNet, Temperature Scaling=Post T2025.03 | 0.77 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 0.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=ResNet110, Temperature Scaling=Post T2025.03 | 0.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 0.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 0.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=DenseNet, Temperature Scaling=Post T2025.03 | 0.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 0.88 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=ResNet50, Temperature Scaling=Post T2025.03 | 0.88 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=WideResNet, Temperature Scaling=Post T2025.03 | 0.94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=WideResNet, Temperature Scaling=Post T2025.03 | 0.95 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 0.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=ResNet110, Temperature Scaling=Post T2025.03 | 0.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=ResNet50, Temperature Scaling=Post T2025.03 | 1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 1.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=ResNet110, Temperature Scaling=Post T2025.03 | 1.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 1.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=DenseNet, Temperature Scaling=Post T2025.03 | 1.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=WideResNet, Temperature Scaling=Post T2025.03 | 1.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=WideResNet, Temperature Scaling=Post T2025.03 | 1.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=WideResNet, Temperature Scaling=Post T2025.03 | 1.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=ResNet50, Temperature Scaling=Post T2025.03 | 1.15 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=WideResNet, Temperature Scaling=Post T2025.03 | 1.16 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=ResNet110, Temperature Scaling=Post T2025.03 | 1.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=DenseNet, Temperature Scaling=Post T2025.03 | 1.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 1.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=WideResNet, Temperature Scaling=Post T2025.03 | 1.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=ResNet110, Temperature Scaling=Post T2025.03 | 1.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=ResNet50, Temperature Scaling=Post T2025.03 | 1.32 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 1.37 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=ResNet50, Temperature Scaling=Post T2025.03 | 1.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRABackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 1.46 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=DenseNet, Temperature Scaling=Post T2025.03 | 1.46 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=DenseNet, Temperature Scaling=Post T2025.03 | 1.46 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=ResNet50, Temperature Scaling=Post T2025.03 | 1.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=ResNet110, Temperature Scaling=Post T2025.03 | 1.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=ResNet110, Temperature Scaling=Post T2025.03 | 1.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=DenseNet, Temperature Scaling=Post T2025.03 | 1.67 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEBackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 1.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 1.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLSDBackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 1.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaCSBackbone=ResNet-502026.03 | 2.48 | — | 0.329 | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 2.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MixupBackbone=ResNet-502026.03 | 3.26 | — | 0.379 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Focal LossBackbone=ResNet-502026.03 | 3.31 | — | 0.408 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLBackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 3.32 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 3.35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=WideResNet, Temperature Scaling=Pre T2025.03 | 3.65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 3.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-EDLevaluation_protocol=single run2026.05 | 4.08 | — | — | — | 0.1348 | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 4.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 4.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DAEDLevaluation_protocol=single run2026.05 | 4.43 | — | — | — | 0.1622 | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=ResNet50, Temperature Scaling=Pre T2025.03 | 4.48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLBackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 4.48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 4.64 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GEDLevaluation_protocol=mean ± std over 5 runs2026.05 | 4.64 | — | — | — | 0.1525 | — | — | — | — | — | — | — | — | — | — | |
| CEBackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 4.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=ResNet110, Temperature Scaling=Pre T2025.03 | 4.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MMCEBackbone=DenseNet, Temperature Scaling=Pre T2025.03 | 4.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| REDevaluation_protocol=single run2026.05 | 4.97 | — | — | — | 0.172 | — | — | — | — | — | — | — | — | — | — | |
| R-EDLevaluation_protocol=single run2026.05 | 5.34 | — | — | — | 0.182 | — | — | — | — | — | — | — | — | — | — | |
| Label SmoothingBackbone=ResNet-502026.03 | 5.44 | — | 0.463 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Baseline (CE)Backbone=ResNet-502026.03 | 9.1 | — | 0.779 | — | — | — | — | — | — | — | — | — | — | — | — | |
| EDLevaluation_protocol=single run2026.05 | 12.94 | — | — | — | 0.1965 | — | — | — | — | — | — | — | — | — | — | |
| I-EDLevaluation_protocol=single run2026.05 | 38.58 | — | — | — | 0.3145 | — | — | — | — | — | — | — | — | — | — | |
| BLModel=ResNet50, Temperature Scaling=Pre T2025.03 | — | 4.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BLModel=ResNet50, Temperature Scaling=Post T2025.03 | — | 1.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEModel=ResNet50, Temperature Scaling=Pre T2025.03 | — | 0.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCEModel=ResNet50, Temperature Scaling=Post T2025.03 | — | 0.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=ResNet50, Temperature Scaling=Pre T2025.03 | — | 0.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=ResNet50, Temperature Scaling=Post T2025.03 | — | 0.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=ResNet110, Temperature Scaling=Pre T2025.03 | — | 1.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=ResNet110, Temperature Scaling=Post T2025.03 | — | 1.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=WideResNet, Temperature Scaling=Pre T2025.03 | — | 1.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=WideResNet, Temperature Scaling=Post T2025.03 | — | 1.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=DenseNet, Temperature Scaling=Pre T2025.03 | — | 1.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BSCE-GRAModel=DenseNet, Temperature Scaling=Post T2025.03 | — | 1.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEModel=ResNet50, Temperature Scaling=Pre T2025.03 | — | 4.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEModel=ResNet50, Temperature Scaling=Post T2025.03 | — | 2.09 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEModel=ResNet110, Temperature Scaling=Pre T2025.03 | — | 4.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEModel=ResNet110, Temperature Scaling=Post T2025.03 | — | 2.46 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CEDL+2026.05 | — | — | — | — | 0.1502 | — | — | — | — | — | — | — | 98.82 | — | — | |
| DAEDL2026.05 | — | — | — | — | 0.1427 | — | — | — | — | — | — | — | 99.08 | — | — | |
| DAEDLCalibration Protocol=pre-calibration2026.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | 7.22 | 14.27 | |
| DFLModel=ResNet50, Temperature Scaling=Pre T2025.03 | — | 1.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DFLModel=ResNet50, Temperature Scaling=Post T2025.03 | — | 1.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DirichletBackbone=densenet402025.11 | — | — | — | — | — | — | — | — | — | — | — | 0.206 | — | — | — | |
| DirichletBackbone=lenet52025.11 | — | — | — | — | — | — | — | — | — | — | — | 0.076 | — | — | — | |
| DirichletBackbone=resnet1102025.11 | — | — | — | — | — | — | — | — | — | — | — | 0.149 | — | — | — | |
| DirichletBackbone=resnet110-SD2025.11 | — | — | — | — | — | — | — | — | — | — | — | 0.121 | — | — | — | |
| DirichletBackbone=resnet-wide322025.11 | — | — | — | — | — | — | — | — | — | — | — | 0.192 | — | — | — | |
| Dropout2026.05 | — | — | — | — | 0.2715 | — | — | — | — | — | — | — | 97.15 | — | — | |
| EDL2026.05 | — | — | — | — | 0.2338 | — | — | — | — | — | — | — | 97.86 | — | — | |
| EDLCalibration Protocol=pre-calibration2026.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | 11.56 | 27.34 | |
| Empirical Bayes (EB)Calibration set=X*h,EB, n=20002026.02 | — | — | — | — | — | — | — | 0.337 | 2.5 | 93.2 | 0.8 | — | — | — | — | |
| FIRType=Average Ranking2025.12 | — | — | 4 | 5.8 | 3.2 | 5.5 | 4.5 | — | — | — | — | — | — | — | — | |
| FLSDModel=ResNet50, Temperature Scaling=Pre T2025.03 | — | 1.56 | — | — | — | — | — | — | — | — | — | — | — | — | — |