Image Classification on ImageNet-1K (Acc, Brier, NLL, ECE, aECE)
82.19AccuracyDEEP ENSEMBLES
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DEEP ENSEMBLESHeads=122025.10 | 82.19 | 0.25 | 0.65 | 0.01 | 0.01 | |
| HYDRA ENS (TAYLOR)Heads=8x32025.10 | 81.2 | 0.26 | 0.75 | 0.01 | 0.01 | |
| HYDRA ENS (CIRC)Heads=8x32025.10 | 80.88 | 0.27 | 0.74 | 0.01 | 0.01 | |
| LORA ENSEMBLESHeads=122025.10 | 80.68 | 0.27 | 0.71 | 0.01 | 0.01 | |
| TAYLORHeads=82025.10 | 80.68 | 0.28 | 0.79 | 0.02 | 0.02 | |
| SINGLEHeads=122025.10 | 80.67 | 0.27 | 0.71 | 0.01 | 0.01 | |
| MIMOHeads=122025.10 | 80.59 | 0.27 | 0.72 | 0.01 | 0.01 | |
| BATCH ENSEMBLESHeads=122025.10 | 80.53 | 0.27 | 0.72 | 0.01 | 0.01 | |
| MC DROPOUTHeads=122025.10 | 80.3 | 0.28 | 0.73 | 0.02 | 0.02 | |
| CIRCAVGHeads=82025.10 | 80.22 | 0.28 | 0.77 | 0.02 | 0.02 | |
| PACKED ENSEMBLESHeads=122025.10 | 79.23 | 0.29 | 0.78 | 0.01 | 0.01 | |
| OBA2025.10 | 78.52 | 0.3 | 0.85 | 0.03 | 0.03 |