Image Classification on CIFAR-100 40% label flips (test)
0.1928AccuracyREFINE
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| REFINEBackbone=CNNs2025.05 | 0.1928 | 0.8555 | 0.1805 | 0.4 | |
| DistillBackbone=CNNs2025.05 | 0.1873 | 0.8605 | 0.1631 | 0 | |
| NoTransBackbone=CNNs2025.05 | 0.1782 | 0.8259 | 0.1684 | 0.6 | |
| LinearProbeBackbone=CNNs2025.05 | 0.1735 | 0.8605 | 0.1472 | 0 | |
| LoRABackbone=CNNs2025.05 | 0.1724 | 0.8568 | 0.1463 | 0 | |
| AdapterBackbone=CNNs2025.05 | 0.1619 | 0.8578 | 0.1303 | 0 | |
| DANN-GateBackbone=CNNs2025.05 | 0.1502 | 0.8472 | 0.1239 | 0 |