Image Classification on CIFAR-10 Low-pass Filtered (test)
88.73Test AccuracyClean
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CleanBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 88.73 | 0 | |
| LSPBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 88.66 | 0.07 | |
| TUEBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 88.65 | 0.08 | |
| GUEBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 88.41 | 0.32 | |
| EMNBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 87.33 | 1.4 | |
| PUEBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 87.27 | 1.46 | |
| FUSEBackbone=ResNet50×4, Evaluation Protocol=Fine-tune (last block + pooling)2026.06 | 82.67 | 6.06 | |
| Zero-shotBackbone=ResNet50×4, Evaluation Protocol=Zero-shot2026.06 | 75.63 | — |