Image Classification on CIFAR-100 Instance-dependent Noise (1k clean samples)
67.42Accuracy (eta=0.2)Robust Self-Training with Closed-loop Label Correction
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Robust Self-Training with Closed-loop Label CorrectionModel Selection=Best2026.03 | 67.42 | 56.08 | 48.09 | |
| FasTENModel Selection=Best2026.03 | 66.98 | 57.78 | 34.6 | |
| Robust Self-Training with Closed-loop Label CorrectionModel Selection=Last2026.03 | 66.25 | 55.43 | 46.51 | |
| MW-NetModel Selection=Best2026.03 | 65.65 | 49.39 | 27.5 | |
| L2BModel Selection=Best2026.03 | 64.86 | 58.79 | 45.4 | |
| L2BModel Selection=Last2026.03 | 64.64 | 58.79 | 45.35 | |
| Cross-EntropyModel Selection=Best2026.03 | 63.09 | 47.32 | 27.3 | |
| MW-NetModel Selection=Last2026.03 | 61.38 | 42.06 | 23.38 | |
| FasTENModel Selection=Last2026.03 | 60.85 | 43.07 | 13.03 | |
| Cross-EntropyModel Selection=Last2026.03 | 59.87 | 41.41 | 22.59 |