Image Classification on CIFAR-10 ST-50 step-imbalance (test)
97.56AccuracyGRAIN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GRAINBackbone=ViT-base, Fine-tuning strategy=FFT + GRAIN2026.06 | 97.56 | 0.13 | |
| FocalLossBackbone=ViT-base, Fine-tuning strategy=FFT + FocalLoss2026.06 | 97.15 | 0.57 | |
| PCGradBackbone=ViT-base, Fine-tuning strategy=FFT + PCGrad2026.06 | 96.84 | 0.15 | |
| NoisyTuneBackbone=ViT-base, Fine-tuning strategy=FFT + NoisyTune2026.06 | 96.8 | 0.24 | |
| SAMBackbone=ViT-base, Fine-tuning strategy=FFT + SAM2026.06 | 96.57 | 0.22 | |
| ViT (FFT)Backbone=ViT-base, Fine-tuning strategy=FFT2026.06 | 96.09 | 0.51 | |
| SWABackbone=ViT-base, Fine-tuning strategy=FFT + SWA2026.06 | 71.37 | 16.83 |