Image Classification on CIFAR-100-N
80.89AccuracyCoFT+
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CoFT+2026.02 | 80.89 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CoFT2026.02 | 79.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InfoBatch + AlignPrunePrune Ratio=~30%, Backbone=ResNet-182026.04 | 79.3 | — | — | — | — | — | 59.4 | 71.8 | 66 | 41.8 | 72.6 | 68 | 2.7 | — | |
| SeTa + AlignPrunePrune Ratio=~30%, Backbone=ResNet-182026.04 | 79.3 | — | — | — | — | — | 56.3 | 70.8 | 60.5 | 41.6 | 71.9 | 64.3 | 0.7 | — | |
| DEFT2026.02 | 79.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InfoBatchPrune Ratio=~30%, Backbone=ResNet-182026.04 | 79 | — | — | — | — | — | 56.1 | 71.4 | 59.7 | 41.8 | 71.9 | 64.2 | 0.6 | — | |
| SeTaPrune Ratio=~30%, Backbone=ResNet-182026.04 | 79 | — | — | — | — | — | 55.6 | 70.2 | 59 | 41.6 | 71.4 | 63.2 | 0 | — | |
| InfoBatch + AlignPrunePrune Ratio=~50%, Backbone=ResNet-182026.04 | 78.5 | — | — | — | — | — | 60.7 | 71.6 | 62 | 42.6 | 72.6 | 68.6 | 2.4 | — | |
| SeTa + AlignPrunePrune Ratio=~50%, Backbone=ResNet-182026.04 | 78.4 | — | — | — | — | — | 56.3 | 71.2 | 61 | 41.9 | 72.2 | 66 | 1 | — | |
| Full-trainingBackbone=ResNet-182026.04 | 78.2 | — | — | — | — | — | 56.1 | 71.4 | 58.6 | 39.8 | 72.4 | 63.3 | — | — | |
| SeTa + AlignPrunePrune Ratio=~70%, Backbone=ResNet-182026.04 | 77.8 | — | — | — | — | — | 55.5 | 72.3 | 63.3 | 42.7 | 72.7 | 67.6 | 1.7 | — | |
| InfoBatchPrune Ratio=~50%, Backbone=ResNet-182026.04 | 77.7 | — | — | — | — | — | 56 | 71.3 | 60.5 | 42.2 | 71.8 | 65.2 | 0.7 | — | |
| UNICON2026.02 | 77.68 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeTaPrune Ratio=~50%, Backbone=ResNet-182026.04 | 77.5 | — | — | — | — | — | 55.7 | 70.7 | 60 | 40.5 | 71.9 | 64.5 | 0.2 | — | |
| InfoBatch + AlignPrunePrune Ratio=~70%, Backbone=ResNet-182026.04 | 77.5 | — | — | — | — | — | 58.3 | 72.2 | 64.7 | 42.8 | 72.5 | 69.1 | 2.5 | — | |
| Dynamic RandomPrune Ratio=~30%, Backbone=ResNet-182026.04 | 77.3 | — | — | — | — | — | 54.7 | 69.9 | 58.8 | 40.1 | 71.5 | 63.2 | -0.6 | — | |
| SeTaPrune Ratio=~70%, Backbone=ResNet-182026.04 | 77.2 | — | — | — | — | — | 55.2 | 71.6 | 62.4 | 42.4 | 72 | 67.4 | 1.2 | — | |
| InfoBatchPrune Ratio=~70%, Backbone=ResNet-182026.04 | 77.1 | — | — | — | — | — | 55 | 71.4 | 63.6 | 42.4 | 72.2 | 67.8 | 1.4 | — | |
| GMM2026.02 | 76.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProMix2026.02 | 75.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RoLT2026.02 | 75.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| fSGLDModel=ViT-B/16, Evaluation Protocol=Fine-tuning2025.10 | 75.67 | — | — | — | — | — | — | — | — | — | — | — | — | 345.8 | |
| Dynamic RandomPrune Ratio=~50%, Backbone=ResNet-182026.04 | 75.3 | — | — | — | — | — | 54.1 | 70.4 | 59.5 | 40.7 | 70.1 | 64.8 | -0.7 | — | |
| Dynamic RandomPrune Ratio=~70%, Backbone=ResNet-182026.04 | 75.2 | — | — | — | — | — | 53.3 | 70.2 | 62.7 | 41 | 71.9 | 67.3 | 0.3 | — | |
| ASAMModel=ViT-B/16, Evaluation Protocol=Fine-tuning2025.10 | 74.86 | — | — | — | — | — | — | — | — | — | — | — | — | 662.5 | |
| SAMModel=ViT-B/16, Evaluation Protocol=Fine-tuning2025.10 | 74.66 | — | — | — | — | — | — | — | — | — | — | — | — | 656.7 | |
| LongReMix2026.02 | 73.94 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Static RandomPrune Ratio=~30%, Backbone=ResNet-182026.04 | 73.8 | — | — | — | — | — | 53.3 | 67.9 | 51.1 | 30.3 | 68.4 | 57.8 | -5.3 | — | |
| ELR2026.02 | 72.83 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SCE2026.02 | 72.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CE2026.02 | 72.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdamWModel=ViT-B/16, Evaluation Protocol=Fine-tuning2025.10 | 72.3 | — | — | — | — | — | — | — | — | — | — | — | — | 344.5 | |
| Static RandomPrune Ratio=~50%, Backbone=ResNet-182026.04 | 72.1 | — | — | — | — | — | 51.8 | 64 | 47.2 | 22.9 | 65.3 | 54.6 | -8.8 | — | |
| SGDModel=ViT-B/16, Evaluation Protocol=Fine-tuning2025.10 | 71.8 | — | — | — | — | — | — | — | — | — | — | — | — | 343.2 | |
| Static RandomPrune Ratio=~70%, Backbone=ResNet-182026.04 | 69.7 | — | — | — | — | — | 45.9 | 56.7 | 39 | 15.8 | 58.4 | 49.5 | -14.9 | — | |
| IP EnsembleBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 62.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Self-TracInBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 61.99 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Self-LiSSABackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 61.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| fSGLDModel=ResNet-34, Wall-clock time per iteration (s/epoch)=23.72025.10 | 61.51 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| fSGLDModel=ResNet-50, Wall-clock time per iteration (s/epoch)=34.12025.10 | 61.26 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ASAMModel=ResNet-34, Wall-clock time per iteration (s/epoch)=41.42025.10 | 60.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| IPBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 60.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ASAMModel=ResNet-50, Wall-clock time per iteration (s/epoch)=60.92025.10 | 60.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EKFACBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 59.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| α-MAE2026.05 | 59.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LiSSABackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 59.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAMModel=ResNet-34, Wall-clock time per iteration (s/epoch)=41.32025.10 | 59.18 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAMModel=ResNet-50, Wall-clock time per iteration (s/epoch)=60.72025.10 | 59.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TDABackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 58.91 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GEXBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 58.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SGDModel=ResNet-34, Wall-clock time per iteration (s/epoch)=22.02025.10 | 58.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DataInfBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 58.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SGDModel=ResNet-50, Wall-clock time per iteration (s/epoch)=31.92025.10 | 57.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdamWModel=ResNet-50, Wall-clock time per iteration (s/epoch)=32.32025.10 | 57.28 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AdamWModel=ResNet-34, Wall-clock time per iteration (s/epoch)=22.52025.10 | 56.77 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TracInBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 56.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cross EntropyBackbone=ResNet-34, Pre-trained=ImageNet-1K, Removal Ratio=5%2024.05 | 56.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ANL-CE2026.05 | 56.37 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SGCE2026.05 | 56.31 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NCE+AGCE2026.05 | 55.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Unhinged2026.05 | 54.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NCE+RCE2026.05 | 54.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| (ELR+)+VIBCategory=Denoise + IB2025.12 | — | — | — | — | — | 61.06 | — | — | — | — | — | — | — | — | |
| CMW-NetBackbone=PARes182022.11 | — | 0.7011 | 0.6584 | 0.5693 | 0.6429 | — | — | — | — | — | — | — | — | — | |
| CoteachingBackbone=PARes182022.11 | — | 0.7081 | 0.6265 | 0.4155 | 0.5834 | — | — | — | — | — | — | — | — | — | |
| Cross EntropyBackbone=PARes182022.11 | — | 0.6038 | 0.4692 | 0.3182 | 0.4637 | — | — | — | — | — | — | — | — | — | |
| DivideMixBackbone=PARes182022.11 | — | 0.772 | 0.7337 | 0.7075 | 0.7377 | — | — | — | — | — | — | — | — | — | |
| DT-JSCCCategory=Improved IB2025.12 | — | — | — | — | — | 43.61 | — | — | — | — | — | — | — | — | |
| Dynamic LossBackbone=PARes182022.11 | — | 0.7826 | 0.7528 | 0.6918 | 0.7424 | — | — | — | — | — | — | — | — | — | |
| GJSBackbone=PARes182022.11 | — | 0.7331 | 0.7133 | 0.6692 | 0.7052 | — | — | — | — | — | — | — | — | — | |
| JoCoR+VIBCategory=Denoise + IB2025.12 | — | — | — | — | — | 54.24 | — | — | — | — | — | — | — | — | |
| LaT-IBCategory=Ours2025.12 | — | — | — | — | — | 63.59 | — | — | — | — | — | — | — | — | |
| MOIT+Backbone=PARes182022.11 | — | 0.7589 | 0.7088 | 0.653 | 0.7069 | — | — | — | — | — | — | — | — | — | |
| NCRBackbone=PARes182022.11 | — | 0.766 | 0.742 | 0.3825 | 0.6302 | — | — | — | — | — | — | — | — | — | |
| NCTBackbone=PARes182022.11 | — | 0.6765 | 0.5797 | 0.4501 | 0.5688 | — | — | — | — | — | — | — | — | — | |
| NIBCategory=Classic IB2025.12 | — | — | — | — | — | 48.11 | — | — | — | — | — | — | — | — | |
| PLCBackbone=PARes182022.11 | — | 0.5966 | 0.4924 | 0.3318 | 0.4736 | — | — | — | — | — | — | — | — | — | |
| Promix+VIBCategory=Denoise + IB2025.12 | — | — | — | — | — | 63.91 | — | — | — | — | — | — | — | — | |
| SELFIEBackbone=PARes182022.11 | — | 0.5571 | 0.5114 | 0.4385 | 0.5023 | — | — | — | — | — | — | — | — | — | |
| SIBCategory=Improved IB2025.12 | — | — | — | — | — | 50.82 | — | — | — | — | — | — | — | — | |
| VIBCategory=Classic IB2025.12 | — | — | — | — | — | 53.29 | — | — | — | — | — | — | — | — | |
| VIB (LSCE)Category=Robust Loss2025.12 | — | — | — | — | — | 50.71 | — | — | — | — | — | — | — | — |