Image Classification on CIFAR-100 LT (IF=50)
63.9Top-1 AccGLMC + Loss + LR
Evaluation results
| Method | Links | |
|---|---|---|
| GLMC + Loss + LRreweighting=Inverse-view, LR Schedule=MiLeLR2026.05 | 63.9 | |
| GLMC + Lossreweighting=Inverse-view2026.05 | 63.6 | |
| GLMCVenue=CVPR232026.05 | 61.1 | |
| GLMC + SEL2026.05 | 58.9 | |
| OursBackbone=ResNet-322026.03 | 57.6 | |
| SELBackbone=ResNet-322026.03 | 57.3 | |
| GBGBackbone=ResNet-322026.03 | 57.2 | |
| RBLVenue=ICML232026.05 | 57.2 | |
| ProCoBackbone=ResNet-32, Training Epochs=2002024.03 | 57.1 | |
| FeatReconBackbone=ResNet-322026.03 | 57 | |
| FeatReconVenue=ICLR252026.05 | 57 | |
| DiffuLT + BBNBackbone=ResNet-322026.03 | 56.7 | |
| BCLBackbone=ResNet-322026.03 | 56.6 | |
| BCLBackbone=ResNet-32, Training Epochs=2002024.03 | 56.4 | |
| DiffuLTBackbone=ResNet-322026.03 | 56.3 | |
| PaCoStrategy=Decoupled / Two-stage2026.01 | 56 | |
| Logit AdjustmentBackbone=ResNet-32, Training Epochs=2002024.03 | 54.9 | |
| LLM-AutoDA†Backbone=ResNet-322026.03 | 54.8 | |
| H2TBackbone=ResNet-322026.03 | 53.8 | |
| ETF-DR + Loss + LRreweighting=Inverse-view, LR Schedule=MiLeLR2026.05 | 52.9 | |
| CE* + Loss + LRreweighting=Inverse-view, LR Schedule=MiLeLR2026.05 | 52.8 | |
| CCAR + LAStrategy=Logit-Adjustment, Adjustment=Logit Adjustment2026.01 | 52.78 | |
| ResLTBackbone=ResNet-32, Training Epochs=2002024.03 | 52.7 | |
| CMO + Loss + LRreweighting=Inverse-view, LR Schedule=MiLeLR2026.05 | 52.7 | |
| CE* + Lossreweighting=Inverse-view2026.05 | 52.6 | |
| MetaSAug-LDAMBackbone=ResNet-32, Training Epochs=2002024.03 | 52.3 | |
| Logit AdjustStrategy=Logit-Adjustment2026.01 | 52.3 | |
| MiSLASVenue=CVPR212026.05 | 52.3 | |
| CMO + Lossreweighting=Inverse-view2026.05 | 52.3 | |
| Focal + LAStrategy=Logit-Adjustment, Base Loss=Focal Loss, Adjustment=Logit Adjustment2026.01 | 52.01 | |
| CCAR + BSStrategy=Logit-Adjustment, Adjustment=Balanced Softmax2026.01 | 51.94 | |
| Hybrid-SCBackbone=ResNet-32, Training Epochs=2002024.03 | 51.9 | |
| HCLVenue=CVPR212026.05 | 51.9 | |
| INC-DRWVenue=AISTATS232026.05 | 51.7 | |
| RIDE-3 expertsVenue=ICLR212026.05 | 51.4 | |
| ETF-DR + Lossreweighting=Inverse-view2026.05 | 51.4 | |
| DiVEBackbone=ResNet-322026.03 | 51.1 | |
| CUDABackbone=ResNet-322026.03 | 51.1 | |
| ETF-DR + DisA2026.05 | 51 | |
| cRT + mixupStrategy=Decoupled / Two-stage2026.01 | 50.86 | |
| FedYoYoVenue=ICCV252026.05 | 50.8 | |
| Balanced SoftmaxStrategy=Logit-Adjustment2026.01 | 50.5 | |
| ETF-DRVenue=NeurIPS222026.05 | 50.4 | |
| Casual ModelBackbone=ResNet-32, Training Epochs=2002024.03 | 50.3 | |
| Focal + BSStrategy=Logit-Adjustment, Base Loss=Focal Loss, Adjustment=Balanced Softmax2026.01 | 49.98 | |
| DisAVenue=ICML242026.05 | 49.6 | |
| SELVenue=ICCV252026.05 | 49.6 | |
| CMOVenue=CVPR222026.05 | 48.3 | |
| Focal-SAMVenue=ICML252026.05 | 48.1 | |
| LDAM-DRWStrategy=Margin-based2026.01 | 47.62 | |
| TSCBackbone=ResNet-32, Training Epochs=2002024.03 | 47.4 | |
| TSCVenue=CVPR222026.05 | 47.4 | |
| SSPBackbone=ResNet-32, Training Epochs=2002024.03 | 47.1 | |
| BBNVenue=CVPR202026.05 | 47.1 | |
| BBNBackbone=ResNet-32, Training Epochs=2002024.03 | 47 | |
| BBNBackbone=ResNet-322026.03 | 47 | |
| cRTBackbone=ResNet-322026.03 | 46.8 | |
| LDAM-DRWBackbone=ResNet-32, Training Epochs=2002024.03 | 46.6 | |
| LDAM-DRWBackbone=ResNet-322026.03 | 46.6 | |
| KCLVenue=ICLR212026.05 | 46.3 | |
| CAM-BSBackbone=ResNet-322026.03 | 46 | |
| CE*reproduced=true2026.05 | 45.5 | |
| CCAR + CEStrategy=Standard / Re-weighting, Base Loss=Cross-Entropy2026.01 | 45.25 | |
| CB-FocalBackbone=ResNet-32, Training Epochs=2002024.03 | 45.2 | |
| Class-BalancedStrategy=Standard / Re-weighting2026.01 | 45.17 | |
| Focal LossStrategy=Standard / Re-weighting2026.01 | 44.3 | |
| Focal LossBackbone=ResNet-322026.03 | 44.3 | |
| Cross-EntropyStrategy=Standard / Re-weighting2026.01 | 43.9 | |
| CEBackbone=ResNet-322026.03 | 43.9 | |
| MLLM-LLaVA-FLType=Our framework2024.09 | 42.34 | |
| CLIP2FLType=SOTA2024.09 | 41.29 | |
| CReFFType=Classifier-retraining2024.09 | 37.64 | |
| CCVRType=Heterogeneity-oriented FL methods2024.09 | 36.98 | |
| Ratio LossType=Imbalance-oriented FL methods2024.09 | 36.88 | |
| FedNovaType=Heterogeneity-oriented FL methods2024.09 | 36.62 | |
| FedBEType=Heterogeneity-oriented FL methods2024.09 | 36.39 | |
| FedAvgType=Heterogeneity-oriented FL methods2024.09 | 36.35 | |
| FedProxType=Heterogeneity-oriented FL methods2024.09 | 36.3 | |
| FedDFType=Heterogeneity-oriented FL methods2024.09 | 36.22 | |
| FedAvgMType=Heterogeneity-oriented FL methods2024.09 | 35.33 | |
| Fed-Focal LossType=Imbalance-oriented FL methods2024.09 | 35.25 | |
| FedAvg+ τ-normType=Imbalance-oriented FL methods2024.09 | 33.71 |