Image Classification on Food-101 (test)
93.1AccuracyFinetune
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Finetunelabeled_data_ratio=100%, backbone=ViT-Base2022.08 | 93.1 | — | — | — | — | — | |
| Semi-ViTlabeled_data_ratio=10%, backbone=ViT-Base2022.08 | 91.3 | — | — | — | — | — | |
| 2SFSBackbone=ViT-L/14, Shots=162025.03 | 91.1 | — | — | — | — | — | |
| VLPCookBackbone=ViT, Linear probing=true, Frozen backbone=true2022.12 | 89.14 | — | — | — | — | — | |
| ReLICv2Evaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.7 | — | — | — | — | — | |
| BYOLEvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.5 | — | — | — | — | — | |
| Supervised-INEvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.3 | — | — | — | — | — | |
| SimCLREvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 88.2 | — | — | — | — | — | |
| GAMRBackbone=ResNet-50, Pre-trained=true2026.05 | 87.4 | — | — | — | — | — | |
| Random InitEvaluation protocol=Fine-tuned, Backbone=ResNet-50, Pre-trained=None2022.01 | 86.9 | — | — | — | — | — | |
| SNSCL+Backbone=ResNet-50, Pre-trained=true2026.05 | 86.4 | — | — | — | — | — | |
| PSSCLBackbone=ResNet-50, Pre-trained=true2026.05 | 86.4 | — | — | — | — | — | |
| R-KalmanBackbone=ViT-B16, Evaluation Protocol=Fine-tuning, LoRA rank (r)=4, LoRA alpha (α)=42026.01 | 86.2 | — | — | — | — | — | |
| LongReMixBackbone=ResNet-50, Pre-trained=true2026.05 | 86.2 | — | — | — | — | — | |
| 2SFSBackbone=ViT-B/16, Shots=162025.03 | 86.1 | — | — | — | — | — | |
| DivideMixBackbone=ResNet-50, Pre-trained=true2026.05 | 85.9 | — | — | — | — | — | |
| WarPIBackbone=ResNet-50, Pre-trained=true2026.05 | 85.9 | — | — | — | — | — | |
| RENGBackbone=ViT-B16, Evaluation Protocol=Fine-tuning, LoRA rank (r)=4, LoRA alpha (α)=42026.01 | 85.8 | — | — | — | — | — | |
| NGDBackbone=ViT-B16, Evaluation Protocol=Fine-tuning, LoRA rank (r)=4, LoRA alpha (α)=42026.01 | 85.4 | — | — | — | — | — | |
| SNSCLBackbone=ResNet-50, Pre-trained=true2026.05 | 85.4 | — | — | — | — | — | |
| PLCBackbone=ResNet-50, Pre-trained=true2026.05 | 85.3 | — | — | — | — | — | |
| AdamWBackbone=ViT-B16, Evaluation Protocol=Fine-tuning, LoRA rank (r)=4, LoRA alpha (α)=42026.01 | 85.2 | — | — | — | — | — | |
| NoiseRankBackbone=ResNet-50, Pre-trained=true2026.05 | 85.2 | — | — | — | — | — | |
| SMPBackbone=ResNet-50, Pre-trained=true2026.05 | 85.1 | — | — | — | — | — | |
| RINGBackbone=ViT-B16, Evaluation Protocol=Fine-tuning, LoRA rank (r)=4, LoRA alpha (α)=42026.01 | 85 | — | — | — | — | — | |
| Finetunelabeled_data_ratio=10%, backbone=ViT-Base2022.08 | 84.5 | — | — | — | — | — | |
| H-T (ViT)Backbone=ViT, Linear probing=true, Frozen backbone=true2022.12 | 84.44 | — | — | — | — | — | |
| CleanNetBackbone=ResNet-50, Pre-trained=true2026.05 | 83.5 | — | — | — | — | — | |
| Semi-ViTlabeled_data_ratio=1%, backbone=ViT-Base2022.08 | 82.1 | — | — | — | — | — | |
| Standard CEBackbone=ResNet-50, Pre-trained=true2026.05 | 81.4 | — | — | — | — | — | |
| ImageNet (ViT)Backbone=ViT, Linear probing=true, Frozen backbone=true2022.12 | 80.99 | — | — | — | — | — | |
| 2SFSBackbone=ViT-B/32, Shots=162025.03 | 80.8 | — | — | — | — | — | |
| ReLICv2Evaluation protocol=Linear evaluation, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 80.6 | — | — | — | — | — | |
| Fully SupervisedLabel Usage=With Labels, Pruning Rate=30%2026.05 | 80.2 | — | — | — | — | — | |
| SophiaBackbone=ViT-B16, Evaluation Protocol=Fine-tuning, LoRA rank (r)=4, LoRA alpha (α)=42026.01 | 80 | — | — | — | — | — | |
| DnCBackbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=4500 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.01 | 78.7 | — | — | — | — | — | |
| ReLICv2Backbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=5000 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.01 | 78.3 | — | — | — | — | — | |
| Semi-AUM+CutoffLabel Usage=With 10% Labels, Pruning Rate=30%2026.05 | 77.9 | — | — | — | — | — | |
| Semi-DUAL+BetaLabel Usage=With 10% Labels, Pruning Rate=30%2026.05 | 77.9 | — | — | — | — | — | |
| ReLICv2Backbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=1000 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.01 | 77.5 | — | — | — | — | — | |
| Fully SupervisedLabel Usage=With Labels, Pruning Rate=50%2026.05 | 77.5 | — | — | — | — | — | |
| ELFS (DINO)Label Usage=Without Labels, Pruning Rate=30%2026.05 | 77.2 | — | — | — | — | — | |
| NNCLREvaluation protocol=Linear evaluation, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 76.7 | — | — | — | — | — | |
| ZCoreLabel Usage=Without Labels, Pruning Rate=30%2026.05 | 76.7 | — | — | — | — | — | |
| RandomLabel Usage=Without Labels, Pruning Rate=30%2026.05 | 76.6 | — | — | — | — | — | |
| Score ExtrapolationLabel Usage=With 10% Labels, Pruning Rate=30%2026.05 | 76.3 | — | — | — | — | — | |
| SPRLBackbone=ResNet-18, Pre-trained=true2026.05 | 76.1 | — | — | — | — | — | |
| BYOLEvaluation protocol=Linear evaluation, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 75.3 | — | — | — | — | — | |
| ZENITHBackbone=DLA2026.01 | 75.3 | — | — | — | — | 1.4 | |
| SPSBackbone=DLA2026.01 | 73.9 | — | — | — | — | 1.33 | |
| Semi-AUM+CutoffLabel Usage=With 10% Labels, Pruning Rate=50%2026.05 | 73.9 | — | — | — | — | — | |
| ELFS (Self-Encoder)Label Usage=Without Labels, Pruning Rate=30%2026.05 | 73.7 | — | — | — | — | — | |
| Semi-DUAL+BetaLabel Usage=With 10% Labels, Pruning Rate=50%2026.05 | 73.4 | — | — | — | — | — | |
| BYOLBackbone=ResNet-50, Pre-training dataset=JFT, Pre-training epochs=5000 ImageNet-equivalent, Evaluation protocol=Linear transfer2022.01 | 73.3 | — | — | — | — | — | |
| LQABackbone=DLA2026.01 | 73.3 | — | — | — | — | 2.69 | |
| ELFS (DINO)Label Usage=Without Labels, Pruning Rate=50%2026.05 | 73.3 | — | — | — | — | — | |
| RandomLabel Usage=Without Labels, Pruning Rate=50%2026.05 | 72.9 | — | — | — | — | — | |
| DAdaptBackbone=DLA2026.01 | 72.5 | — | — | — | — | 2.32 | |
| HamBRBackbone=ResNet-50, Pre-trained=true2026.05 | 72.5 | — | — | — | — | — | |
| Supervised-INEvaluation protocol=Linear evaluation, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 72.3 | — | — | — | — | — | |
| ALIGBackbone=DLA2026.01 | 72.3 | — | — | — | — | 0.68 | |
| PrototypicalityLabel Usage=Without Labels, Pruning Rate=30%2026.05 | 72.3 | — | — | — | — | — | |
| ZCoreLabel Usage=Without Labels, Pruning Rate=50%2026.05 | 72.1 | — | — | — | — | — | |
| ProdigyBackbone=DLA2026.01 | 72 | — | — | — | — | 2.04 | |
| Score ExtrapolationLabel Usage=With 10% Labels, Pruning Rate=50%2026.05 | 71.8 | — | — | — | — | — | |
| Fully SupervisedLabel Usage=With Labels, Pruning Rate=70%2026.05 | 71.3 | — | — | — | — | — | |
| PSSCLBackbone=ResNet-50, Pre-trained=true2026.05 | 70.7 | — | — | — | — | — | |
| UNICONBackbone=ResNet-50, Pre-trained=true2026.05 | 70.4 | — | — | — | — | — | |
| ProMixBackbone=ResNet-50, Pre-trained=true2026.05 | 70.1 | — | — | — | — | — | |
| RRLBackbone=ResNet-50, Pre-trained=true2026.05 | 69.7 | — | — | — | — | — | |
| L2BBackbone=ResNet-50, Pre-trained=true2026.05 | 68.9 | — | — | — | — | — | |
| LongReMixBackbone=ResNet-50, Pre-trained=true2026.05 | 68.7 | — | — | — | — | — | |
| SimCLREvaluation protocol=Linear evaluation, Backbone=ResNet-50, Pre-trained=ImageNet2022.01 | 68.4 | — | — | — | — | — | |
| PALBackbone=DLA2026.01 | 68.4 | — | — | — | — | 0.85 | |
| DoWGBackbone=DLA2026.01 | 68.1 | — | — | — | — | 1.39 | |
| DoGBackbone=DLA2026.01 | 67.8 | — | — | — | — | 2.86 | |
| DivideMixBackbone=ResNet-50, Pre-trained=true2026.05 | 67.2 | — | — | — | — | — | |
| Fully SupervisedLabel Usage=With Labels, Pruning Rate=80%2026.05 | 67.2 | — | — | — | — | — | |
| Semi-AUM+CutoffLabel Usage=With 10% Labels, Pruning Rate=70%2026.05 | 67.2 | — | — | — | — | — | |
| PrototypicalityLabel Usage=Without Labels, Pruning Rate=50%2026.05 | 66.9 | — | — | — | — | — | |
| Semi-DUAL+BetaLabel Usage=With 10% Labels, Pruning Rate=70%2026.05 | 66.3 | — | — | — | — | — | |
| ELFS (DINO)Label Usage=Without Labels, Pruning Rate=70%2026.05 | 65.6 | — | — | — | — | — | |
| COCOBBackbone=DLA2026.01 | 64.3 | — | — | — | — | 1.58 | |
| RandomLabel Usage=Without Labels, Pruning Rate=70%2026.05 | 63.9 | — | — | — | — | — | |
| ZCoreLabel Usage=Without Labels, Pruning Rate=70%2026.05 | 63.4 | — | — | — | — | — | |
| Score ExtrapolationLabel Usage=With 10% Labels, Pruning Rate=70%2026.05 | 63.2 | — | — | — | — | — | |
| Semi-AUM+CutoffLabel Usage=With 10% Labels, Pruning Rate=80%2026.05 | 63.1 | — | — | — | — | — | |
| ELFS (Self-Encoder)Label Usage=Without Labels, Pruning Rate=50%2026.05 | 62.8 | — | — | — | — | — | |
| Semi-DUAL+BetaLabel Usage=With 10% Labels, Pruning Rate=80%2026.05 | 62.2 | — | — | — | — | — | |
| ELFS (DINO)Label Usage=Without Labels, Pruning Rate=80%2026.05 | 61.1 | — | — | — | — | — | |
| Finetunelabeled_data_ratio=1%, backbone=ViT-Base2022.08 | 60.9 | — | — | — | — | — | |
| ELFS (Self-Encoder)Label Usage=Without Labels, Pruning Rate=70%2026.05 | 58.6 | — | — | — | — | — | |
| Score ExtrapolationLabel Usage=With 10% Labels, Pruning Rate=80%2026.05 | 57.6 | — | — | — | — | — | |
| PrototypicalityLabel Usage=Without Labels, Pruning Rate=70%2026.05 | 56.6 | — | — | — | — | — | |
| ZCoreLabel Usage=Without Labels, Pruning Rate=80%2026.05 | 56.3 | — | — | — | — | — | |
| RandomLabel Usage=Without Labels, Pruning Rate=80%2026.05 | 55.5 | — | — | — | — | — | |
| Fully SupervisedLabel Usage=With Labels, Pruning Rate=90%2026.05 | 55.3 | — | — | — | — | — | |
| ELFS (Self-Encoder)Label Usage=Without Labels, Pruning Rate=80%2026.05 | 52.9 | — | — | — | — | — | |
| Semi-DUAL+BetaLabel Usage=With 10% Labels, Pruning Rate=90%2026.05 | 51.8 | — | — | — | — | — | |
| Semi-AUM+CutoffLabel Usage=With 10% Labels, Pruning Rate=90%2026.05 | 51.2 | — | — | — | — | — |