Image Classification on ImageNet (val) (Standard Top-1/Top-5 Accuracy)
84.1Top-1 AccuracyQUEST
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| QUESTModel=ViT-L/16†, Epochs=400 + 20, Training Recipe=DeiT-32026.03 | 84.1 | — | — | — | — | |
| StandardModel=ViT-L/16†, Epochs=400 + 20, Training Recipe=DeiT-32026.03 | 83.9 | — | — | — | — | |
| Swin-BF. (G)=154, #P. (M)=87.8, Input Resolution=224 x 2242025.05 | 83.4 | — | — | — | — | |
| QUESTModel=ViT-H/14†, Epochs=400 + 20, Training Recipe=DeiT-32026.03 | 83.4 | — | — | — | — | |
| QUESTModel=ViT-B/16, Epochs=400 + 20, Training Recipe=DeiT-32026.03 | 83.2 | — | — | — | — | |
| StandardModel=ViT-H/14†, Epochs=400 + 20, Training Recipe=DeiT-32026.03 | 83.2 | — | — | — | — | |
| Swin-SF. (G)=87, #P. (M)=49.6, Input Resolution=224 x 2242025.05 | 83.1 | — | — | — | — | |
| StandardModel=ViT-B/16, Epochs=400 + 20, Training Recipe=DeiT-32026.03 | 82.7 | — | — | — | — | |
| CLIP ViT-B/16 + EDBackbone=ViT-B/16, Regularization=Effective Degree (ED)2026.05 | 82.19 | — | — | — | — | |
| PVTv2-b2F. (G)=40, #P. (M)=25.4, Input Resolution=224 x 2242025.05 | 82 | — | — | — | — | |
| CLIP ViT-B/16Backbone=ViT-B/162026.05 | 81.35 | — | — | — | — | |
| Swin-TF. (G)=45, #P. (M)=28.3, Input Resolution=224 x 2242025.05 | 81.2 | — | — | — | — | |
| GRF++2025.10 | 80.31 | — | — | — | — | |
| QUESTModel=ViT-S/16, Epochs=200, Training Recipe=DeiT-12026.03 | 80.2 | — | — | — | — | |
| ViTVariant=standard2025.10 | 80.1 | — | — | — | — | |
| QUESTModel=ViT-B/16, Epochs=100, Training Recipe=DeiT-12026.03 | 79.7 | — | — | — | — | |
| StandardModel=ViT-S/16, Epochs=200, Training Recipe=DeiT-12026.03 | 79.6 | — | — | — | — | |
| LE-SAMBackbone=ResNet-1012026.05 | 79.31 | — | — | — | — | |
| ESAMBackbone=ResNet-1012026.05 | 79.1 | — | — | — | — | |
| QKNorm-DS‡Model=ViT-B/16, Epochs=100, Training Recipe=DeiT-12026.03 | 79 | — | — | — | — | |
| Eigen-SAMBackbone=ResNet-1012026.05 | 78.96 | — | — | — | — | |
| F-SAMBackbone=ResNet-1012026.05 | 78.72 | — | — | — | — | |
| PVTv2-b1F. (G)=21, #P. (M)=13.1, Input Resolution=224 x 2242025.05 | 78.7 | — | — | — | — | |
| SAMBackbone=ResNet-1012026.05 | 78.54 | — | — | — | — | |
| ASAMBackbone=ResNet-1012026.05 | 78.43 | — | — | — | — | |
| R-101-P2HCT-SF. (G)=140, #P. (M)=54.4, Input Resolution=224 x 2242025.05 | 78.4 | 94.1 | — | — | — | |
| VeLONumber of Parameters=86M2025.06 | 78.39 | — | — | — | — | |
| STEAMBackbone=ResNet-101, Params=+0.57K, GFLOPs=7.86832024.12 | 78.38 | 94.33 | — | — | — | |
| RepLBackbone=ResNet-1522026.05 | 78.31 | 94.14 | — | 24.19 | 633.89 | |
| MCABackbone=ResNet-101, Params=+130.4K, GFLOPs=7.86562024.12 | 78.17 | 94.18 | — | — | — | |
| E2EBackbone=ResNet-1522026.05 | 78.16 | 94.03 | — | 27.58 | 738.74 | |
| RepLBackbone=ResNet-1012026.05 | 78.13 | 94.02 | — | 18.05 | 616.23 | |
| GCTBackbone=ResNet-101, Params=+0.03K, GFLOPs=7.88302024.12 | 78.08 | 94.15 | — | — | — | |
| R-50-P2HCT-NF. (G)=72, #P. (M)=29.2, Input Resolution=224 x 2242025.05 | 77.9 | 94 | — | — | — | |
| ViT-BF. (G)=176, #P. (M)=86.6, Input Resolution=224 x 2242025.05 | 77.9 | — | — | — | — | |
| CBAMBackbone=ResNet-101, Params=+4.78M, GFLOPs=7.89422024.12 | 77.9 | 93.87 | — | — | — | |
| ECABackbone=ResNet-101, Params=+0.17K, GFLOPs=7.88322024.12 | 77.85 | 93.88 | — | — | — | |
| SGDBackbone=ResNet-1012026.05 | 77.81 | — | — | — | — | |
| SEBackbone=ResNet-101, Params=+4.74M, GFLOPs=7.88772024.12 | 77.59 | 93.81 | — | — | — | |
| E2EBackbone=ResNet-1012026.05 | 77.55 | 93.8 | — | 20.95 | 720.11 | |
| LE-SAMBackbone=ResNet-502026.05 | 77.41 | — | — | — | — | |
| R-101F. (G)=132, #P. (M)=46.4, Input Resolution=224 x 2242025.05 | 77.4 | — | — | — | — | |
| Adam + CosineNumber of Parameters=86M2025.06 | 77.22 | — | — | — | — | |
| STEAMBackbone=ResNet-50, Params=+0.32K, GFLOPs=4.13602024.12 | 77.2 | 93.64 | — | — | — | |
| CLIP ViT-B/32 + EDBackbone=ViT-B/32, Regularization=Effective Degree (ED)2026.05 | 77.14 | — | — | — | — | |
| EfficientNet-B0F. (G)=3.9, #P. (M)=5.3, Input Resolution=224 x 2242025.05 | 77.1 | 93.3 | — | — | — | |
| MCABackbone=ResNet-50, Params=+6.06K, GFLOPs=4.13272024.12 | 77.05 | 93.5 | — | — | — | |
| GCTBackbone=ResNet-50, Params=+0.02K, GFLOPs=4.14352024.12 | 77.03 | 93.52 | — | — | — | |
| ESAMBackbone=ResNet-502026.05 | 77.02 | — | — | — | — | |
| Eigen-SAMBackbone=ResNet-502026.05 | 76.94 | — | — | — | — | |
| CBAMBackbone=ResNet-50, Params=+2.53M, GFLOPs=4.15002024.12 | 76.94 | 93.46 | — | — | — | |
| F-SAMBackbone=ResNet-502026.05 | 76.93 | — | — | — | — | |
| ECABackbone=ResNet-50, Params=+0.08K, GFLOPs=4.14362024.12 | 76.88 | 93.38 | — | — | — | |
| ResNet-101Backbone=ResNet-101, Params=44.55M, GFLOPs=7.86512024.12 | 76.71 | 93.26 | — | — | — | |
| SAMBackbone=ResNet-502026.05 | 76.7 | — | — | — | — | |
| ASAMBackbone=ResNet-502026.05 | 76.68 | — | — | — | — | |
| SEBackbone=ResNet-50, Params=+2.51M, GFLOPs=4.14602024.12 | 76.61 | 93.18 | — | — | — | |
| YOLOv11-cls-P2HCT-SF. (G)=19, #P. (M)=11.5, Input Resolution=224 x 2242025.05 | 76.2 | 93 | — | — | — | |
| R-50F. (G)=71, #P. (M)=27.4, Input Resolution=224 x 2242025.05 | 76.2 | — | — | — | — | |
| CLIP ViT-B/32Backbone=ViT-B/322026.05 | 76.2 | — | — | — | — | |
| TeacherTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=N/A2024.11 | 76.16 | 92.86 | — | — | — | |
| TeacherTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 76.16 | 92.86 | — | — | — | |
| SGDBackbone=ResNet-502026.05 | 76.02 | — | — | — | — | |
| RepLBackbone=ResNet-342026.05 | 75.44 | 91.47 | — | 8.06 | 410.53 | |
| YOLOv11-cls-SF. (G)=16, #P. (M)=5.5, Input Resolution=224 x 2242025.05 | 75.4 | 92.7 | — | — | — | |
| ResNet-50Backbone=ResNet-50, Params=25.56M, GFLOPs=4.13242024.12 | 75.22 | 92.52 | — | — | — | |
| R-18-P2HCT-NF. (G)=43, #P. (M)=19.6, Input Resolution=224 x 2242025.05 | 75 | 92.5 | — | — | — | |
| QUESTModel=ViT-L/16, Epochs=100, Training Recipe=DeiT-12026.03 | 74.9 | — | — | — | — | |
| E2EBackbone=ResNet-342026.05 | 74.82 | 91.04 | — | 9.21 | 463.23 | |
| YOLOv8-cls-SF. (G)=17, #P. (M)=6.4, Input Resolution=224 x 2242025.05 | 73.8 | 91.7 | — | — | — | |
| TeacherTeacher architecture=ResNet-342024.11 | 73.31 | 91.42 | — | — | — | |
| TeacherTeacher/Student architecture pair=ResNet34/ResNet182026.05 | 73.31 | 91.42 | — | — | — | |
| DISTTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=logits2024.11 | 73.24 | 91.12 | — | — | — | |
| DHKDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=logits2024.11 | 72.99 | 91.45 | — | — | — | |
| KD+OursTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 72.75 | 91.41 | — | — | — | |
| ReviewKDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=features2024.11 | 72.56 | 91 | — | — | — | |
| ReviewKDTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 72.56 | 91 | — | — | — | |
| QKNorm-DS‡Model=ViT-L/16, Epochs=100, Training Recipe=DeiT-12026.03 | 72.5 | — | — | — | — | |
| SimKDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=features2024.11 | 72.25 | 90.86 | — | — | — | |
| CAT-KDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=features2024.11 | 72.24 | 91.13 | — | — | — | |
| KD+LSTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 72.18 | 90.8 | — | — | — | |
| DHKDdistillation manner=logits, Teacher architecture=ResNet-34, Student architecture=ResNet-182024.11 | 72.15 | 90.89 | — | — | — | |
| SeAl-KDKnowledge Distillation=Yes, Backbone=ResNet-342026.05 | 72.11 | — | — | — | — | |
| DISTdistillation manner=logits, Teacher architecture=ResNet-34, Student architecture=ResNet-182024.11 | 72.07 | 90.42 | — | — | — | |
| DKDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=logits2024.11 | 72.05 | 91.05 | — | — | — | |
| DKDTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 72.05 | 91.05 | — | — | — | |
| KD+OursTeacher/Student architecture pair=ResNet34/ResNet182026.05 | 71.97 | 90.73 | — | — | — | |
| DKDdistillation manner=logits, Teacher architecture=ResNet-34, Student architecture=ResNet-182024.11 | 71.7 | 90.41 | — | — | — | |
| DKDTeacher/Student architecture pair=ResNet34/ResNet182026.05 | 71.7 | 90.41 | — | — | — | |
| ReviewKDdistillation manner=features, Teacher architecture=ResNet-34, Student architecture=ResNet-182024.11 | 71.61 | 90.51 | — | — | — | |
| ReviewKDTeacher/Student architecture pair=ResNet34/ResNet182026.05 | 71.61 | 90.51 | — | — | — | |
| SimKDdistillation manner=features, Teacher architecture=ResNet-34, Student architecture=ResNet-182024.11 | 71.59 | 90.48 | — | — | — | |
| KD+LSTeacher/Student architecture pair=ResNet34/ResNet182026.05 | 71.42 | 90.29 | — | — | — | |
| CTKDTeacher/Student architecture pair=ResNet34/ResNet182026.05 | 71.38 | 90.27 | — | — | — | |
| CRDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=features2024.11 | 71.37 | 90.41 | — | — | — | |
| CRDTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 71.37 | 90.41 | — | — | — | |
| STEAMBackbone=ResNet-18, Params=+0.25K, GFLOPs=1.82612024.12 | 71.36 | 90.1 | — | — | — | |
| CAT-KDdistillation manner=features, Teacher architecture=ResNet-34, Student architecture=ResNet-182024.11 | 71.26 | 90.45 | — | — | — | |
| OFDTeacher architecture=ResNet-50, Student architecture=MobileNet, distillation manner=features2024.11 | 71.25 | 90.34 | — | — | — | |
| OFDTeacher/Student architecture pair=ResNet50/MN-V12026.05 | 71.25 | 90.34 | — | — | — |