Image Classification on ImageNet-1K (Top-1 and Top-5 Accuracy)
84.6Top-1 AccDaViT-B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DaViT-B#Params=87.9M, FLOPs=15.5G2025.12 | 84.6 | 96.9 | |
| DaViT-S#Params=49.7M, FLOPs=8.8G2025.12 | 84.2 | 96.9 | |
| DaViT-T#Params=28.3M, FLOPs=4.5G2025.12 | 82.8 | 96.2 | |
| DisentangleFormer-Tiny#Params=25.6M, FLOPs=3.7G2025.12 | 80 | 94.8 | |
| SWEETModel Scale=L6H12, Para.(M)=44.0, FLOPs (G)=17.0, Training Epochs=302026.01 | 77.42 | 93.25 | |
| WAVEModel Scale=L6H12, Para.(M)=44.0, FLOPs (G)=17.0, Training Epochs=302026.01 | 76.97 | 93.19 | |
| WT-SelectModel Scale=L6H12, Para.(M)=44.0, FLOPs (G)=17.0, Training Epochs=302026.01 | 75.91 | 92.42 | |
| Iso. PruningModel Scale=L6H12, Para.(M)=44.0, FLOPs (G)=17.0, Training Epochs=302026.01 | 75.85 | 92.54 | |
| DMAEModel Scale=L6H12, Para.(M)=44.0, FLOPs (G)=17.0, Training Epochs=302026.01 | 73.95 | 91.17 | |
| FCFDTeacher=ResNet50, Student=MobileNetV12025.11 | 73.37 | 91.35 | |
| HCDTeacher=ResNet50, Student=MobileNetV12025.11 | 73.23 | 91.46 | |
| LDRLDTeacher=ResNet50, Student=MobileNetV12025.11 | 73.12 | 91.43 | |
| WTTMTeacher=ResNet50, Student=MobileNetV12025.11 | 73.09 | — | |
| RSDTeacher=ResNet50, Student=MobileNetV12025.11 | 73.08 | — | |
| TeKAPTeacher=ResNet50, Student=MobileNetV12025.11 | 72.87 | 91.05 | |
| RLDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.75 | 91.18 | |
| CLIPSetting=Zero-shot2026.02 | 72.7 | — | |
| IPWDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.65 | 91.08 | |
| ReviewKDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.56 | 91 | |
| FCFDTeacher=ResNet34, Student=ResNet182025.11 | 72.24 | 90.74 | |
| CAT-KDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.24 | 91.13 | |
| SDDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.24 | 90.71 | |
| WTTMTeacher=ResNet34, Student=ResNet182025.11 | 72.19 | — | |
| RSDTeacher=ResNet34, Student=ResNet182025.11 | 72.18 | — | |
| HCDTeacher=ResNet34, Student=ResNet182025.11 | 72.18 | 90.64 | |
| LSKDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.18 | 90.8 | |
| DKDTeacher=ResNet50, Student=MobileNetV12025.11 | 72.05 | 91.05 | |
| RLDTeacher=ResNet34, Student=ResNet182025.11 | 71.91 | 90.54 | |
| LDRLDTeacher=ResNet34, Student=ResNet182025.11 | 71.88 | 90.59 | |
| SWEETModel Scale=L12H6, Para.(M)=22.1, FLOPs (G)=8.5, Training Epochs=302026.01 | 71.7 | 90.24 | |
| DKDTeacher=ResNet34, Student=ResNet182025.11 | 71.7 | 90.31 | |
| ReviewKDTeacher=ResNet34, Student=ResNet182025.11 | 71.61 | 90.51 | |
| CAT-KDTeacher=ResNet34, Student=ResNet182025.11 | 71.61 | 90.45 | |
| SDDTeacher=ResNet34, Student=ResNet182025.11 | 71.44 | 90.05 | |
| LSKDTeacher=ResNet34, Student=ResNet182025.11 | 71.42 | 90.29 | |
| TeKAPTeacher=ResNet34, Student=ResNet182025.11 | 71.35 | 90.54 | |
| CTKDTeacher=ResNet34, Student=ResNet182025.11 | 71.32 | 90.27 | |
| CRDTeacher=ResNet50, Student=MobileNetV12025.11 | 71.32 | 90.41 | |
| WAVEModel Scale=L12H6, Para.(M)=22.1, FLOPs (G)=8.5, Training Epochs=302026.01 | 71.27 | 90.01 | |
| OFDTeacher=ResNet50, Student=MobileNetV12025.11 | 71.25 | 90.34 | |
| CRDTeacher=ResNet34, Student=ResNet182025.11 | 71.17 | 90.13 | |
| Iso. PruningModel Scale=L12H6, Para.(M)=22.1, FLOPs (G)=8.5, Training Epochs=302026.01 | 70.86 | 89.67 | |
| OFDTeacher=ResNet34, Student=ResNet182025.11 | 70.81 | 90.34 | |
| ATTeacher=ResNet34, Student=ResNet182025.11 | 70.69 | 90.01 | |
| KDTeacher=ResNet34, Student=ResNet182025.11 | 70.66 | 89.88 | |
| KDTeacher=ResNet50, Student=MobileNetV12025.11 | 70.49 | 89.92 | |
| SWEETModel Scale=L6H6, Para.(M)=11.4, FLOPs (G)=4.3, Training Epochs=302026.01 | 70.34 | 89.22 | |
| ATTeacher=ResNet50, Student=MobileNetV12025.11 | 70.18 | 89.68 | |
| From ScratchTeacher=ResNet34, Student=ResNet182025.11 | 69.75 | 89.07 | |
| From ScratchTeacher=ResNet50, Student=MobileNetV12025.11 | 68.87 | 88.76 | |
| SWEETModel Scale=Average, Training Epochs=302026.01 | 68.79 | 87.99 | |
| WAVEModel Scale=L6H6, Para.(M)=11.4, FLOPs (G)=4.3, Training Epochs=302026.01 | 68.38 | 88.16 | |
| Original (age)Backbone=ViT-B/16, Training Dataset=FairFace, Evaluation Attribute=age2024.11 | 68.31 | 91.83 | |
| Iso. PruningModel Scale=L6H6, Para.(M)=11.4, FLOPs (G)=4.3, Training Epochs=302026.01 | 68.25 | 88 | |
| SWEETModel Scale=L3H12, Para.(M)=22.8, FLOPs (G)=8.6, Training Epochs=302026.01 | 67.21 | 86.93 | |
| WAVEModel Scale=Average, Training Epochs=302026.01 | 67.19 | 87.12 | |
| WT-SelectModel Scale=L6H6, Para.(M)=11.4, FLOPs (G)=4.3, Training Epochs=302026.01 | 67.03 | 86.93 | |
| DMAEModel Scale=L3H12, Para.(M)=22.8, FLOPs (G)=8.6, Training Epochs=302026.01 | 66.67 | 86.71 | |
| WT-SelectModel Scale=L12H6, Para.(M)=22.1, FLOPs (G)=8.5, Training Epochs=302026.01 | 66.51 | 86.81 | |
| Iso. PruningModel Scale=L3H12, Para.(M)=22.8, FLOPs (G)=8.6, Training Epochs=302026.01 | 66.46 | 86.35 | |
| WT-SelectModel Scale=L3H12, Para.(M)=22.8, FLOPs (G)=8.6, Training Epochs=302026.01 | 66.44 | 86.46 | |
| DMAEModel Scale=L12H6, Para.(M)=22.1, FLOPs (G)=8.5, Training Epochs=302026.01 | 65.7 | 86.17 | |
| Iso. PruningModel Scale=Average, Training Epochs=302026.01 | 65.19 | 85.09 | |
| DMAEModel Scale=L6H6, Para.(M)=11.4, FLOPs (G)=4.3, Training Epochs=302026.01 | 64.86 | 85.65 | |
| DMAEModel Scale=Average, Training Epochs=302026.01 | 64.05 | 84.69 | |
| WT-SelectModel Scale=Average, Training Epochs=302026.01 | 63.87 | 84.24 | |
| WAVEModel Scale=L3H12, Para.(M)=22.8, FLOPs (G)=8.6, Training Epochs=302026.01 | 63.75 | 84.91 | |
| SWEETModel Scale=L12H3, Para.(M)=5.7, FLOPs (G)=2.2, Training Epochs=302026.01 | 57.28 | 80.31 | |
| WAVEModel Scale=L12H3, Para.(M)=5.7, FLOPs (G)=2.2, Training Epochs=302026.01 | 55.56 | 79.35 | |
| STELLARSetting=Zero-shot2026.02 | 51.53 | 80.04 | |
| iBOTSetting=Zero-shot2026.02 | 50.01 | 80.43 | |
| DMAEModel Scale=L12H3, Para.(M)=5.7, FLOPs (G)=2.2, Training Epochs=302026.01 | 49.07 | 73.77 | |
| Iso. PruningModel Scale=L12H3, Para.(M)=5.7, FLOPs (G)=2.2, Training Epochs=302026.01 | 44.53 | 68.89 | |
| WT-SelectModel Scale=L12H3, Para.(M)=5.7, FLOPs (G)=2.2, Training Epochs=302026.01 | 43.49 | 68.58 | |
| MAESetting=Zero-shot2026.02 | 23.18 | 50.43 | |
| Ours (age)Backbone=ViT-B/16, Training Dataset=FairFace, Evaluation Attribute=age2024.11 | — | 91.68 |