5-way Few-shot Image Classification on CIFAR-FS (test)
87.281-shot AccMCT (Pair)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MCT (Pair)Backbone=ResNet-12, Inference Protocol=Transductive2020.02 | 87.28 | 90.5 | — | |
| MCT (Instance)Backbone=ResNet-12, Inference Protocol=Transductive2020.02 | 85.61 | 90.03 | — | |
| SIBBackbone=WRN-28-10, Inference Protocol=Transductive2020.02 | 80 | 85.3 | — | |
| Transductive fine-tuningBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 78.36 | 87.54 | — | |
| DPGNBackbone=ResNet-12, Inference Protocol=Transductive2020.02 | 77.9 | 90.2 | — | |
| Ours-DistillBackbone=ResNet-122021.03 | 77.87 | 89.74 | — | |
| MCI (Instance)Backbone=ResNet-12, Inference Protocol=Inductive2020.02 | 77.84 | 89.11 | — | |
| SKD-GEN1Backbone=ResNet-12, Training Set=train+val2020.06 | 76.9 | 88.9 | — | |
| OursBackbone=ResNet-122021.03 | 76.83 | 89.26 | — | |
| Fine-tuningBackbone=WRN-28-10, Inference Protocol=Transductive2020.02 | 76.58 | 85.79 | — | |
| Transductive fine-tuningBackbone=WRN-28-10, Pre-training split=train2019.09 | 76.58 | 85.79 | — | |
| Fine-tuningBackbone=WRN-28-102021.03 | 76.58 | 85.79 | — | |
| MCI (Pair)Backbone=ResNet-12, Inference Protocol=Inductive2020.02 | 76.23 | 88.39 | — | |
| DSN-MRBackbone=ResNet-122021.03 | 75.6 | 86.2 | — | |
| SKD-GEN0Backbone=ResNet-12, Training Set=train+val2020.06 | 74.5 | 88 | — | |
| Transductive fine-tuningBackbone=ResNet-12, Pre-training split=train + val2019.09 | 74.35 | 84.57 | — | |
| RFS-distillBackbone=ResNet-12, Training Set=train+val2020.06 | 73.9 | 86.9 | — | |
| RFS-DistillBackbone=ResNet-122021.03 | 73.9 | 86.9 | — | |
| BoostingBackbone=WRN-28-10, Training Set=train+val2020.06 | 73.6 | 86 | — | |
| BoostingBackbone=WRN-28-102021.03 | 73.6 | 86 | — | |
| MABASBackbone=ResNet-122021.03 | 73.51 | 85.49 | — | |
| MetaOpt SVMBackbone=ResNet-12*, Pre-training split=train + val2019.09 | 72.8 | 85 | — | |
| MetaOptNetBackbone=ResNet-12, Training Set=train+val2020.06 | 72.6 | 84.3 | — | |
| MetaOptNetBackbone=ResNet-122021.03 | 72.6 | 84.3 | — | |
| MetaOptNetBackbone=ResNet-122021.06 | 72.6 | 84.3 | — | |
| Prototypical Networks+Backbone=ResNet-12, Training Set=train+val2020.06 | 72.2 | 83.5 | — | |
| Proto-Net+Backbone=ResNet-122021.03 | 72.2 | 83.5 | — | |
| ProtoNetBackbone=ResNet-122021.06 | 72.2 | 83.5 | — | |
| Support-based initializationBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 72.14 | 85.21 | — | |
| MetaOpt-SVMBackbone=ResNet-12, Inference Protocol=Inductive2020.02 | 72 | 84.2 | — | |
| MetaOpt SVMBackbone=ResNet-12*, Pre-training split=train2019.09 | 72 | 84.2 | — | |
| RFS-simpleBackbone=ResNet-12, Training Set=train+val2020.06 | 71.5 | 86 | — | |
| RFS-SimpleBackbone=ResNet-122021.03 | 71.5 | 86 | — | |
| Transductive fine-tuningBackbone=ResNet-12, Pre-training split=train2019.09 | 70.76 | 81.56 | — | |
| TEWAMBackbone=ResNet-12, Training Set=train+val2020.06 | 70.4 | 81.3 | — | |
| TEWAMBackbone=ResNet-122021.03 | 70.4 | 81.3 | — | |
| Support-based initializationBackbone=WRN-28-10, Pre-training split=train2019.09 | 70.26 | 83.82 | — | |
| Fine-tuningBackbone=WRN-28-10, Pre-training split=train + val2019.09 | 70.07 | 87.26 | — | |
| MTLBackbone=ResNet-122021.06 | 69.5 | 84.1 | — | |
| Support-based initializationBackbone=ResNet-12, Pre-training split=train + val2019.09 | 69.39 | 83.27 | — | |
| Shot-FreeBackbone=ResNet-12, Training Set=train+val2020.06 | 69.2 | 84.7 | — | |
| Shot-FreeBackbone=ResNet-122021.03 | 69.2 | 84.7 | — | |
| Fine-tuningBackbone=WRN-28-10, Pre-training split=train2019.09 | 68.72 | 86.11 | — | |
| Fine-tuningBackbone=ResNet-12, Pre-training split=train + val2019.09 | 68.11 | 85.19 | — | |
| Support-based initializationBackbone=ResNet-12, Pre-training split=train2019.09 | 65.69 | 79.95 | — | |
| R2D2Backbone=conv (96)x4, Pre-training split=train2019.09 | 65.4 | 79.4 | — | |
| R2D2Backbone=96-192-384-512, Training Set=train+val2020.06 | 65.3 | 79.4 | — | |
| R2D2Backbone=96-192-384-5122021.03 | 65.3 | 79.4 | — | |
| R2D2Backbone=96-192-384-5122021.06 | 65.3 | 79.4 | — | |
| Fine-tuningBackbone=ResNet-12, Pre-training split=train2019.09 | 64.66 | 82.13 | — | |
| Support-based initializationBackbone=conv (64)x4, Pre-training split=train + val2019.09 | 64.01 | 78.46 | — | |
| Transductive fine-tuningBackbone=conv (64)x4, Pre-training split=train + val2019.09 | 63.89 | 78.48 | — | |
| Fine-tuningBackbone=conv (64)x4, Pre-training split=train + val2019.09 | 62.12 | 77.98 | — | |
| Support-based initializationBackbone=conv (64)x4, Pre-training split=train2019.09 | 61.77 | 76.4 | — | |
| Transductive fine-tuningBackbone=conv (64)x4, Pre-training split=train2019.09 | 61.73 | 76.92 | — | |
| Fine-tuningBackbone=conv (64)x4, Pre-training split=train2019.09 | 59.74 | 76.37 | — | |
| MAMLBackbone=32-32-32-32, Training Set=train+val2020.06 | 58.9 | 71.5 | — | |
| MAMLBackbone=32-32-32-322021.03 | 58.9 | 71.5 | — | |
| MAMLBackbone=32-32-32-322021.06 | 58.9 | 71.5 | — | |
| Prototypical Networks+Backbone=64-64-64-64, Training Set=train+val2020.06 | 55.5 | 72 | — | |
| Proto-Net+Backbone=64-64-64-642021.03 | 55.5 | 72 | — | |
| Relation NetworksBackbone=64-96-128-256, Training Set=train+val2020.06 | 55 | 69.3 | — | |
| Relation NetBackbone=64-96-128-2562021.03 | 55 | 69.3 | — | |
| Adaptive PlacementBackbone=ViT-Small2026.06 | — | — | 87.55 | |
| Adaptive PlacementBackbone=DeiT-Small2026.06 | — | — | 87.64 | |
| Adaptive PlacementBackbone=Swin-Tiny2026.06 | — | — | 81.19 | |
| Adaptive PlasticityBackbone=ViT-Small2026.06 | — | — | 87.21 | |
| Adaptive PlasticityBackbone=DeiT-Small2026.06 | — | — | 84.38 | |
| Adaptive PlasticityBackbone=Swin-Tiny2026.06 | — | — | 81.09 | |
| BECLRNumber of shots=1-shot2024.02 | — | — | 70.39 | |
| BECLRNumber of shots=5-shot2024.02 | — | — | 81.56 | |
| DeiT-SmallHebbian Configuration=Base2026.06 | — | — | 60.56 | |
| Fixed HebbianBackbone=ViT-Small2026.06 | — | — | 72.58 | |
| Fixed HebbianBackbone=DeiT-Small2026.06 | — | — | 80.48 | |
| Fixed HebbianBackbone=Swin-Tiny2026.06 | — | — | 74.03 | |
| Fully Adaptive Hebbian RoutingBackbone=ViT-Small2026.06 | — | — | 88.59 | |
| Fully Adaptive Hebbian RoutingBackbone=DeiT-Small2026.06 | — | — | 86.5 | |
| Fully Adaptive Hebbian RoutingBackbone=Swin-Tiny2026.06 | — | — | 81.38 | |
| HMSNumber of shots=1-shot2024.02 | — | — | 54.65 | |
| HMSNumber of shots=5-shot2024.02 | — | — | 73.7 | |
| LF2CSNumber of shots=1-shot2024.02 | — | — | 55.04 | |
| LF2CSNumber of shots=5-shot2024.02 | — | — | 70.62 | |
| MoCo v2Number of shots=1-shot2024.02 | — | — | 52.73 | |
| MoCo v2Number of shots=5-shot2024.02 | — | — | 67.81 | |
| SimCLRNumber of shots=1-shot2024.02 | — | — | 54.56 | |
| SimCLRNumber of shots=5-shot2024.02 | — | — | 71.19 | |
| Swin-TinyHebbian Configuration=Base2026.06 | — | — | 52.52 | |
| ViT-SmallHebbian Configuration=Base2026.06 | — | — | 69.58 |