Image Classification on SUN
77.6Accuracy+DP-FM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| +DP-FMShot=162026.05 | 77.6 | — | |
| TIP-AdapterShot=162026.05 | 76 | — | |
| PLOT++Shot=162026.05 | 76 | — | |
| CLIP-LoRAShot=162026.05 | 76 | — | |
| ProGradShot=162026.05 | 75.1 | — | |
| CoOpShot=162026.05 | 74.9 | — | |
| TransCLIP-FSShots=16, Backbone=ViT-B/16, Text Regularization=KL penalty2024.06 | 74.7 | — | |
| CLIP-AdapterShot=162026.05 | 74.2 | — | |
| GPUASamples per class=full training set2026.06 | 74.2 | — | |
| +DP-FMShot=42026.05 | 73.4 | — | |
| KgCoOpShot=162026.05 | 73.3 | — | |
| CLIP-LoRAShot=42026.05 | 72.8 | — | |
| CoCoOpShot=162026.05 | 72.8 | — | |
| DINOv2-ARLN=20M, Zero-shot=true2024.09 | 72.6 | — | |
| GPUA*Samples per class=162026.06 | 72.6 | — | |
| COSMICInference-time adaptation=true2026.06 | 72.3 | — | |
| InMaP + ZLaPBackbone=ViT-B/16, Evaluation Protocol=Transductive Zero-Shot2024.04 | 71.9 | 0.6 | |
| TransCLIP-FSShots=4, Backbone=ViT-B/16, Text Regularization=KL penalty2024.06 | 71.9 | — | |
| DINOv2-MpNetN=20M, Zero-shot=true2024.09 | 71.9 | — | |
| PLOT++Shot=42026.05 | 71.7 | — | |
| ProGradShot=42026.05 | 71.7 | — | |
| KgCoOpShot=42026.05 | 71.5 | — | |
| InMaPBackbone=ViT-B/16, Evaluation Protocol=Transductive Zero-Shot2024.04 | 71.3 | — | |
| +DP-FMShot=12026.05 | 71 | — | |
| LAION-CLIP VIT-LN=400M, Zero-shot=true2024.09 | 70.9 | — | |
| TIP-AdapterShot=42026.05 | 70.8 | — | |
| TransCLIP-FSShots=1, Backbone=ViT-B/16, Text Regularization=KL penalty2024.06 | 70.6 | — | |
| CoCoOpShot=42026.05 | 70.4 | — | |
| CLIP-LoRAShot=12026.05 | 70.3 | — | |
| DMN2026.06 | 70.2 | — | |
| DPE2026.06 | 70.1 | — | |
| CoOpShot=42026.05 | 69.7 | — | |
| CoCoOpShot=12026.05 | 68.7 | — | |
| StatA2026.06 | 68.7 | — | |
| KgCoOpShot=12026.05 | 68.4 | — | |
| TIPPLEInference-time adaptation=true2026.06 | 68.1 | — | |
| CLIP-AdapterShot=42026.05 | 68 | — | |
| CLIP + ZLaPBackbone=ViT-B/16, Evaluation Protocol=Transductive Zero-Shot2024.04 | 67.8 | 1.5 | |
| OpenAI-CLIP VIT-LN=400M, Zero-shot=true2024.09 | 67.7 | — | |
| ZERO2026.06 | 67.6 | — | |
| TDA2026.06 | 67.6 | — | |
| ZLaP2026.06 | 67.4 | — | |
| TIP-AdapterShot=12026.05 | 67.2 | — | |
| CoOpShot=12026.05 | 67 | — | |
| ProGradShot=12026.05 | 67 | — | |
| CoCoOpFMA=true2025.10 | 66.9 | — | |
| PLOT++Shot=12026.05 | 66.8 | — | |
| MTA2026.06 | 66.7 | — | |
| CLIP-DNBackbone=ViT-B/16, Evaluation Protocol=Transductive Zero-Shot2024.04 | 66.5 | — | |
| LoRAFMA=false2025.10 | 66.5 | — | |
| CoCoOpFMA=false2025.10 | 66.5 | — | |
| CLIPBackbone=ViT-B/16, Evaluation Protocol=Transductive Zero-Shot2024.04 | 66.3 | — | |
| LoRAFMA=true2025.10 | 66.3 | — | |
| InMaP + ZLaPBackbone=ResNet-50, Evaluation Protocol=Transductive Zero-Shot2024.04 | 65.9 | 0.6 | |
| TPTBackbone=ViT-B/16, Evaluation Protocol=Transductive Zero-Shot2024.04 | 65.5 | — | |
| CLIP-AdapterShot=12026.05 | 65.4 | — | |
| InMaPBackbone=ResNet-50, Evaluation Protocol=Transductive Zero-Shot2024.04 | 65.3 | — | |
| AdapterFMA=true2025.10 | 63.9 | — | |
| AdapterFMA=false2025.10 | 63.6 | — | |
| CLIP + ZLaPBackbone=ResNet-50, Evaluation Protocol=Transductive Zero-Shot2024.04 | 63.3 | 2.1 | |
| CLIPFMA=true2025.10 | 63.1 | — | |
| CLIPFMA=false2025.10 | 62.6 | — | |
| CLIPShot=02026.05 | 62.6 | — | |
| CLIP-ViT-B/16Shots=0, Backbone=ViT-B/162024.06 | 62.5 | — | |
| LiT L16LN=112M, Zero-shot=true2024.09 | 62.5 | — | |
| CLIP2026.06 | 62.5 | — | |
| TPTBackbone=ResNet-50, Evaluation Protocol=Transductive Zero-Shot2024.04 | 61.5 | — | |
| CLIPBackbone=ResNet-50, Evaluation Protocol=Transductive Zero-Shot2024.04 | 61.2 | — | |
| CLIP-DNBackbone=ResNet-50, Evaluation Protocol=Transductive Zero-Shot2024.04 | 60.9 | — | |
| CLIPzero-shot=true2021.02 | 58.5 | — | |
| CoOpFMA=false2025.10 | 58 | — | |
| CoOpFMA=true2025.10 | 57.6 | — | |
| X-InstructBLIP2024.10 | 42.8 | — | |
| ChatBridge2024.10 | 42.2 | — | |
| Ours2024.10 | 42.2 | — | |
| OneLLM2024.10 | 29.4 | — | |
| Visual N-Gramszero-shot=true2021.02 | 23 | — |