Image Classification on StanfordCars
92.74AccuracyFP32
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FP32Model=SigLIP2-B/16, Evaluation Protocol=Zero-shot2025.10 | 92.74 | — | — | — | — | — | |
| MMRLmode=Target Evaluation2025.05 | 91.43 | — | — | — | — | — | |
| MMRL++mode=Target Evaluation2025.05 | 91.43 | — | — | — | — | — | |
| TCPmode=Target Evaluation2025.05 | 91.25 | — | — | — | — | — | |
| FP32Model=SigLIP-B/16, Evaluation Protocol=Zero-shot2025.10 | 90.81 | — | — | — | — | — | |
| MaPLemode=Target Evaluation2025.05 | 90.49 | — | — | — | — | — | |
| MMAmode=Target Evaluation2025.05 | 90.3 | — | — | — | — | — | |
| PromptSRCmode=Target Evaluation2025.05 | 90.25 | — | — | — | — | — | |
| CoOpOpmode=Target Evaluation2025.05 | 90.14 | — | — | — | — | — | |
| Full-FTProjection Type=Baseline, Number of Parameters=7.00M2024.11 | 90.06 | — | — | — | — | — | |
| RegCacheModel=SigLIP-B/16, Evaluation Protocol=Zero-shot2025.10 | 89.73 | — | — | — | — | — | |
| FTProjection Type=Both-Proj, Number of Parameters=5.35M2024.11 | 89.67 | — | — | — | — | — | |
| FTProjection Type=In-Proj, Number of Parameters=3.58M2024.11 | 89.62 | — | — | — | — | — | |
| CoOpmode=Target Evaluation2025.05 | 89.14 | — | — | — | — | — | |
| SGD-DualOpt_Fine-tuneBackbone=ConvNeXt-B2026.04 | 88.99 | — | — | — | — | — | |
| FTBackbone=ViT-Large, Memory=4629.8 MB2026.05 | 88.9 | — | — | — | — | — | |
| FFBackbone=ViT-Large, Method=FF, # Params=303.3M2026.05 | 88.9 | — | — | — | — | — | |
| FTProjection Type=Out-Proj, Number of Parameters=1.81M2024.11 | 88.86 | — | — | — | — | — | |
| L2-SPBackbone=ConvNeXt-B2026.04 | 88.83 | — | — | — | — | — | |
| FullBackbone=ConvNeXt-B2026.04 | 88.67 | — | — | — | — | — | |
| RegCacheModel=SigLIP2-B/16, Evaluation Protocol=Zero-shot2025.10 | 88.2 | — | — | — | — | — | |
| FP32Model=OpenCLIP-B/16, Evaluation Protocol=Zero-shot2025.10 | 88.07 | — | — | — | — | — | |
| NaiveModel=SigLIP-B/16, Evaluation Protocol=Zero-shot2025.10 | 87.97 | — | — | — | — | — | |
| TPTEvaluation Protocol=Cross-dataset transfer evaluation2024.02 | 87.79 | — | — | — | — | — | |
| CERSABackbone=ViT-Large, Memory=1232.3 MB2026.05 | 87.6 | — | — | — | — | — | |
| PiSSABackbone=ViT-Large, Memory=1229.9 MB2026.05 | 86.7 | — | — | — | — | — | |
| IOTAShot=16-shot, Backbone=ViT-B/162026.01 | 86.41 | — | — | — | — | — | |
| RegCacheModel=OpenCLIP-B/16, Evaluation Protocol=Zero-shot2025.10 | 85.85 | — | — | — | — | — | |
| ProDiaLProjection Type=Both-Proj, Number of Parameters=0.67M2024.11 | 85.38 | — | — | — | — | — | |
| GPartBackbone=ViT-Large, Method=GPart (Ours), # Params=144K2026.05 | 85.2 | — | — | — | — | — | |
| DoRAProjection Type=Both-Proj, Number of Parameters=0.69M2024.11 | 85.18 | — | — | — | — | — | |
| LoRAProjection Type=Both-Proj, Number of Parameters=0.63M2024.11 | 85.06 | — | — | — | — | — | |
| SVFTBackbone=ViT-Large, Memory=1355.8 MB2026.05 | 84.9 | — | — | — | — | — | |
| StrongProjection Type=Baseline, Number of Parameters=0.96M2024.11 | 84.78 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Keff=Very Low2025.01 | 84.4 | — | — | — | — | — | |
| Hypernetwork (PASS)Channel Sparsity=10%, Transfer Source=Tiny-ImageNet2024.07 | 84.31 | — | — | — | — | — | |
| CERSABackbone=ViT-Base, α=0.95, β=0.952026.05 | 83.9 | — | — | — | — | — | |
| Transfer Channel Mask (PASS)Channel Sparsity=10%, Transfer Source=Tiny-ImageNet2024.07 | 83.5 | — | — | — | — | — | |
| SVFitBackbone=ViT-Large, Memory=1351.5 MB2026.05 | 83.3 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Scenario=Separate, Batch Size=1282025.01 | 83.2 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=Separate2025.01 | 83.2 | — | — | — | — | — | |
| Uni-LoRABackbone=ViT-Large, Method=Uni-LoRA, # Params=144K2026.05 | 83.16 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Scenario=High, Batch Size=1282025.01 | 82.9 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=High2025.01 | 82.9 | — | — | — | — | — | |
| ManiPT2026.02 | 82.53 | — | — | — | — | — | |
| ProDiaLProjection Type=In-Proj, Number of Parameters=0.42M2024.11 | 82.45 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Keff=Low2025.01 | 82.4 | — | — | — | — | — | |
| DoRAProjection Type=In-Proj, Number of Parameters=0.43M2024.11 | 82.17 | — | — | — | — | — | |
| LoRAProjection Type=In-Proj, Number of Parameters=0.41M2024.11 | 82.12 | — | — | — | — | — | |
| CLIP-Adapter (best α)Backbone=ViT-B/16, Shots=16, Blending ratio (alpha) selection strategy=Best residual ratio per dataset (oracle)2026.03 | 82.1 | — | — | — | — | — | |
| CLIP-AdapterBackbone=ViT-B/16, Shots=16, Blending ratio (alpha)=best2026.03 | 82.1 | — | — | — | — | — | |
| Adam-DualOpt_Fine-tuneBackbone=ViT-B2026.04 | 82.02 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, K_eff=High2025.01 | 82 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Scenario=Medium, Batch Size=1282025.01 | 81.9 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=Medium2025.01 | 81.9 | — | — | — | — | — | |
| CS-AlignerBackbone=ViT-B/16, Number of Shots=16, Evaluation Protocol=fine-tuned2025.02 | 81.9 | — | — | — | — | — | |
| TAC2026.02 | 81.63 | — | — | — | — | — | |
| LLaMP2026.02 | 81.56 | — | — | — | — | — | |
| HOSO-AdapterBackbone=ViT-B/16, Shots=16, Blending ratio (alpha) selection strategy=Optimised on hold-one-shot-out cache2026.03 | 81.5 | — | — | — | — | — | |
| HOSO-AdapterBackbone=ViT-B/16, Shots=16, Number of runs=32026.03 | 81.5 | — | — | — | — | — | |
| IOTAShot=8-shot, Backbone=ViT-B/162026.01 | 80.82 | — | — | — | — | — | |
| TAP2026.02 | 80.7 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Keff=Medium2025.01 | 80.6 | — | — | — | — | — | |
| LoRABackbone=ViT-Large, Memory=1229.9 MB2026.05 | 80.3 | — | — | — | — | — | |
| ProTextShot=16-shot, Backbone=ViT-B/162026.01 | 80.23 | — | — | — | — | — | |
| Hypernetwork (PASS)Channel Sparsity=30%, Transfer Source=Tiny-ImageNet2024.07 | 79.88 | — | — | — | — | — | |
| CoOpShot=16-shot, Backbone=ViT-B/162026.01 | 79.84 | — | — | — | — | — | |
| FTBackbone=ViT-Base2026.05 | 79.8 | — | — | — | — | — | |
| FFBackbone=ViT-Base, Method=FF, # Params=85.8M2026.05 | 79.78 | — | — | — | — | — | |
| LP++Shot=16-shot, Backbone=ViT-B/162026.01 | 79.75 | — | — | — | — | — | |
| Transfer Channel Mask (PASS)Channel Sparsity=30%, Transfer Source=Tiny-ImageNet2024.07 | 79.7 | — | — | — | — | — | |
| FourierFTBackbone=ViT-Large, Method=FourierFT, # Params=480K2026.05 | 79.14 | — | — | — | — | — | |
| CoFT+mode=unsupervised fine-tuning2026.02 | 79.13 | — | — | — | — | — | |
| CLIP-1+DINOv3Setting=Inductive, Test-Time Adaptation (TTA)=false2025.06 | 78.8 | — | — | — | — | — | |
| TCPShot=16-shot, Backbone=ViT-B/162026.01 | 78.66 | — | — | — | — | — | |
| CLIP-AdapterBackbone=ViT-B/16, Number of Shots=16, Evaluation Protocol=fine-tuned2025.02 | 78.6 | — | — | — | — | — | |
| PromptSRC2026.02 | 78.27 | — | — | — | — | — | |
| CoOp2026.02 | 78.12 | — | — | — | — | — | |
| StatAEncoder=ViT-L/14, K_eff=All2025.01 | 78 | — | — | — | — | — | |
| ProDiaLProjection Type=Out-Proj, Number of Parameters=0.27M2024.11 | 78 | — | — | — | — | — | |
| Clip-AdapterShot=16-shot, Backbone=ViT-B/162026.01 | 77.99 | — | — | — | — | — | |
| KgCoOpShot=16-shot, Backbone=ViT-B/162026.01 | 77.85 | — | — | — | — | — | |
| LoRAProjection Type=Out-Proj, Number of Parameters=0.26M2024.11 | 77.81 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Scenario=Low, Batch Size=1282025.01 | 77.7 | — | — | — | — | — | |
| StatABackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=true, Scenario=Low2025.01 | 77.7 | — | — | — | — | — | |
| DoRAProjection Type=Out-Proj, Number of Parameters=0.28M2024.11 | 77.7 | — | — | — | — | — | |
| CLIP-1+DINO(v2+v3)Setting=Inductive, Test-Time Adaptation (TTA)=false2025.06 | 77.5 | — | — | — | — | — | |
| GPartBackbone=ViT-Base, Method=GPart (Ours), # Params=72K2026.05 | 77.12 | — | — | — | — | — | |
| CoFTmode=unsupervised fine-tuning2026.02 | 77.08 | — | — | — | — | — | |
| CoPrompt2026.02 | 76.97 | — | — | — | — | — | |
| CLIPBackbone=ViT-L/142025.01 | 76.9 | — | — | — | — | — | |
| CLIPEncoder=ViT-L/14, K_eff=All2025.01 | 76.9 | — | — | — | — | — | |
| CLIPBackbone=ViT-L/14, Setting=Zero-shot, Batch Size=1282025.01 | 76.9 | — | — | — | — | — | |
| CLIPBackbone=ViT-L/14, Batch Size=128, Online Test-Time Adaptation=false2025.01 | 76.9 | — | — | — | — | — | |
| Hypernetwork (PASS)Channel Sparsity=50%, Transfer Source=Tiny-ImageNet2024.07 | 76.8 | — | — | — | — | — | |
| Transfer Channel Mask (PASS)Channel Sparsity=50%, Transfer Source=Tiny-ImageNet2024.07 | 76.6 | — | — | — | — | — | |
| MapleShot=16-shot, Backbone=ViT-B/162026.01 | 76.58 | — | — | — | — | — | |
| CoOpShot=8-shot, Backbone=ViT-B/162026.01 | 76.5 | — | — | — | — | — | |
| Group-L1Channel Sparsity=10%, Transfer Source=Tiny-ImageNet2024.07 | 76.5 | — | — | — | — | — | |
| KgCoOpShot=8-shot, Backbone=ViT-B/162026.01 | 76.46 | — | — | — | — | — |