Image Classification on UCF101
87.7Top-1 AccSkip Tuning
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Skip TuningShot=16, Time (s)=962, Mem. (M)=5282024.12 | 87.7 | — | |
| MMRLShots=16 shots2025.03 | 87.6 | — | |
| AWTShots=162024.07 | 87.53 | — | |
| GalLoP+OursShots=42026.05 | 87.3 | — | |
| PromptSRC+OursShots=42026.05 | 87.2 | — | |
| CLIP-LoRA + FMAShots=162025.10 | 87.1 | — | |
| TCP+OursShots=42026.05 | 87.1 | — | |
| GalLoPShots=42026.05 | 86.9 | — | |
| PromptSRCShots=42026.05 | 86.5 | — | |
| ProSRCShot=16, Time (s)=14403, Mem. (M)=33732024.12 | 86.47 | — | |
| PromptSRCShots=16 shots2025.03 | 86.47 | — | |
| MMAShots=16 shots2025.03 | 86.3 | — | |
| TCP*Shots=42026.05 | 86.3 | — | |
| CLIP-LoRAShots=162025.10 | 86.2 | — | |
| IOTAShot=16-shot, Backbone=ViT-B/162026.01 | 86.02 | — | |
| SCANShot=162025.05 | 85.78 | — | |
| Dual Prompt TuningBackbone=ViT-L/14, PEFT Paradigm=CoOp, Annotation Budget (B%)=5%2026.02 | 85.7 | — | |
| Skip TuningShot=8, Time (s)=508, Mem. (M)=5282024.12 | 85.6 | — | |
| PLOT++Shots=162025.10 | 85.3 | — | |
| PLOTShots=42026.05 | 85.3 | — | |
| ProDA2023.11 | 85.23 | — | |
| MaPLeShot=16, Time (s)=14070, Mem. (M)=30612024.12 | 85.03 | — | |
| MaPLeShots=16 shots2025.03 | 85.03 | — | |
| OURSShots=8, Backbone=ViT-B/16, Templates=12025.12 | 85 | — | |
| MaPLeShots=42026.05 | 85 | — | |
| BASIC-LBackbone=BASIC-L2021.11 | 84.8 | — | |
| MMRLShots=8 shots2025.03 | 84.67 | — | |
| ProSRCShot=8, Time (s)=7636, Mem. (M)=33732024.12 | 84.3 | — | |
| PromptSRCShots=8 shots2025.03 | 84.3 | — | |
| ProTextShot=16-shot, Backbone=ViT-B/162026.01 | 84.22 | — | |
| OURSShots=8, Backbone=ViT-B/16, Templates=72025.12 | 84.2 | — | |
| PromptAlign2023.11 | 84.11 | — | |
| CLIP-LoRAShots=8, Backbone=ViT-B/16, Templates=12025.12 | 84.1 | — | |
| CS-AlignerBackbone=ViT-B/16, Number of Shots=16, Evaluation Protocol=fine-tuned2025.02 | 84 | — | |
| CLIP-Adapter (best α)Backbone=ViT-B/16, Shots=16, Blending ratio (alpha) selection strategy=Best residual ratio per dataset (oracle)2026.03 | 84 | — | |
| TIP-AdapterShots=162025.10 | 83.9 | — | |
| KgCoOpShot=16-shot, Backbone=ViT-B/162026.01 | 83.72 | — | |
| TCPShot=16-shot, Backbone=ViT-B/162026.01 | 83.64 | — | |
| MMAShots=8 shots2025.03 | 83.43 | — | |
| HOSO-AdapterBackbone=ViT-B/16, Shots=16, Blending ratio (alpha) selection strategy=Optimised on hold-one-shot-out cache2026.03 | 83.43 | — | |
| LP++Shot=16-shot, Backbone=ViT-B/162026.01 | 83.35 | — | |
| ProDAShots=42026.05 | 83.3 | — | |
| CLIP-LoRAShots=8, Backbone=ViT-B/16, Templates=72025.12 | 83.2 | — | |
| CoOpShots=162025.10 | 83.1 | — | |
| MaPLe2023.11 | 82.97 | — | |
| BiCLIPShots=16, Base Model=CLIP, Adaptation Method=Bilinear adaptation2026.03 | 82.95 | 14.88 | |
| CLIP-AdapterBackbone=ViT-B/16, Number of Shots=16, Evaluation Protocol=fine-tuned2025.02 | 82.8 | — | |
| Clip-AdapterShot=16-shot, Backbone=ViT-B/162026.01 | 82.77 | — | |
| ProGradShots=162025.10 | 82.7 | — | |
| MMRLShots=4 shots2025.03 | 82.67 | — | |
| Skip TuningShot=4, Time (s)=280, Mem. (M)=5282024.12 | 82.53 | — | |
| OURSShots=4, Backbone=ViT-B/16, Templates=72025.12 | 82.4 | — | |
| CLIP-LoRA + FMAShots=42025.10 | 82.4 | — | |
| CoOpShot=16-shot, Backbone=ViT-B/162026.01 | 82.37 | — | |
| SCANShot=82025.05 | 82.34 | — | |
| MaPLe+TPTtest-time prompt tuning=true2023.11 | 82.23 | — | |
| CoOpShot=16, Time (s)=3879, Mem. (M)=45512024.12 | 82.23 | — | |
| CoOpShots=16 shots2025.03 | 82.23 | — | |
| OURSShots=4, Backbone=ViT-B/16, Templates=12025.12 | 82.2 | — | |
| CoOpShots=42026.05 | 82.2 | — | |
| Linear probe CLIPShots=16 shots2025.03 | 82.11 | — | |
| TransCLIP-FSShots=16, Backbone=ViT-B/162024.06 | 82.1 | — | |
| CLIP-LoRA+OursShots=42026.05 | 82.1 | — | |
| MapleShot=16-shot, Backbone=ViT-B/162026.01 | 81.89 | — | |
| IOTAShot=8-shot, Backbone=ViT-B/162026.01 | 81.89 | — | |
| KgCoOpShots=162025.10 | 81.7 | — | |
| LoCoOpShots=42026.05 | 81.6 | — | |
| GPT-4V2024.03 | 81.6 | — | |
| ProSRCShot=4, Time (s)=3841, Mem. (M)=33732024.12 | 81.57 | — | |
| PromptSRCShots=4 shots2025.03 | 81.57 | — | |
| TCPShot=8-shot, Backbone=ViT-B/162026.01 | 81.53 | — | |
| RARBase Model=Qwen-VL2024.03 | 81.4 | — | |
| MaPLeShot=8, Time (s)=7060, Mem. (M)=30612024.12 | 81.37 | — | |
| MaPLeShots=8 shots2025.03 | 81.37 | — | |
| KgCoOpShot=8-shot, Backbone=ViT-B/162026.01 | 81.36 | — | |
| StatAEncoder=ViT-L/14, K_eff=All2025.01 | 81.3 | — | |
| LoRA-CLIPShots=4, Backbone=ViT-B/16, Templates=12025.12 | 81.1 | — | |
| CLIP-LoRAShots=42025.10 | 81.1 | — | |
| ProTextShot=8-shot, Backbone=ViT-B/162026.01 | 81.08 | — | |
| LP++Shot=8-shot, Backbone=ViT-B/162026.01 | 80.92 | — | |
| CLIP-LoRAShots=42026.05 | 80.9 | — | |
| Clip-AdapterShot=8-shot, Backbone=ViT-B/162026.01 | 80.7 | — | |
| Dual Prompt TuningBackbone=ViT-L/14, PEFT Paradigm=CoOp, Annotation Budget (B%)=2%2026.02 | 80.6 | — | |
| CLIP-Adapter (α = 0.2)Backbone=ViT-B/16, Shots=16, Blending ratio (alpha) selection strategy=Fixed residual ratio (α = 0.2), selected on ImageNet2026.03 | 80.4 | — | |
| LoRA-CLIPShots=4, Backbone=ViT-B/16, Templates=72025.12 | 80.3 | — | |
| SimNLShot=162025.05 | 80.28 | — | |
| OSTBProtocol=Final Ensemble2026.06 | 80.28 | — | |
| CoFT+mode=unsupervised fine-tuning2026.02 | 80.23 | — | |
| CoOpShot=8, Time (s)=1988, Mem. (M)=45512024.12 | 80.2 | — | |
| CoOpShots=8 shots2025.03 | 80.2 | — | |
| CLIP-AdapterShots=162025.10 | 80.2 | — | |
| MMAShots=4 shots2025.03 | 80.1 | — | |
| CoOpShot=8-shot, Backbone=ViT-B/162026.01 | 79.88 | — | |
| PEFTShots=102023.05 | 79.8 | — | |
| PLOT++Shots=42025.10 | 79.8 | — | |
| Co-CoOpShot=16-shot, Backbone=ViT-B/162026.01 | 79.46 | — | |
| Linear probe CLIPShots=8 shots2025.03 | 79.34 | — | |
| MapleShot=8-shot, Backbone=ViT-B/162026.01 | 79.12 | — | |
| TransCLIP-FSShots=4, Backbone=ViT-B/162024.06 | 79.1 | — | |
| RARBase Model=LLaVA1.52024.03 | 79.1 | — |