Image Classification on EuroSAT (4-shot)
82.4AccuracyProKeR
Evaluation results
| Method | Links | |
|---|---|---|
| ProKeRModel=DFN, Hyperparameter Tuning=true2026.03 | 82.4 | |
| HEC-VTModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=true2026.03 | 82 | |
| ProbingModel=DFN, Hyperparameter Tuning=true2026.03 | 81.6 | |
| HEC-VModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=false2026.03 | 78.8 | |
| GDAModel=DFN, Hyperparameter Tuning=true2026.03 | 78 | |
| ProKeRModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=true2026.03 | 74.4 | |
| ProbingModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=true2026.03 | 74 | |
| Tip-AdapterProtocol=Inductive, Shots=4, Backbone=CLIP2024.12 | 72.5 | |
| SAVsModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=false2026.03 | 72 | |
| TipAdapterModel=DFN, Hyperparameter Tuning=true2026.03 | 69.2 | |
| GDAModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=true2026.03 | 69.2 | |
| CoOpProtocol=Inductive, Shots=4, Backbone=CLIP2024.12 | 63.2 | |
| HEC-TModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=false2026.03 | 55.6 | |
| UNEM-DirichletProtocol=Transductive, Shots=4, Backbone=CLIP2024.12 | 53.8 | |
| Laplacian ShotProtocol=Transductive, Shots=4, Backbone=CLIP2024.12 | 53 | |
| Zero-ShotModel=DFN, Hyperparameter Tuning=false2026.03 | 51.6 | |
| EM-DirichletProtocol=Transductive, Shots=4, Backbone=CLIP2024.12 | 50.8 | |
| TipAdapterModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=true2026.03 | 50.4 | |
| Zero-ShotModel=DFN+LLM (Qwen2-VL), Hyperparameter Tuning=false2026.03 | 48.8 | |
| BDSCPNProtocol=Transductive, Shots=4, Backbone=CLIP2024.12 | 46.1 | |
| α-TIMProtocol=Transductive, Shots=4, Backbone=CLIP2024.12 | 46.1 | |
| PADDLEProtocol=Transductive, Shots=4, Backbone=CLIP2024.12 | 45.9 |