Image Classification on Resisc45 (test)
99.08Top-1 AccuracyRS-M-CLIP + multi-VLM GMM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RS-M-CLIP + multi-VLM GMMBackbone Model=RS-M-CLIP, Adaptation Protocol=multi-VLM GMM2026.06 | 99.08 | — | |
| RS-M-CLIP + single-VLM GMMBackbone Model=RS-M-CLIP, Adaptation Protocol=single-VLM GMM2026.06 | 99.06 | — | |
| OSTBAdaptation Protocol=OSTB Adapted Ensemble2026.06 | 98.8 | — | |
| RSDiX-CLIP-B/32 + single-VLM GMMBackbone Model=RSDiX-CLIP-B/32, Adaptation Protocol=single-VLM GMM2026.06 | 98.34 | — | |
| RSDiX-CLIP-B/32 + multi-VLM GMMBackbone Model=RSDiX-CLIP-B/32, Adaptation Protocol=multi-VLM GMM2026.06 | 98.33 | — | |
| RSDiX-CLIP-B/16 + single-VLM GMMBackbone Model=RSDiX-CLIP-B/16, Adaptation Protocol=single-VLM GMM2026.06 | 98.11 | — | |
| RSDiX-CLIP-B/16 + multi-VLM GMMBackbone Model=RSDiX-CLIP-B/16, Adaptation Protocol=multi-VLM GMM2026.06 | 98.1 | — | |
| RSDiX-CLIP-B/16Backbone Model=RSDiX-CLIP-B/16, Adaptation Protocol=Original Zero-shot2026.06 | 97.83 | — | |
| RSDiX-CLIP-B/32Backbone Model=RSDiX-CLIP-B/32, Adaptation Protocol=Original Zero-shot2026.06 | 97.04 | — | |
| RS-M-CLIPBackbone Model=RS-M-CLIP, Adaptation Protocol=Original Zero-shot2026.06 | 95.82 | — | |
| Fine-tuned (STL)Backbone=CLIP-ViT-B/32, Evaluation Protocol=Fine-tuned (STL)2026.05 | 95.2 | — | |
| WUDI-Merging w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging, Refinement=FEATCAL2026.05 | 93 | — | |
| Traditional MTLBackbone=CLIP-ViT-B/32, Evaluation Protocol=Traditional MTL2026.05 | 92.2 | — | |
| AdaMerging w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging, Refinement=FEATCAL2026.05 | 91.3 | — | |
| WUDI-Merging w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging, Refinement=ProbSurgery2026.05 | 90.6 | — | |
| AdaMerging w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging, Refinement=ProbSurgery2026.05 | 90.4 | — | |
| WUDI-Merging w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging, Refinement=Surgery2026.05 | 89.1 | — | |
| AdaMerging w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging, Refinement=Surgery2026.05 | 88.7 | — | |
| GeoRSCLIP-B/32 + multi-VLM GMMBackbone Model=GeoRSCLIP-B/32, Adaptation Protocol=multi-VLM GMM2026.06 | 88.32 | — | |
| Task Arithmetic w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic, Refinement=FEATCAL2026.05 | 88.1 | — | |
| StreetCLIP + multi-VLM GMMBackbone Model=StreetCLIP, Adaptation Protocol=multi-VLM GMM2026.06 | 87.84 | — | |
| AiRPrompting Strategy=Visual Prompt, Learning Paradigm=TRZSL2025.04 | 87.8 | — | |
| AiRPrompting Strategy=Text Prompt, Learning Paradigm=TRZSL2025.04 | 87.5 | — | |
| CLIP-B/16 + multi-VLM GMMBackbone Model=CLIP-B/16, Adaptation Protocol=multi-VLM GMM2026.06 | 86.88 | — | |
| RemoteCLIP-B/32 + multi-VLM GMMBackbone Model=RemoteCLIP-B/32, Adaptation Protocol=multi-VLM GMM2026.06 | 86.68 | — | |
| CPLYear=2024, Prompting Strategy=Visual Prompt, Learning Paradigm=TRZSL2025.04 | 86.6 | — | |
| SkyCLIP50-B/32 + multi-VLM GMMBackbone Model=SkyCLIP50-B/32, Adaptation Protocol=multi-VLM GMM2026.06 | 86.33 | — | |
| asymmetric knowledge distillation frameworkDistortion type=Gaussian blur with kernel size 37 and σ = 9 (GB), Evaluation protocol=Transfer learning2026.04 | 86.03 | — | |
| CPLYear=2024, Prompting Strategy=Text Prompt, Learning Paradigm=TRZSL2025.04 | 85.8 | — | |
| Simple Averaging w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging, Refinement=ProbSurgery2026.05 | 85.8 | — | |
| Simple Averaging w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging, Refinement=FEATCAL2026.05 | 85 | — | |
| WUDI-MergingBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging2026.05 | 85 | — | |
| BaselineDistortion type=Gaussian blur with kernel size 37 and σ = 9 (GB), Evaluation protocol=Transfer learning2026.04 | 84.68 | — | |
| AdaMergingBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging2026.05 | 84 | — | |
| asymmetric knowledge distillation frameworkDistortion type=90% random masking (RM), Evaluation protocol=Transfer learning2026.04 | 83.75 | — | |
| asymmetric knowledge distillation frameworkDistortion type=Gaussian noise with σ = 0.5 (GN), Evaluation protocol=Transfer learning2026.04 | 83.53 | — | |
| CLIP-B/32 + multi-VLM GMMBackbone Model=CLIP-B/32, Adaptation Protocol=multi-VLM GMM2026.06 | 83.28 | — | |
| AiRPrompting Strategy=Text Prompt, Learning Paradigm=SSL2025.04 | 82.5 | — | |
| GRIPYear=2023, Prompting Strategy=Visual Prompt, Learning Paradigm=TRZSL2025.04 | 82.1 | — | |
| RemoteCLIP-B/32 + single-VLM GMMBackbone Model=RemoteCLIP-B/32, Adaptation Protocol=single-VLM GMM2026.06 | 81.28 | — | |
| BaselineDistortion type=Gaussian noise with σ = 0.5 (GN), Evaluation protocol=Transfer learning2026.04 | 81.21 | — | |
| GRIPYear=2023, Prompting Strategy=Text Prompt, Learning Paradigm=TRZSL2025.04 | 81.1 | — | |
| AiRPrompting Strategy=Visual Prompt, Learning Paradigm=SSL2025.04 | 81.1 | — | |
| Task Arithmetic w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic, Refinement=ProbSurgery2026.05 | 81 | — | |
| CPLYear=2024, Prompting Strategy=Text Prompt, Learning Paradigm=SSL2025.04 | 80.9 | — | |
| BaselineDistortion type=90% random masking (RM), Evaluation protocol=Transfer learning2026.04 | 80.88 | — | |
| Simple Averaging w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging, Refinement=Surgery2026.05 | 80.5 | — | |
| AiRPrompting Strategy=Text Prompt, Learning Paradigm=UL2025.04 | 79.9 | — | |
| CIDistortion type=90% random masking (RM), Evaluation protocol=Transfer learning2026.04 | 79.79 | — | |
| StreetCLIP + single-VLM GMMBackbone Model=StreetCLIP, Adaptation Protocol=single-VLM GMM2026.06 | 79.45 | — | |
| AiRPrompting Strategy=Visual Prompt, Learning Paradigm=UL2025.04 | 79.3 | — | |
| CPLYear=2024, Prompting Strategy=Visual Prompt, Learning Paradigm=SSL2025.04 | 78.4 | — | |
| CPLYear=2024, Prompting Strategy=Text Prompt, Learning Paradigm=UL2025.04 | 77.3 | — | |
| SkyCLIP50-B/32 + single-VLM GMMBackbone Model=SkyCLIP50-B/32, Adaptation Protocol=single-VLM GMM2026.06 | 77.1 | — | |
| GeoRSCLIP-B/32 + single-VLM GMMBackbone Model=GeoRSCLIP-B/32, Adaptation Protocol=single-VLM GMM2026.06 | 76.48 | — | |
| Task Arithmetic w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic, Refinement=Surgery2026.05 | 76.1 | — | |
| CIDistortion type=Gaussian blur with kernel size 37 and σ = 9 (GB), Evaluation protocol=Transfer learning2026.04 | 74.6 | — | |
| CLIP-B/16 + single-VLM GMMBackbone Model=CLIP-B/16, Adaptation Protocol=single-VLM GMM2026.06 | 74.51 | — | |
| GRIPYear=2023, Prompting Strategy=Text Prompt, Learning Paradigm=SSL2025.04 | 74.1 | — | |
| CLIP-B/32 + single-VLM GMMBackbone Model=CLIP-B/32, Adaptation Protocol=single-VLM GMM2026.06 | 73.1 | — | |
| CPLYear=2024, Prompting Strategy=Visual Prompt, Learning Paradigm=UL2025.04 | 72.9 | — | |
| FPLYear=2023, Prompting Strategy=Text Prompt, Learning Paradigm=TRZSL2025.04 | 72.1 | — | |
| GRIPYear=2023, Prompting Strategy=Visual Prompt, Learning Paradigm=SSL2025.04 | 71.2 | — | |
| GRIPYear=2023, Prompting Strategy=Text Prompt, Learning Paradigm=UL2025.04 | 70.6 | — | |
| Simple AveragingBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging2026.05 | 70.6 | — | |
| StreetCLIPBackbone Model=StreetCLIP, Adaptation Protocol=Original Zero-shot2026.06 | 69.61 | — | |
| GeoRSCLIP-B/32Backbone Model=GeoRSCLIP-B/32, Adaptation Protocol=Original Zero-shot2026.06 | 68.49 | — | |
| GRIPYear=2023, Prompting Strategy=Visual Prompt, Learning Paradigm=UL2025.04 | 68.4 | — | |
| CIDistortion type=Gaussian noise with σ = 0.5 (GN), Evaluation protocol=Transfer learning2026.04 | 68.11 | — | |
| FPLYear=2023, Prompting Strategy=Text Prompt, Learning Paradigm=SSL2025.04 | 68.1 | — | |
| RemoteCLIP-B/32Backbone Model=RemoteCLIP-B/32, Adaptation Protocol=Original Zero-shot2026.06 | 68.04 | — | |
| FPLYear=2023, Prompting Strategy=Visual Prompt, Learning Paradigm=TRZSL2025.04 | 67.8 | — | |
| VPTYear=2022, Prompting Strategy=Visual Prompt, Learning Paradigm=TRZSL2025.04 | 67.1 | — | |
| SkyCLIP50-B/32Backbone Model=SkyCLIP50-B/32, Adaptation Protocol=Original Zero-shot2026.06 | 66.57 | — | |
| FPLYear=2023, Prompting Strategy=Visual Prompt, Learning Paradigm=SSL2025.04 | 65.1 | — | |
| Task ArithmeticBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic2026.05 | 64.7 | — | |
| CLIP-B/16Backbone Model=CLIP-B/16, Adaptation Protocol=Original Zero-shot2026.06 | 64.01 | — | |
| CoOpYear=2022, Prompting Strategy=Text Prompt, Learning Paradigm=TRZSL2025.04 | 63.4 | — | |
| FPLYear=2023, Prompting Strategy=Text Prompt, Learning Paradigm=UL2025.04 | 63 | — | |
| FPLYear=2023, Prompting Strategy=Visual Prompt, Learning Paradigm=UL2025.04 | 62.2 | — | |
| VPTYear=2022, Prompting Strategy=Visual Prompt, Learning Paradigm=SSL2025.04 | 60.8 | — | |
| Pre-trainedBackbone=CLIP-ViT-B/32, Evaluation Protocol=Zero-shot2026.05 | 60.7 | — | |
| CLIP-B/32Backbone Model=CLIP-B/32, Adaptation Protocol=Original Zero-shot2026.06 | 60.69 | — | |
| CoOpYear=2022, Prompting Strategy=Text Prompt, Learning Paradigm=SSL2025.04 | 58.5 | — | |
| CLIPYear=2021, Prompting Strategy=Text Prompt, Learning Paradigm=SSL2025.04 | 54.5 | — | |
| CLIPYear=2021, Prompting Strategy=Text Prompt, Learning Paradigm=UL2025.04 | 54.5 | — | |
| CLIPYear=2021, Prompting Strategy=Text Prompt, Learning Paradigm=TRZSL2025.04 | 54.5 | — | |
| CLIPYear=2021, Prompting Strategy=Visual Prompt, Learning Paradigm=SSL2025.04 | 54.5 | — | |
| CLIPYear=2021, Prompting Strategy=Visual Prompt, Learning Paradigm=UL2025.04 | 54.5 | — | |
| CLIPYear=2021, Prompting Strategy=Visual Prompt, Learning Paradigm=TRZSL2025.04 | 54.5 | — | |
| BOFTBackbone=ViT-L, m=4, r=42024.05 | — | 74.7 | |
| CAT-MergingBackbone=ViT-B/32, #Params=x1, Merging Strategy=Static Merging2026.05 | — | 80.5 | |
| DiDi-Merging-MBackbone=ViT-B/32, #Params=x1.4, Merging Strategy=Dynamic Merging2026.05 | — | 93.5 | |
| DiDi-Merging-SBackbone=ViT-B/32, #Params=x1.24, Merging Strategy=Dynamic Merging2026.05 | — | 93.2 | |
| DoRABackbone=ViT-L, r=82024.05 | — | 78.32 | |
| DoRABackbone=ViT-L, r=12024.05 | — | 78.02 | |
| EMR-MergingBackbone=ViT-B/32, #Params=x4, Merging Strategy=Dynamic Merging2026.05 | — | 91.9 | |
| FREE-MergingBackbone=ViT-B/32, #Params=x2.16, Merging Strategy=Dynamic Merging2026.05 | — | 93.4 | |
| Full-FTBackbone=ViT-L2024.05 | — | 76.83 | |
| HeadBackbone=ViT-L2024.05 | — | 64.1 |