Uncertainty Quantification for Sketch Classification on ImageNet-Sketch (test)
77.7Rejection Accuracy (90% Rej)REPVLM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 77.7 | 0.948 | |
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 77.6 | 0.933 | |
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 74.9 | 0.909 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 71.6 | 1 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 71.6 | 1 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 71.6 | 1 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 68.2 | 0.864 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 67.2 | 0.185 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 65.4 | -0.413 |