Uncertainty Quantification for Fine-grained Image Recognition on Food101 (test)
0.93Accuracy @ 90% RejectionREPVLM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 0.93 | 0.755 | |
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 0.928 | 0.907 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 0.919 | 0.648 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 0.919 | 0.648 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 0.919 | 0.648 | |
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 0.917 | 0.855 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 0.899 | -0.376 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 0.894 | -0.959 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 0.888 | -0.985 |