Uncertainty Quantification for Object Classification on ObjectNet (test)
68.7Rejection Accuracy @ 90%REPVLM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 68.7 | 0.998 | |
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 68.7 | 0.947 | |
| REPVLMBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 64.4 | 0.976 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 60.5 | 0.738 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 60.5 | 0.738 | |
| MCDOBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 60.5 | 0.738 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=DataComp2026.01 | 50.1 | -0.792 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=Conceptual Caption2026.01 | 47 | -0.99 | |
| ProbVLMBackbone=SigLIP ViT-B/16, Pre-training Data=LAION2026.01 | 46.1 | -0.998 |