Gender Classification on COCO 95% spurious correlation
78.1Average ScoreBAP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BAPBackbone=CLIP, Evaluation Protocol=Linear Probing2026.05 | 78.1 | 75.7 | |
| Data-Matched ControlBackbone=SigLIP2, Evaluation Protocol=Trained Evaluation2026.05 | 78.1 | 76.2 | |
| BAPBackbone=SigLIP2, Evaluation Protocol=Linear Probing2026.05 | 77.6 | 75.8 | |
| BAPBackbone=SigLIP2, Evaluation Protocol=Zero-Shot2026.05 | 77.5 | 71.8 | |
| Data-Matched ControlBackbone=CLIP, Evaluation Protocol=Trained Evaluation2026.05 | 75.2 | 73.8 | |
| DFRBackbone=SigLIP2, Evaluation Protocol=Trained Evaluation2026.05 | 75.2 | 71.4 | |
| DIALBackbone=CLIP, Evaluation Protocol=Zero-Shot2026.05 | 75.1 | 53.8 | |
| BAPBackbone=CLIP, Evaluation Protocol=Zero-Shot2026.05 | 74.2 | 57.1 | |
| DFRBackbone=CLIP, Evaluation Protocol=Trained Evaluation2026.05 | 74.2 | 66.8 | |
| AFRBackbone=CLIP, Evaluation Protocol=Trained Evaluation2026.05 | 73.3 | 64.7 | |
| RoboShotBackbone=CLIP, Evaluation Protocol=Zero-Shot2026.05 | 73.2 | 52.5 | |
| RoboShotBackbone=SigLIP2, Evaluation Protocol=Zero-Shot2026.05 | 72.9 | 67 | |
| DIALBackbone=SigLIP2, Evaluation Protocol=Zero-Shot2026.05 | 72.4 | 65.2 | |
| Native BackboneBackbone=CLIP, Evaluation Protocol=Zero-Shot2026.05 | 71.4 | 48.3 | |
| Native BackboneBackbone=SigLIP2, Evaluation Protocol=Zero-Shot2026.05 | 71.3 | 64.2 | |
| WiSE-FTBackbone=SigLIP2, Evaluation Protocol=Trained Evaluation2026.05 | 65.7 | 53.2 | |
| ERMBackbone=SigLIP2, Evaluation Protocol=Linear Probing & Fine-Tuning2026.05 | 63.3 | 41.2 | |
| WiSE-FTBackbone=CLIP, Evaluation Protocol=Trained Evaluation2026.05 | 63.1 | 57.4 | |
| ERMBackbone=CLIP, Evaluation Protocol=Linear Probing & Fine-Tuning2026.05 | 60.3 | 54.2 | |
| Native BackboneBackbone=SigLIP2, Evaluation Protocol=Linear Probing2026.05 | 58.9 | 32.3 | |
| PruSCBackbone=SigLIP2, Evaluation Protocol=Trained Evaluation2026.05 | 57.5 | 45.9 | |
| Native BackboneBackbone=CLIP, Evaluation Protocol=Linear Probing2026.05 | 57.3 | 23.4 | |
| AFRBackbone=SigLIP2, Evaluation Protocol=Trained Evaluation2026.05 | 56.8 | 68.6 | |
| PruSCBackbone=CLIP, Evaluation Protocol=Trained Evaluation2026.05 | 55.6 | 48.6 |