Binary Classification on CounterAnimal Pair 1: Brambling vs. Bulbul
93.2Average AccuracyBAP (LP)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BAP (LP)Backbone=SigLIP2, Evaluation Protocol=Trained2026.05 | 93.2 | 88.3 | |
| Native Backbone (LP)Backbone=SigLIP2, Evaluation Protocol=Trained2026.05 | 92.4 | 71.8 | |
| BAP (LP)Backbone=CLIP, Evaluation Protocol=Trained2026.05 | 92 | 86.4 | |
| Native Backbone (LP)Backbone=CLIP, Evaluation Protocol=Trained2026.05 | 91.7 | 74.2 | |
| Data-Matched ControlBackbone=CLIP, Evaluation Protocol=Trained2026.05 | 86.8 | 70.4 | |
| Data-Matched ControlBackbone=SigLIP2, Evaluation Protocol=Trained2026.05 | 85.2 | 73.6 | |
| BAP (ZS)Backbone=CLIP, Evaluation Protocol=Zero-shot2026.05 | 83.5 | 64.9 | |
| BAP (ZS)Backbone=SigLIP2, Evaluation Protocol=Zero-shot2026.05 | 81.9 | 56.7 | |
| DIALBackbone=CLIP, Evaluation Protocol=Zero-shot2026.05 | 79.2 | 61.6 | |
| Native Backbone (ZS)Backbone=CLIP, Evaluation Protocol=Zero-shot2026.05 | 78.5 | 56.3 | |
| DIALBackbone=SigLIP2, Evaluation Protocol=Zero-shot2026.05 | 74.6 | 42 | |
| ERM (LP-FT)Backbone=SigLIP2, Evaluation Protocol=Trained2026.05 | 73.1 | 31.9 | |
| Native Backbone (ZS)Backbone=SigLIP2, Evaluation Protocol=Zero-shot2026.05 | 70.4 | 36.3 | |
| ERM (LP-FT)Backbone=CLIP, Evaluation Protocol=Trained2026.05 | 68.6 | 23.2 | |
| PRuSCBackbone=SigLIP2, Evaluation Protocol=Trained2026.05 | 57.9 | 41.3 | |
| PRuSCBackbone=CLIP, Evaluation Protocol=Trained2026.05 | 52.1 | 34.6 |