Image Classification on ImageNet Top V2 (Top-1 accuracy)
83.5Top-1 AccuracyDREAM
Evaluation Results
| Method | Links | |
|---|---|---|
| DREAMEvaluation Protocol=Fine-tuning2026.03 | 83.5 | |
| REPAEvaluation Protocol=Fine-tuning2026.03 | 83.1 | |
| MAREvaluation Protocol=Fine-tuning2026.03 | 81.7 | |
| CLIPEvaluation Protocol=Fine-tuning2026.03 | 81.7 | |
| FLUIDEvaluation Protocol=Fine-tuning2026.03 | 81.6 | |
| DINORANKCLIP (full)Evaluation Protocol=Zero-shot2026.05 | 13.8 | |
| Ours w/o DINOEvaluation Protocol=Zero-shot2026.05 | 12.95 | |
| RANKCLIPEvaluation Protocol=Zero-shot2026.05 | 12.31 | |
| ALIPEvaluation Protocol=Zero-shot2026.05 | 11.43 | |
| CyCLIPEvaluation Protocol=Zero-shot2026.05 | 11.2 | |
| CLIPEvaluation Protocol=Zero-shot2026.05 | 10.76 |