Image-to-Text Retrieval on CC3M (val)
33.65R@1EfficientNet-B0
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| EfficientNet-B0Distillation=CLIP-KD2026.02 | 33.65 | 58.72 | 68.28 | |
| InherNet2026.02 | 32.17 | 57.17 | 67.82 | |
| ResNet-18Distillation=CLIP-KD2026.02 | 31.03 | 56.41 | 65.82 | |
| ResNet-101Role=Teacher2026.02 | 30.58 | 53.95 | 63.49 | |
| MobileNetV3Distillation=CLIP-KD2026.02 | 29.07 | 53.43 | 63.59 |