Text Retrieval on MS COCO 2014 (test)
23.81Deliberate R@1CLIPSurgery
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CLIPSurgeryBackbone=EVA-CLIP (ViT-B/16)2025.12 | 23.81 | 42.92 | |
| RolloutBackbone=EVA-CLIP (ViT-B/16)2025.12 | 22.68 | 42.38 | |
| Grad-CAMBackbone=EVA-CLIP (ViT-B/16)2025.12 | 21.61 | 40.08 | |
| M2IBBackbone=EVA-CLIP (ViT-B/16)2025.12 | 20.57 | 39.05 | |
| GAMEBackbone=EVA-CLIP (ViT-B/16)2025.12 | 19.82 | 38 | |
| B-cosifiedBackbone=EVA-CLIP (ViT-B/16), Training Data=CC3M2025.12 | 16.32 | 29.75 | |
| ECLIPBackbone=EVA-CLIP (ViT-B/16)2025.12 | 15.5 | 29.33 | |
| MaskCLIPBackbone=EVA-CLIP (ViT-B/16)2025.12 | 15.16 | 29.49 | |
| SVOBackbone=EVA-CLIP (ViT-B/16), Training Data=CC3M2025.12 | 14.01 | 28.21 |