Image-Text Retrieval on DCI (Text Perturbation Robustness)
61.43Performance (Word Dropout p=0.3)HyFL-CLIP
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| HyFL-CLIP2026.07 | 61.43 | 61.74 | 58.51 | — | 67.69 | 28.73 | 34.56 | |
| HiMo-CLIPimplementation=evaluated using our own implementation2026.07 | 59.16 | 47.05 | 54.93 | — | 64.15 | 25.12 | 30.04 | |
| FineLIPimplementation=evaluated using our own implementation2026.07 | 58.27 | 47.19 | 54.18 | — | 64.33 | 23.79 | 28.78 | |
| Long-CLIPsource=obtained using checkpoints provided by the original authors2026.07 | 49.48 | 38.74 | 45.77 | — | 53.09 | 23.58 | 28.7 |