Image-Text Retrieval Robustness Evaluation on DOCCI
72.92Word Dropout (p=0.3)FineLIP
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| FineLIPimplementation=evaluated using our own implementation2026.07 | 72.92 | 72.84 | 61.54 | — | 72.93 | 26.35 | 33.24 | |
| HiMo-CLIPimplementation=evaluated using our own implementation2026.07 | 71.5 | 71.46 | 60.9 | — | 71.65 | 26.84 | 33.81 | |
| HyFL-CLIP2026.07 | 68.18 | 68.48 | 65.13 | — | 75.53 | 30.93 | 37.2 | |
| Long-CLIPsource=obtained using checkpoints provided by the original authors2026.07 | 59.56 | 59.66 | 51.03 | — | 59.72 | 25.41 | 30.17 |