Image-Text Retrieval Robustness on DCI-Long
48.32Retrieval Acc (Word Dropout p=0.3)HyFL-CLIP
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| HyFL-CLIP2026.07 | 48.32 | 48.18 | 45.56 | — | 54.14 | 18.93 | 23.63 | |
| HiMo-CLIPimplementation=evaluated using our own implementation2026.07 | 45.7 | 33.97 | 41.81 | — | 50.8 | 16.35 | 20.55 | |
| FineLIPimplementation=evaluated using our own implementation2026.07 | 45.41 | 33.8 | 42.17 | — | 51 | 15.84 | 19.65 | |
| Long-CLIPsource=obtained using checkpoints provided by the original authors2026.07 | 34.84 | 25.45 | 32.23 | — | 38.53 | 15.65 | 18.43 |