Cross-modal Retrieval on InstVL img 10K
44.05T2V Recall@1InstAP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| InstAP2026.04 | 44.05 | 45.76 | |
| UMT-L (InstVL; g+i)training_corpus=all InstVL captions treated as global2026.04 | 34.83 | 35.15 | |
| SigLIP2026.04 | 29.76 | 37.83 | |
| OpenCLIP2026.04 | 29.21 | 37.76 | |
| UMT-L (InstVL; g)training_corpus=InstVL global captions2026.04 | 22.87 | 30.37 | |
| VideoPrism2026.04 | 22.75 | 29.51 | |
| UMT-L2026.04 | 21.34 | 23.08 | |
| ViCLIP2026.04 | 19.46 | 20.02 | |
| CLIP4Clip2026.04 | 18.68 | 28.19 | |
| CLIP-ViP2026.04 | 14.38 | 21.85 | |
| MCQ2026.04 | 9.63 | 11.13 | |
| Coca2026.04 | 7.36 | 13.33 |