Cross-modal Retrieval (te → vi → ta) on ObjF AVE real 1.0
19.5AccuracyBrokenBind
Evaluation Results
| Method | Links | |
|---|---|---|
| BrokenBindBackbone=TVL2026.02 | 19.5 | |
| BrokenBindBackbone=TVL2026.02 | 18.4 | |
| TVLBackbone=TVL2026.02 | 14.2 | |
| TVLBackbone=TVL, Fine-tuned=true2026.02 | 13.8 | |
| TVLBackbone=TVL2026.02 | 9.6 | |
| TVLBackbone=TVL, Fine-tuned=true2026.02 | 9 |