Zero-shot Retrieval (T+V → A) on AudioCaps
95.2Recall@1PEAVL
Evaluation Results
| Method | Links | |
|---|---|---|
| PEAVLJoint Query Strategy=Native, Model Scale=Large, Training Domain=In-domain2025.12 | 95.2 | |
| PEAVL-OODJoint Query Strategy=Native, Model Scale=Large, Training Domain=Out-of-domain2025.12 | 93.3 | |
| PEAVBJoint Query Strategy=Native, Model Scale=Base, Training Domain=In-domain2025.12 | 93.2 | |
| PEAVB-OODJoint Query Strategy=Native, Model Scale=Base, Training Domain=Out-of-domain2025.12 | 92.6 | |
| PEAVSJoint Query Strategy=Native, Model Scale=Small, Training Domain=In-domain2025.12 | 90.3 | |
| PEAVS-OODJoint Query Strategy=Native, Model Scale=Small, Training Domain=Out-of-domain2025.12 | 89.1 | |
| PEAVLJoint Query Strategy=Approximated, Model Scale=Large, Training Domain=In-domain2025.12 | 88.3 | |
| PEAVL-OODJoint Query Strategy=Approximated, Model Scale=Large, Training Domain=Out-of-domain2025.12 | 86.1 | |
| PEAVBJoint Query Strategy=Approximated, Model Scale=Base, Training Domain=In-domain2025.12 | 83.7 | |
| PEAVB-OODJoint Query Strategy=Approximated, Model Scale=Base, Training Domain=Out-of-domain2025.12 | 82.2 | |
| PEAVSJoint Query Strategy=Approximated, Model Scale=Small, Training Domain=In-domain2025.12 | 77.4 | |
| PEAVS-OODJoint Query Strategy=Approximated, Model Scale=Small, Training Domain=Out-of-domain2025.12 | 75.3 | |
| ImageBindJoint Query Strategy=Approximated2025.12 | 51.3 | |
| LanguageBindJoint Query Strategy=Approximated2025.12 | 19.7 |