Audio-Visual Instance Editing on AvED-Bench
46.8AVSInstructAV2AV
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| InstructAV2AV2026.05 | 46.8 | 41.8 | 43.6 | |
| AVI-Editinput=ground-truth instance masks and text description2025.12 | 45.6 | 42.4 | 41.2 | |
| AVI-Edit2026.05 | 33 | 31.6 | 31.2 | |
| Oviinput=ground-truth instance masks and text description2025.12 | 31.6 | 31.2 | 32 | |
| VACE-Foleyinput=ground-truth instance masks and text description2025.12 | 19.2 | 21.6 | 22.8 | |
| CoherentAVEdit2026.05 | 18.4 | 23.2 | 23 | |
| AvEDinput=ground-truth instance masks and text description2025.12 | 3.6 | 4.8 | 4 | |
| AvED2026.05 | 1.8 | 3.4 | 2.2 |