Audio-visual temporal grounding on VUE (TR)
46.1Vision ScoreOmniAgent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OmniAgentSize=7B2026.06 | 46.1 | 36.5 | |
| VidiSize=7B2026.06 | 43.9 | 33.4 | |
| VITALSize=7B2026.06 | 35.3 | — | |
| Gemini-2.0-FlashSize=-2026.06 | 25.4 | 18.4 | |
| GPT-4oSize=-2026.06 | 22 | 11.1 | |
| Gemini-2.5-ProSize=-2026.06 | 21.8 | 12.1 | |
| Qwen2.5-OmniSize=7B2026.06 | 8 | 3.5 |