Visual-Only QA on VideoHolmes
47.63AccuracyQwen2.5-Omni7B + OutRo
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-Omni7B + OutRoBase Model=Qwen2.5-Omni7B, OutRo=true2026.03 | 47.63 | |
| Qwen2.5-Omni7BBase Model=Qwen2.5-Omni7B, OutRo=false2026.03 | 47.46 | |
| Qwen2.5-Omni3BBase Model=Qwen2.5-Omni3B, OutRo=false2026.03 | 43.54 | |
| Qwen2.5-Omni3B + OutRoBase Model=Qwen2.5-Omni3B, OutRo=true2026.03 | 43.54 | |
| video-SALMONN2+3B + OutRoBase Model=video-SALMONN2+3B, OutRo=true2026.03 | 42.9 | |
| video-SALMONN2+3BBase Model=video-SALMONN2+3B, OutRo=false2026.03 | 42.68 |