Safety Evaluation on OmniBench
42.47AccuracyOmniSteer
Evaluation results
| Method | Links | |
|---|---|---|
| OmniSteerModel=Qwen2.5-Omni-7B2026.02 | 42.47 | |
| VanillaModel=Qwen2.5-Omni-7B2026.02 | 40.43 | |
| VanillaModel=MiniCPM-o-2.62026.02 | 39.06 | |
| OmniGuardModel=MiniCPM-o-2.62026.02 | 38.58 | |
| OmniSteerModel=MiniCPM-o-2.62026.02 | 37.87 | |
| Self-ReminderModel=Qwen2.5-Omni-7B2026.02 | 36.46 | |
| Self-ReminderModel=MiniCPM-o-2.62026.02 | 32.65 | |
| VanillaModel=Baichuan-Omni-1d52026.02 | 31.18 | |
| OmniSteerModel=Baichuan-Omni-1d52026.02 | 29.2 | |
| OmniGuardModel=Baichuan-Omni-1d52026.02 | 28.64 | |
| OmniGuardModel=Qwen2.5-Omni-7B2026.02 | 19.61 | |
| Self-ReminderModel=Baichuan-Omni-1d52026.02 | 17.32 |