Safety Evaluation on BeaverTails Text
99.3Overall ScoreOmniSteer
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OmniSteerBase Model=Qwen2.5-7B-Omni2026.02 | 99.3 | — | — | |
| Self-ReminderBase Model=Qwen2.5-7B-Omni2026.02 | 96.5 | — | — | |
| OmniGuardBase Model=Qwen2.5-7B-Omni2026.02 | 94.8 | — | — | |
| Self-ReminderModel=Baichuan-Omni-1d52026.02 | 90.43 | 98.33 | 79.81 | |
| Self-ReminderModel=MiniCPM-o-2.62026.02 | 88.55 | 98.73 | 74.84 | |
| VanillaBase Model=Qwen2.5-7B-Omni2026.02 | 88.4 | — | — | |
| OmniSteerModel=Qwen2.5-Omni-7B2026.02 | 87.92 | 95.9 | 77.17 | |
| Self-ReminderModel=Qwen2.5-Omni-7B2026.02 | 87.16 | 91.23 | 81.68 | |
| VanillaModel=Baichuan-Omni-1d52026.02 | 85.87 | 95.38 | 73.06 | |
| OmniSteerModel=Baichuan-Omni-1d52026.02 | 85.67 | 96.42 | 71.2 | |
| OmniSteerModel=MiniCPM-o-2.62026.02 | 84.24 | 86.61 | 81.06 | |
| VanillaModel=Qwen2.5-Omni-7B2026.02 | 83.35 | 81.59 | 85.71 | |
| OmniGuardModel=Qwen2.5-Omni-7B2026.02 | 83.25 | 81.77 | 85.25 | |
| VanillaModel=MiniCPM-o-2.62026.02 | 78.95 | 77.26 | 81.21 | |
| OmniGuardModel=MiniCPM-o-2.62026.02 | 78.91 | 77.44 | 80.9 | |
| OmniGuardModel=Baichuan-Omni-1d52026.02 | 64.91 | 99.88 | 17.86 |