Multimodal Safety Evaluation on BeaverTails-V
2.9Safety ScoreQwen2.5-VL-7B + TIS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL-7B + TISBase Model=Qwen2.5-VL-7B, Enhancement=TIS2026.03 | 2.9 | 2.77 | 2.88 | |
| Qwen2.5-VL-3B + SFT+GRPOBase Model=Qwen2.5-VL-3B, Stage=SFT+GRPO2026.03 | 2.85 | 2.75 | 2.89 | |
| Qwen2.5-VL-3B + SFT+DPO+GRPOBase Model=Qwen2.5-VL-3B, Stage=SFT+DPO+GRPO2026.03 | 2.84 | 2.78 | 2.87 | |
| Qwen2.5-VL-7B + SFT+GRPOBase Model=Qwen2.5-VL-7B, Stage=SFT+GRPO2026.03 | 2.84 | 2.84 | 2.91 | |
| Qwen2.5-VL-7B + SaFeR-VLMBase Model=Qwen2.5-VL-7B, Method=SaFeR-VLM2026.03 | 2.83 | 2.33 | 2.54 | |
| Qwen2.5-VL-7B + SPA-VLBase Model=Qwen2.5-VL-7B, Enhancement=SPA-VL2026.03 | 2.82 | 2.6 | 2.27 | |
| Qwen2.5-VL-7B + SFT+DPO+GRPOBase Model=Qwen2.5-VL-7B, Stage=SFT+DPO+GRPO2026.03 | 2.82 | 2.88 | 2.93 | |
| Qwen2.5-VL-3B + TISBase Model=Qwen2.5-VL-3B, Enhancement=TIS2026.03 | 2.75 | 2.37 | 2.69 | |
| Qwen2.5-VL-3B + SPA-VLBase Model=Qwen2.5-VL-3B, Enhancement=SPA-VL2026.03 | 2.74 | 2.59 | 2.23 | |
| Qwen2.5-VL-3B + SaFeR-VLMBase Model=Qwen2.5-VL-3B, Method=SaFeR-VLM2026.03 | 2.72 | 2.27 | 2.28 | |
| Qwen2.5-VL-7B + SFT+DPOBase Model=Qwen2.5-VL-7B, Stage=SFT+DPO2026.03 | 2.72 | 2.81 | 2.91 | |
| Qwen2.5-VL-7B + VLGuardBase Model=Qwen2.5-VL-7B, Enhancement=VLGuard2026.03 | 2.64 | 1.73 | 1.9 | |
| Qwen2.5-VL-7B + SFTBase Model=Qwen2.5-VL-7B, Stage=SFT2026.03 | 2.62 | 2.66 | 2.86 | |
| Qwen2.5-VL-3B + VLGuardBase Model=Qwen2.5-VL-3B, Enhancement=VLGuard2026.03 | 2.61 | 1.69 | 1.86 | |
| Qwen2.5-VL-3B + SFT+DPOBase Model=Qwen2.5-VL-3B, Stage=SFT+DPO2026.03 | 2.58 | 2.66 | 2.77 | |
| Qwen2.5-VL-3B + SFTBase Model=Qwen2.5-VL-3B, Stage=SFT2026.03 | 2.57 | 2.61 | 2.77 | |
| Qwen2.5-VL-7B + SIABase Model=Qwen2.5-VL-7B, Enhancement=SIA2026.03 | 2.5 | 2.33 | 2.17 | |
| Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2026.03 | 1.67 | 2.7 | 2.14 | |
| Qwen2.5-VL-7B + ECSOBase Model=Qwen2.5-VL-7B, Enhancement=ECSO2026.03 | 1.58 | 2.67 | 2.15 | |
| Qwen2.5-VL-3B + SIABase Model=Qwen2.5-VL-3B, Enhancement=SIA2026.03 | 0.82 | 2.01 | 1.39 | |
| Qwen2.5-VL-3B + ECSOBase Model=Qwen2.5-VL-3B, Enhancement=ECSO2026.03 | 0.73 | 2.41 | 1.66 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B2026.03 | 0.59 | 2.56 | 1.68 |