Multimodal Safety Evaluation on ToolkitBench
2.49Safety ScoreQwen2.5-VL-7B + SFT+DPO+GRPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL-7B + SFT+DPO+GRPOBase Model=Qwen2.5-VL-7B, Stage=SFT+DPO+GRPO2026.03 | 2.49 | 2.5 | 2.45 | |
| Qwen2.5-VL-7B + SFT+GRPOBase Model=Qwen2.5-VL-7B, Stage=SFT+GRPO2026.03 | 2.47 | 2.35 | 2.31 | |
| Qwen2.5-VL-3B + SFT+DPO+GRPOBase Model=Qwen2.5-VL-3B, Stage=SFT+DPO+GRPO2026.03 | 2.45 | 2.21 | 2.23 | |
| Qwen2.5-VL-3B + SFT+GRPOBase Model=Qwen2.5-VL-3B, Stage=SFT+GRPO2026.03 | 2.39 | 2.12 | 2.15 | |
| Qwen2.5-VL-7B + TISBase Model=Qwen2.5-VL-7B, Enhancement=TIS2026.03 | 2.37 | 1.94 | 1.98 | |
| Qwen2.5-VL-7B + SFT+DPOBase Model=Qwen2.5-VL-7B, Stage=SFT+DPO2026.03 | 2.29 | 2.34 | 2.22 | |
| Qwen2.5-VL-7B + SaFeR-VLMBase Model=Qwen2.5-VL-7B, Method=SaFeR-VLM2026.03 | 2.26 | 1.62 | 1.66 | |
| Qwen2.5-VL-7B + SPA-VLBase Model=Qwen2.5-VL-7B, Enhancement=SPA-VL2026.03 | 2.23 | 1.97 | 1.45 | |
| Qwen2.5-VL-3B + SFT+DPOBase Model=Qwen2.5-VL-3B, Stage=SFT+DPO2026.03 | 2.18 | 1.94 | 1.85 | |
| Qwen2.5-VL-7B + SFTBase Model=Qwen2.5-VL-7B, Stage=SFT2026.03 | 2.15 | 2.15 | 2.01 | |
| Qwen2.5-VL-3B + SPA-VLBase Model=Qwen2.5-VL-3B, Enhancement=SPA-VL2026.03 | 2.13 | 1.78 | 1.27 | |
| Qwen2.5-VL-7B + SIABase Model=Qwen2.5-VL-7B, Enhancement=SIA2026.03 | 2.13 | 1.79 | 1.44 | |
| Qwen2.5-VL-3B + SaFeR-VLMBase Model=Qwen2.5-VL-3B, Method=SaFeR-VLM2026.03 | 2.08 | 1.46 | 1.43 | |
| Qwen2.5-VL-3B + SFTBase Model=Qwen2.5-VL-3B, Stage=SFT2026.03 | 2.07 | 1.87 | 1.74 | |
| Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2026.03 | 2.07 | 2.05 | 1.54 | |
| Qwen2.5-VL-7B + ECSOBase Model=Qwen2.5-VL-7B, Enhancement=ECSO2026.03 | 2.04 | 2.02 | 1.52 | |
| Qwen2.5-VL-3B + TISBase Model=Qwen2.5-VL-3B, Enhancement=TIS2026.03 | 1.97 | 1.45 | 1.61 | |
| Qwen2.5-VL-7B + VLGuardBase Model=Qwen2.5-VL-7B, Enhancement=VLGuard2026.03 | 1.77 | 1.19 | 1.11 | |
| Qwen2.5-VL-3B + VLGuardBase Model=Qwen2.5-VL-3B, Enhancement=VLGuard2026.03 | 1.69 | 1.1 | 1.09 | |
| Qwen2.5-VL-3B + SIABase Model=Qwen2.5-VL-3B, Enhancement=SIA2026.03 | 1.62 | 1.46 | 1.01 | |
| Qwen2.5-VL-3B + ECSOBase Model=Qwen2.5-VL-3B, Enhancement=ECSO2026.03 | 1.58 | 1.66 | 1.13 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B2026.03 | 1.57 | 1.74 | 1.18 |