Multi-turn MLLM Safety Evaluation on STEER-VLS Multi-turn
65.96Safety ScoreSAFER-STEER
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SAFER-STEERBackbone=Qwen2.5VL-7B, Alignment Strategy=SAFER-STEER2026.03 | 65.96 | 72.95 | |
| SaFeR-VLMBackbone=Qwen2.5VL-7B, Alignment Strategy=SaFeR-VLM2026.03 | 54.78 | 73.33 | |
| SAFER-STEERBackbone=Qwen2.5VL-3B, Alignment Strategy=SAFER-STEER2026.03 | 54.43 | 72.48 | |
| SFTBackbone=Qwen2.5VL-7B, Alignment Strategy=SFT2026.03 | 41.96 | 61.01 | |
| TISBackbone=Qwen2.5VL-7B, Alignment Strategy=TIS2026.03 | 40.47 | 43.7 | |
| SFTBackbone=Qwen2.5VL-3B, Alignment Strategy=SFT2026.03 | 34.33 | 53.73 | |
| SPA-VLBackbone=Qwen2.5VL-7B, Alignment Strategy=SPA-VL2026.03 | 32.56 | 51.74 | |
| SaFeR-VLMBackbone=Qwen2.5VL-3B, Alignment Strategy=SaFeR-VLM2026.03 | 29.91 | 53.08 | |
| MMDPOBackbone=Qwen2.5VL-7B, Alignment Strategy=MMDPO2026.03 | 25.24 | 54.95 | |
| Qwen2.5VL-7BBackbone=Qwen2.5VL-7B, Alignment Strategy=Base2026.03 | 24.56 | 51.18 | |
| TISBackbone=Qwen2.5VL-3B, Alignment Strategy=TIS2026.03 | 24.48 | 31.34 | |
| GRLHF-VBackbone=Qwen2.5VL-7B, Alignment Strategy=GRLHF-V2026.03 | 21.45 | 45.22 | |
| GRLHF-VBackbone=Qwen2.5VL-3B, Alignment Strategy=GRLHF-V2026.03 | 18.55 | 37.1 | |
| SafeRLHF-VBackbone=Qwen2.5VL-7B, Alignment Strategy=SafeRLHF-V2026.03 | 18.37 | 28.86 | |
| SPA-VLBackbone=Qwen2.5VL-3B, Alignment Strategy=SPA-VL2026.03 | 17.94 | 36.18 | |
| MMDPOBackbone=Qwen2.5VL-3B, Alignment Strategy=MMDPO2026.03 | 13.53 | 35.59 | |
| SafeRLHF-VBackbone=Qwen2.5VL-3B, Alignment Strategy=SafeRLHF-V2026.03 | 13.12 | 16.91 | |
| Qwen2.5VL-3BBackbone=Qwen2.5VL-3B, Alignment Strategy=Base2026.03 | 11.76 | 34.12 |