List your company

Multi-turn MLLM Safety Evaluation on STEER Average Multi-turn

64.89Safety ScoreSAFER-STEER
6.327621.531336.73551.9387Mar 18, 2026

Evaluation results

MethodLinks
SAFER-STEERBackbone=Qwen2.5VL-7B, Alignment Strategy=SAFER-STEER2026.0364.8972.35
SAFER-STEERBackbone=Qwen2.5VL-3B, Alignment Strategy=SAFER-STEER2026.0355.5870.27
SaFeR-VLMBackbone=Qwen2.5VL-7B, Alignment Strategy=SaFeR-VLM2026.0355.1963.75
SFTBackbone=Qwen2.5VL-7B, Alignment Strategy=SFT2026.0350.3166.86
SFTBackbone=Qwen2.5VL-3B, Alignment Strategy=SFT2026.0341.6157.57
SPA-VLBackbone=Qwen2.5VL-7B, Alignment Strategy=SPA-VL2026.0337.2852.52
TISBackbone=Qwen2.5VL-7B, Alignment Strategy=TIS2026.0336.6833.16
SaFeR-VLMBackbone=Qwen2.5VL-3B, Alignment Strategy=SaFeR-VLM2026.0332.3846.71
TISBackbone=Qwen2.5VL-3B, Alignment Strategy=TIS2026.0326.9427.03
MMDPOBackbone=Qwen2.5VL-7B, Alignment Strategy=MMDPO2026.0326.4751.21
Qwen2.5VL-7BBackbone=Qwen2.5VL-7B, Alignment Strategy=Base2026.0324.6646.48
SPA-VLBackbone=Qwen2.5VL-3B, Alignment Strategy=SPA-VL2026.0323.8136.68
GRLHF-VBackbone=Qwen2.5VL-7B, Alignment Strategy=GRLHF-V2026.0319.5338.96
SafeRLHF-VBackbone=Qwen2.5VL-7B, Alignment Strategy=SafeRLHF-V2026.0317.724.86
MMDPOBackbone=Qwen2.5VL-3B, Alignment Strategy=MMDPO2026.0313.127.16
GRLHF-VBackbone=Qwen2.5VL-3B, Alignment Strategy=GRLHF-V2026.0312.9328.41
Qwen2.5VL-3BBackbone=Qwen2.5VL-3B, Alignment Strategy=Base2026.0312.5527.13
SafeRLHF-VBackbone=Qwen2.5VL-3B, Alignment Strategy=SafeRLHF-V2026.038.5813.04