List your company

Safety Evaluation on SIUO (test)

95.67Effective ScoreSAFE RLHF-v
69.025275.942682.8689.7774Feb 28, 2026

Evaluation results

MethodLinks
SAFE RLHF-vBackbone=Qwen2.5-VL-7B2026.0295.6755.9
Pragma-VLBackbone=Qwen2.5-VL-7B2026.0295.2163.47
Beavertails-V helpBackbone=Qwen2.5-VL-7B2026.0295.234.33
Pragma-VLBackbone=Llava-1.5-7B2026.0294.0155.42
SPA-VLBackbone=Qwen2.5-VL-7B2026.0293.3736.74
SFTBackbone=Qwen2.5-VL-7B2026.0293.2949.39
Beavertails-V allBackbone=Qwen2.5-VL-7B2026.0292.5951.23
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.0292.1738.78
DPOBackbone=Qwen2.5-VL-7B2026.0291.6159.03
Llava-1.5-7BBackbone=Llava-1.5-7B2026.0290.4114.37
Beavertails-V harmBackbone=Qwen2.5-VL-7B2026.0289.7659.64
Beavertails-V helpBackbone=Llava-1.5-7B2026.0288.6230.53
SPA-VLBackbone=Llava-1.5-7B2026.0286.8336.53
SAFE RLHF-vBackbone=Llava-1.5-7B2026.0286.4147.53
SFTBackbone=Llava-1.5-7B2026.0286.2247.3
Beavertails-V allBackbone=Llava-1.5-7B2026.0282.7241.35
MM-RLHFBackbone=Qwen2.5-VL-7B2026.0280.2351.8
DPOBackbone=Llava-1.5-7B2026.0278.3144.91
MM-RLHFBackbone=Llava-1.5-7B2026.0273.6537.95
Beavertails-V harmBackbone=Llava-1.5-7B2026.0270.0532.28