List your company

Safety and Helpfulness Evaluation on VLGuard

95.1Safety ScoreGRPO
34.57250.2866681.714Mar 18, 2026Mar 30, 2026Apr 12, 2026Apr 25, 2026May 7, 2026May 20, 2026Jun 2, 2026

Evaluation results

MethodLinks
GRPOBackbone=Qwen3-VL-4B, Alignment Method=GRPO2026.0695.1—51.79
SAFER-STEERBase Model=Qwen2.5VL-7B, Method Category=Dynamic Methods2026.0388.4881.82—
Safe-RLHF-VBackbone=Qwen3-VL-4B, Alignment Method=Safe-RLHF-V2026.0688.2—77.6
GRPOBackbone=Qwen2.5-VL-7B, Alignment Method=GRPO2026.0687.9—55.38
COPSD (Hybrid)Backbone=Qwen3-VL-4B, Alignment Method=COPSD (Hybrid)2026.0686.3—97.13
SFTBase Model=Qwen2.5VL-7B, Method Category=Dynamic Methods2026.0384.277.6—
COPSD (Standard)Backbone=Qwen2.5-VL-7B, Alignment Method=COPSD (Standard)2026.0683.6—94.62
OPDBackbone=Qwen2.5-VL-7B, Alignment Method=OPD2026.0683—96.42
SPA-VLBase Model=Qwen2.5VL-7B, Method Category=Static Methods2026.0382.674.51—
COPSD (Hybrid)Backbone=Qwen2.5-VL-7B, Alignment Method=COPSD (Hybrid)2026.0681.9—95.34
SFTBackbone=Qwen2.5-VL-7B, Alignment Method=SFT2026.0681.8—95.52
COPSD (Standard)Backbone=Qwen3-VL-4B, Alignment Method=COPSD (Standard)2026.0681.1—94.8
TISBase Model=Qwen2.5VL-7B, Method Category=Static Methods2026.0380.176.7—
SaFeR-VLMBase Model=Qwen2.5VL-7B, Method Category=Static Methods2026.0379.8877.88—
SAFER-STEERBase Model=Qwen2.5VL-3B, Method Category=Dynamic Methods2026.0379.6273.25—
MMDPOBase Model=Qwen2.5VL-7B, Method Category=Static Methods2026.0379.5978.41—
SPA-VLBase Model=Qwen2.5VL-3B, Method Category=Static Methods2026.0377.1168.67—
SFTBase Model=Qwen2.5VL-3B, Method Category=Dynamic Methods2026.0376.966.6—
SaFeR-VLMBase Model=Qwen2.5VL-3B, Method Category=Static Methods2026.0375.175.3—
TISBase Model=Qwen2.5VL-3B, Method Category=Static Methods2026.0374.371—
OPDBackbone=Qwen3-VL-4B, Alignment Method=OPD2026.0674.3—94
SFTBackbone=Qwen3-VL-4B, Alignment Method=SFT2026.0673.7—95.52
Qwen2.5VL-7BBase Model=Qwen2.5VL-7B, Method Category=Base2026.0371.972.37—
MMDPOBase Model=Qwen2.5VL-3B, Method Category=Static Methods2026.0371.5466.13—
Qwen3-VL-4BBackbone=Qwen3-VL-4B, Alignment Method=Base2026.0670.4—94.98
Safe-RLHF-VBackbone=Qwen2.5-VL-7B, Alignment Method=Safe-RLHF-V2026.0669.6—64.87
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B, Alignment Method=Base2026.0667.6—97.31
Qwen2.5VL-3BBase Model=Qwen2.5VL-3B, Method Category=Base2026.0366.256.3—
GRLHF-VBase Model=Qwen2.5VL-7B, Method Category=Static Methods2026.0359.3463.72—
SafeRLHF-VBase Model=Qwen2.5VL-7B, Method Category=Static Methods2026.0352.846.2—
SafeRLHF-VBase Model=Qwen2.5VL-3B, Method Category=Static Methods2026.0348.239—
GRLHF-VBase Model=Qwen2.5VL-3B, Method Category=Static Methods2026.0336.937.8—