List your company

Safety Evaluation on WildGuard (Safe)

30.7Refusal RateTarget feature
2.5169.83317.1524.467Jun 25, 2026

Evaluation results

MethodLinks
Target featureModel=Qwen3-4B-it2026.0630.7
Target featureModel=Qwen3-1.7B-it2026.0618.3
Random featureModel=Qwen3-1.7B-it2026.065.3
No SteeringModel=Qwen3-4B-it2026.064.4
No SteeringModel=Qwen3-1.7B-it2026.064.3
Random featureModel=Qwen3-4B-it2026.063.6