Share your thoughts, 1 month free Claude Pro on usSee more

Safety Guardrailing on HumanEval

0False Positive Rate

Perplexity Filter

Updated 16d ago

Evaluation Results

Method	Links
Perplexity Filter 2026.04		0
GradSafe 2026.04		0
Llama Guard 2026.04		0
Llama Guard 2026.04		0
SelfDefend 2026.04		0
WildGuard 2026.04		0
WildGuard 2026.04		0
GuardReasoner 2026.04		0
GuardReasoner 2026.04		0
SelfGrader 2026.04		0
Perplexity Filter 2026.04		0
GradientCuff 2026.04		0
Llama Guard 2026.04		0
Llama Guard 2026.04		0
SelfDefend 2026.04		0
SelfGrader 2026.04		0
Perplexity Filter 2026.04		0
GradientCuff 2026.04		0
Llama Guard (Pre) 2026.04		0
Llama Guard (Post) 2026.04		0
SelfDefend (Intent) 2026.04		0
WildGuard (Pre) 2026.04		0
WildGuard (Post) 2026.04		0
SelfGrader 2026.04		0
Prompt Guard 2026.04		0.6
Prompt Guard 2026.04		0.6
Prompt Guard 2026.04		0.6
SelfDefend 2026.04		1.21
SelfDefend 2026.04		1.21
SelfDefend (Direct) 2026.04		1.8
GradientCuff 2026.04		51.21
Token Highlighter 2026.04		100