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

Toxicity Detection on ToxicChat

1F1 Score

SafeGPT

Updated 2d ago

Evaluation Results

Method
SafeGPT 2026.01	1	1	1	0	-
ReflectiveProposer 2026.01	0.907	0.916	0.905	-	-
Unoptimized 2026.01	0.849	0.869	0.844	-	-
MIPRO 2026.01	0.796	0.844	0.791	-	-
ConsisGuard 2026.05	0.7905	-	-	-	-
GuardReasoner 2026.05	0.7879	-	-	-	-
GuardReasoner 2026.05	0.7879	-	-	-	-
GuardReasoner 2026.05	0.782	-	-	-	-
GuardReasoner 2026.05	0.782	-	-	-	-
ConsisGuard 2026.05	0.7726	-	-	-	-
COLAGUARD 2026.05	0.7527	-	-	-	-
COLAGUARD 2026.05	0.7526	-	-	-	-
MIPRO 2026.01	0.75	0.819	0.749	-	-
GPT-4o+CoT 2026.05	0.7452	-	-	-	-
GPT-4o+CoT 2026.05	0.7343	-	-	-	-
Aegis Guard Permissive 2026.05	0.73	-	-	-	-
Aegis Guard Per 2026.05	0.73	-	-	-	-
GuardReasoner 2026.05	0.7243	-	-	-	-
WildGuard 2026.05	0.708	-	-	-	-
WildGuard 2026.05	0.708	-	-	-	-
Aegis Guard Defensive 2026.05	0.7	-	-	-	-
Aegis Guard Def 2026.05	0.7	-	-	-	-
qwen3-CoT 2026.05	0.6962	-	-	-	-
ShieldGemma 2026.05	0.6792	-	-	-	-
ShieldGemma 2026.05	0.6792	-	-	-	-
Gemini 1.5+CoT 2026.05	0.666	-	-	-	-
QWQ+CoT 2026.05	0.6653	-	-	-	-
GPT-4o 2026.05	0.6446	-	-	-	-
GPT-4+CoT 2026.05	0.6396	-	-	-	-
o1-pre+CoT 2026.05	0.637	-	-	-	-
LLaMA Guard 2026.05	0.616	-	-	-	-
o1-preview 2026.05	0.5769	-	-	-	-
Claude 3.5+CoT 2026.05	0.5458	-	-	-	-
LLaMA Guard 3 2026.05	0.5312	-	-	-	-
LLaMA Guard 3 2026.05	0.5312	-	-	-	-
LLaMA Guard 2 2026.05	0.471	-	-	-	-
LLaMA Guard 2 2026.05	0.471	-	-	-	-
QwQ-preview 2026.05	0.3481	-	-	-	-
Moderation 2026.05	0.254	-	-	-	-
ShieldGemma 2026.05	0.0691	-	-	-	-
ShieldGemma 2026.05	0.0691	-	-	-	-
Regex-Only 2026.01	0	0	0	0	-
Simple NER 2026.01	0	0	0	0	-
Keyword 2026.01	0	0	0	0	-
Hybrid 2026.01	0	0	0	0	-
SIREN 2026.04	-	-	-	-	0.816
Qwen3Guard 2026.04	-	-	-	-	0.82
SIREN 2026.04	-	-	-	-	0.8
LlamaGuard3 2026.04	-	-	-	-	0.633
SIREN 2026.04	-	-	-	-	0.835
Qwen3Guard 2026.04	-	-	-	-	0.849
SIREN 2026.04	-	-	-	-	0.831
LlamaGuard3 2026.04	-	-	-	-	0.722