List your company
ResearchFormerly WizWand

Response Harmfulness Detection on SafeRLHF

72.1F1 ScoreBeaverDam
7.6224.3641.157.84May 27, 2026May 28, 2026May 29, 2026
Updated 5mo ago

Evaluation results

MethodLinks
BeaverDamModel Size=7B2026.0572.1
COLAGUARDModel Size=8B2026.0570.49
GuardReasonerModel Size=8B2026.0570.04
GuardReasonerCategory=Reasoning-based Guardrails, Model Size=8B2026.0570.04
ConsisGuardCategory=Reasoning-based Guardrails, Model Size=7B2026.0569.98
ConsisGuardCategory=Reasoning-based Guardrails, Model Size=3B2026.0569.7
GuardReasonerModel Size=3B2026.0569.02
GuardReasonerCategory=Reasoning-based Guardrails, Model Size=3B2026.0569.02
COLAGUARDModel Size=3B2026.0568.72
GuardReasonerModel Size=1B2026.0568.39
Claude 3.5+CoTCategory=CoT-Based LLM, Model Size=-2026.0567.54
o1-previewModel Size=Unknown2026.0566.6
GPT-4o+CoTModel Size=Unknown2026.0565.1
o1-pre+CoTCategory=CoT-Based LRM, Model Size=-2026.0565.1
MD-JudgeModel Size=7B2026.0564.7
MD-JudgeCategory=Discriminant-based Guardrails, Model Size=7B2026.0564.7
qwen3-CoTCategory=CoT-Based LRM, Model Size=32B2026.0564.42
GPT-4o+CoTCategory=CoT-Based LLM, Model Size=-2026.0564.34
WildGuardModel Size=7B2026.0564.2
WildGuardCategory=Discriminant-based Guardrails, Model Size=7B2026.0564.2
GPT-4oModel Size=Unknown2026.0564.05
Gemini 1.5+CoTCategory=CoT-Based LLM, Model Size=-2026.0563.73
QWQ+CoTCategory=CoT-Based LRM, Model Size=32B2026.0563.71
QwQ-previewModel Size=32B2026.0562.76
GPT-4+CoTCategory=CoT-Based LLM, Model Size=-2026.0561.25
HarmBench LLaMAModel Size=13B2026.0560
Aegis Guard DefensiveModel Size=7B2026.0559.3
Aegis Guard DefCategory=Discriminant-based Guardrails, Model Size=7B2026.0559.3
Aegis Guard PermissiveModel Size=7B2026.0555.9
Aegis Guard PerCategory=Discriminant-based Guardrails, Model Size=7B2026.0555.9
HarmBench MistralModel Size=7B2026.0552.4
LLaMA Guard 2Model Size=8B2026.0551.6
LLaMA Guard 2Category=Discriminant-based Guardrails, Model Size=8B2026.0551.6
LLaMA GuardModel Size=7B2026.0548.4
ShieldGemmaModel Size=9B2026.0547.07
ShieldGemmaCategory=Discriminant-based Guardrails, Model Size=9B2026.0547.07
LLaMA Guard 3Model Size=8B2026.0544.36
LLaMA Guard 3Category=Discriminant-based Guardrails, Model Size=8B2026.0544.36
ShieldGemmaModel Size=2B2026.0516.92
ShieldGemmaCategory=Discriminant-based Guardrails, Model Size=2B2026.0516.92
ModerationCategory=Discriminant-based Guardrails, Model Size=-2026.0510.1