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

Safety Classification on SafeRLHF

0.94F1 Score

Qwen3Guard

Updated 1mo ago

Evaluation Results

Method	Links
Qwen3Guard 2025.12		0.94
AprielGuard 2025.12		0.93
IBM Granite Guardian 3.2 2025.12		0.92
AprielGuard 2025.12		0.92
IBM Granite Guardian 3.3 2025.12		0.91
Qwen3Guard 2025.12		0.91
gpt-oss-safeguard 2025.12		0.91
IBM Granite Guardian 3.1 2025.12		0.9
IBM Granite Guardian 3.3 2025.12		0.9
Llama Guard 3 2025.12		0.89
NExT-Guard 2026.02		0.888
Llama Guard 2 2025.12		0.88
Llama Guard 4 2025.12		0.88
IBM Granite Guardian 3.2 2025.12		0.88
NExT-Guard* 2026.02		0.833
BeaverDam 7B 2026.02		0.721
Qwen3Guard Gen 8B 2026.02		0.7054
GuardReasoner 8B 2026.02		0.7004
GuardReasoner-Omni 4B 2026.02		0.6844
GuardReasoner-Omni 2B 2026.02		0.6814
GuardReasoner-VL 7B 2026.02		0.6663
GuardReasoner-VL 3B 2026.02		0.6634
MD-Judge 7B 2026.02		0.647
WildGuard 7B 2026.02		0.642
Qwen3Guard-4B-Gen 2026.02		0.637
Qwen3Guard-4B-Stream 2026.02		0.637
Qwen3Guard Gen 8B 2026.02		0.6348
PolyGuard-Qwen 7B 2026.02		0.6344
Qwen3Guard-0.6B-Gen 2026.02		0.634
Qwen3Guard-8B-Gen 2026.02		0.632
WildGuard-7B 2026.02		0.629
Qwen3Guard-8B-Stream 2026.02		0.617
Qwen3Guard-0.6B-Stream 2026.02		0.611
HarmBench LLaMA 13B 2026.02		0.6
SCM-7B 2026.02		0.598
Aegis Guard Defensive 7B 2026.02		0.593
NemoGuard-8B 2026.02		0.565
ShieldGemma 2025.12		0.56
Aegis Guard Permissive 7B 2026.02		0.559
Kelp 2026.02		0.54
HarmBench Mistral 7B 2026.02		0.524
ShieldGemma-27B 2026.02		0.511
ShieldGemma 9B 2026.02		0.4707
LLaMA Guard 3 8B 2026.02		0.4436
LlamaGuard3-8B 2026.02		0.441
LLaMA Guard 4 12B 2026.02		0.4331
ShieldGemma-9B 2026.02		0.431
LlamaGuard4-12B 2026.02		0.415