List your company

Full-response Safety Guardrail Classification on Safe-RLHF (test)

93.2F1 ScoreQwen3Guard-8B
86.02487.88789.7591.613Jun 1, 2026

Evaluation results

MethodLinks
Qwen3Guard-8B2026.0693.2
SentGuardbackbone=Qwen3-4B-Instruct-25072026.0692.5
WildGuard-7B2026.0692.3
GPT-5.5protocol=zero-shot2026.0690.7
LlamaGuard3-8B2026.0688.7
Gemini-3.5-Flashprotocol=zero-shot2026.0686.9
LlamaGuard4-12B2026.0686.3