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

Safety Classification on OpenAI Moderation

81.4F1 Score

ShieldGemma 27B

Updated 29d ago

Evaluation Results

Method	Links
ShieldGemma 27B 2026.06		81.4
Label and intent reward 2026.06		80.9
Label reward 2026.06		79.8
Gemma-3-12B 2026.06		79.3
Human-intent on AIMS 2026.06		79.2
GPT-5.4 2026.06		79.1
Claude Sonnet 4.6 2026.06		78.5
GPT-OSS-Safeguard 120B 2026.06		78
GPT-OSS-120B 2026.06		77.5
IF-DPO 2026.06		76.6
LE-DPO 2026.06		76.5
Llama-3.1-8B 2026.06		76.1
Gemma-3-12B 2026.06		76.1
Nemotron Safety 4B 2026.06		74.7
LlamaGuard 4 2026.06		73.6
Llama-3.1-8B 2026.06		72.8
WildGuard 7B 2026.06		72.4
GuardReasoner 8B 2026.06		70.4