List your company

Safety Classification on WildGuard (test)

88.8F1 ScoreWildGuard 7B
49.69659.8487080.152Sep 30, 2025Nov 13, 2025Dec 28, 2025Feb 11, 2026Mar 27, 2026May 11, 2026Jun 25, 2026

Evaluation results

MethodLinks
WildGuard 7BOptimization Strategy=Dedicated Safety Guard2026.0688.8
GuardReasoner 8BOptimization Strategy=Dedicated Safety Guard2026.0688.8
TPCParams=1,381,8892025.0988.5
GPT-OSS-120BOptimization Strategy=Zero-shot2026.0688.4
MLP probeParams=1,382,1472025.0988.2
GPT-5.4Optimization Strategy=Zero-shot2026.0688
Bilinear probeParams=344,1292025.0987.82
EE-MLPParams=1,381,9182025.0987.76
Human-intent on AIMSOptimization Strategy=Reasoning Distillation, Teacher Model=GPT-OSS-120B, Student Model=Gemma-3-12B2026.0687.6
GPT-OSS-Safeguard 120BOptimization Strategy=Dedicated Safety Guard2026.0687.1
Label rewardOptimization Strategy=GRPO2026.0687.1
Linear probeParams=5,3772025.0986.86
gpt-4o-miniParams=≈ 8,000,000,000*2025.0986.63
Label and intent rewardOptimization Strategy=GRPO2026.0686.3
Gemma-3-12BOptimization Strategy=SFT, Supervision Source=Annotated Intents2026.0685.7
Llama-3.1-8BOptimization Strategy=SFT, Supervision Source=Annotated Intents2026.0685.6
LE-DPOOptimization Strategy=DPO2026.0685.6
Gemma-3-12BOptimization Strategy=Zero-shot2026.0685.3
Nemotron Safety 4BOptimization Strategy=Dedicated Safety Guard2026.0685.2
IF-DPOOptimization Strategy=DPO2026.0685.1
Claude Sonnet 4.6Optimization Strategy=Zero-shot2026.0683.8
claude-3-haikuParams=≈ 8,000,000,000*2025.0983.24
o3-miniParams=unknown2025.0982.03
SentGuardbackbone=Qwen3-4B-Instruct-25072026.0680
Qwen3Guard-8B2026.0677.2
Llama-Guard-3-8BParams=8,030,261,2482025.0976.85
Llama-3.1-8BOptimization Strategy=Zero-shot2026.0676.2
WildGuard-7B2026.0675.3
GPT-5.5protocol=zero-shot2026.0674.4
LlamaGuard 4Optimization Strategy=Dedicated Safety Guard2026.0673.8
LlamaGuard3-8B2026.0670.7
shieldgemma-2bThreshold=optimal, Params=2,614,341,8882025.0969.9
LlamaGuard4-12B2026.0666.7
Gemini-3.5-Flashprotocol=zero-shot2026.0659.4
ShieldGemma 27BOptimization Strategy=Dedicated Safety Guard2026.0651.2