List your company

Agent Safety Auditing on ASSE-Security

97.6AccuracyTRACES
36.86452.63268.484.168May 26, 2026

Evaluation results

MethodLinks
TRACESBackbone=Llama3.1-8B2026.0597.698.296.662.196.7
TRACESBackbone=Qwen3-4B2026.059595.69556.396.9
AgentDoGBackbone=Llama3.1-8B2026.0588.985.790.656.979.3
AgentDoGBackbone=Qwen3-4B2026.0587.888.695.151.976.2
ShieldAgent2026.0582.784.683.647.866.5
AgentAuditor2026.05818386.246.670.7
GPT-5.22026.0573.869.158.547.674.9
Gemini-3-Flash2026.0562.856.748.832.966.4
Llama3.1-8B-Instruct2026.0553.138.525.69.856.9
PolyGuard2026.0551.642.438.245.663.3
Qwen3-4B2026.055066.71007550
LlamaGuard3-8B2026.0547.612.27.37.349
JoySafety2026.0539.232.729.830.655.2