List your company

Trajectory-level safety evaluation on ASSE-Safety (test)

81.1AccuracyAgentDoG-Llama3.1-8B
38.25249.37660.571.624Jan 26, 2026

Evaluation results

MethodLinks
AgentDoG-Llama3.1-8BModel Category=Our Models2026.0181.18087.283.4
AgentDoG-Qwen3-4BModel Category=Our Models2026.0180.485.577.381.2
AgentDoG-Qwen2.5-7BModel Category=Our Models2026.0179.884.377.480.7
ShieldAgentModel Category=Guard Models2026.0179.676.590.582.9
Gemini-3-ProModel Category=Closed-Source2026.0178.585.972.578.6
Qwen3-235B-A22B-Instruct-2507Model Category=Open-Source2026.0177.685.471.177.6
GPT-5.2Model Category=Closed-Source2026.0177.482.474.478.2
Gemini-3-FlashModel Category=Closed-Source2026.0175.989.563.474.2
QwQ-32BModel Category=Open-Source2026.0168.288.747.862.1
Qwen3-4B-Instruct-2507Model Category=Open-Source2026.0160.987.533.148.1
Qwen2.5-7B-InstructModel Category=Open-Source2026.0158.881.331.945.8
LlamaGuard4-12BModel Category=Guard Models2026.0156.379.927.841.2
PolyGuardModel Category=Guard Models2026.0156.362.749.455.2
Llama3.1-8B-InstructModel Category=Open-Source2026.0155.25598.370.6
LlamaGuard3-8BModel Category=Guard Models2026.0154.577.624.737.4
Qwen3-GuardModel Category=Guard Models2026.0148.261.215.825.1
ShieldGemma-27BModel Category=Guard Models2026.0147.265.1712.6
ShieldGemma-9BModel Category=Guard Models2026.0147.2715.510.1
NemoGuardModel Category=Guard Models2026.0143.447.230.336.9
JoySafetyModel Category=Guard Models2026.0139.940.922.629.1