List your company

Safety Evaluation on AdvBench N=520

0.2Flip RateLLaMA-3.1-8B
-3.47221.31446.170.886Jun 1, 2026

Evaluation results

MethodLinks
LLaMA-3.1-8BBits=4, Mitigation Strategy=Group-642026.060.275.3
LLaMA-3.1-8BBits=4, Mitigation Strategy=Unprotected2026.060.8—
LLaMA-3.1-8BBits=4, Mitigation Strategy=FP16 L0-12026.061.2-49
LLaMA-3.1-8BBits=4, Mitigation Strategy=Attn top-2, Selected Layers=L2,L182026.061.5-75
Qwen-2.5-7BBits=4, Mitigation Strategy=FP16 L0-12026.061.997.8
Gemma-2-9BBits=2, Mitigation Strategy=Group-642026.062.797
Mistral-7BBits=4, Mitigation Strategy=Group-642026.067.352
Mistral-7BBits=4, Mitigation Strategy=Attn top-2, Selected Layers=L1,L22026.069.836
Mistral-7BBits=4, Mitigation Strategy=FP16 L0-12026.0610.432
Mistral-7BBits=4, Mitigation Strategy=Unprotected2026.0615.2—
Gemma-2-9BBits=2, Mitigation Strategy=FP16 L0-12026.0679.213.9
Gemma-2-9BBits=2, Mitigation Strategy=Attn top-2, Selected Layers=L1,L22026.0680.812.2
Qwen-2.5-7BBits=4, Mitigation Strategy=Group-642026.0688.22.4
Qwen-2.5-7BBits=4, Mitigation Strategy=Attn top-2, Selected Layers=L5,L222026.0688.91.5
Qwen-2.5-7BBits=4, Mitigation Strategy=Unprotected2026.0690.3—
Gemma-2-9BBits=2, Mitigation Strategy=Unprotected2026.0692—