List your company

Jailbreaking on AdvBench 520

96.7ASR (%)STEER
15.16436.33257.578.668Jul 2, 2026

Evaluation results

MethodLinks
STEERTarget Model=Mistral-7B, Iterations=@82026.0796.7
STEERTarget Model=GLM-4-9B, Iterations=@82026.0796.5
STEERTarget Model=Gemma-7B, Iterations=@82026.0796.2
GCGTarget Model=Gemma-7B, Iterations=@82026.0795.8
CSRTTarget Model=Gemma-7B, Iterations=@82026.0788.1
GCGTarget Model=Mistral-7B, Iterations=@82026.0785.6
CSRTTarget Model=Mistral-7B, Iterations=@82026.0784.2
STEERTarget Model=Llama-3-8B, Iterations=@82026.0782.1
STEERTarget Model=DeepSeek-R1, Iterations=@82026.0777.3
STEERTarget Model=Qwen3-8B, Iterations=@82026.0776.5
CSRTTarget Model=GLM-4-9B, Iterations=@82026.0774.4
CSRTTarget Model=Llama-3-8B, Iterations=@82026.0769.6
CSRTTarget Model=Qwen3-8B, Iterations=@82026.0758.1
CSRTTarget Model=DeepSeek-R1, Iterations=@82026.0755.8
GCGTarget Model=DeepSeek-R1, Iterations=@82026.0747.5
GCGTarget Model=Qwen3-8B, Iterations=@82026.0733.8
GCGTarget Model=GLM-4-9B, Iterations=@82026.0720.2
GCGTarget Model=Llama-3-8B, Iterations=@82026.0718.3