List your company

Jailbreak Defense on AutoDAN AdvE

9.83ASRCKU
9.139213.802118.46523.1279Jan 6, 2026

Evaluation results

MethodLinks
CKUBase Model Architecture=Llama3-8B-Instruct2026.019.83
Safe UnlearningBase Model Architecture=Llama3-8B-Instruct2026.0110.79
JPUBase Model Architecture=Llama3-8B-Instruct2026.0111.15
Circuit BreakerBase Model Architecture=Llama3-8B-Instruct2026.0111.51
JPUBase Model Architecture=Llama2-7B-Chat2026.0111.59
CKUBase Model Architecture=Llama2-7B-Chat2026.0112.71
Circuit BreakerBase Model Architecture=Llama2-7B-Chat2026.0113.19
Safe UnlearningBase Model Architecture=Llama2-7B-Chat2026.0113.67
RSFTBase Model Architecture=Llama3-8B-Instruct2026.0113.91
EraserBase Model Architecture=Llama3-8B-Instruct2026.0115.11
RSFTBase Model Architecture=Llama2-7B-Chat2026.0116.07
EraserBase Model Architecture=Llama2-7B-Chat2026.0117.27
Base modelBase Model Architecture=Llama3-8B-Instruct2026.0124.46
Base modelBase Model Architecture=Llama2-7B-Chat2026.0127.1