List your company

Safety Moderation on WILDJAILBREAK (val)

0.7ASRWILDGUARD
-2.46418.89340.2561.607Jun 26, 2024

Evaluation results

MethodLinks
WILDGUARDCompletion Model=Tulu2+WJ2024.060.71.7
Aegis-Guard-DCompletion Model=Tulu2+WJ2024.060.916.8
WILDGUARDTarget Model=Llama3-8B-Inst2024.061.18
Aegis-Guard-PCompletion Model=Tulu2+WJ2024.061.74
Llama-Guard2Completion Model=Tulu2+WJ2024.061.91.6
MD-JudgeCompletion Model=Tulu2+WJ2024.061.910.1
No FilterCompletion Model=Tulu2+WJ2024.062.31.2
WILDGUARDCompletion Model=Tulu2-dpo2024.062.40.4
Aegis-Guard-DTarget Model=Llama3-8B-Inst2024.06620.8
MD-JudgeTarget Model=Llama3-8B-Inst2024.0610.110.8
Aegis-Guard-PTarget Model=Llama3-8B-Inst2024.0612.210
Aegis-Guard-DCompletion Model=Tulu2-dpo2024.0612.416
Llama-Guard2Target Model=Llama3-8B-Inst2024.0614.28.4
NoneTarget Model=Llama3-8B-Inst2024.0617.47.6
MD-JudgeCompletion Model=Tulu2-dpo2024.0625.74.4
Aegis-Guard-PCompletion Model=Tulu2-dpo2024.0632.73.6
Llama-Guard2Completion Model=Tulu2-dpo2024.0653.10.8
No FilterCompletion Model=Tulu2-dpo2024.0679.80