List your company

Safety Evaluation on WildChat unsafe prompts

99.82Not-Unsafe RateHuman-PPO
28.247246.828665.4183.9914Nov 2, 2024Dec 6, 2024Jan 9, 2025Feb 13, 2025Mar 19, 2025Apr 22, 2025May 27, 2025

Evaluation results

MethodLinks
Human-PPO2024.1199.8298.9998.76
Qwen (SFT_DB)LLM=Qwen, SFT Status=SFT on AIDSAFE-generated CoTs and responses2025.0596.5——
RBR-PPO2024.1196.0395.995.19
Qwen (Base)LLM=Qwen, SFT Status=None2025.0595.5——
Mixtral (SFT_DB)LLM=Mixtral, SFT Status=SFT on AIDSAFE-generated CoTs and responses2025.0585.95——
Helpful-PPO2024.1169.3473.785.67
Qwen (SFT_OG)LLM=Qwen, SFT Status=SFT on original response data without CoTs2025.0559.42——
Mixtral (SFT_OG)LLM=Mixtral, SFT Status=SFT on original response data without CoTs2025.0533.5——
Mixtral (Base)LLM=Mixtral, SFT Status=None2025.0531——