List your company

Safety Evaluation on Refusal Signal Score

7.5ASRMLP
4.117626.948849.7872.6112May 6, 2026

Evaluation results

MethodLinks
MLPBase Model=Llama-3.1-70B, Condition=Cured2026.057.5
MLPBase Model=Qwen2.5-72B, Condition=Cured2026.058.06
MoEBase Model=Llama-3.1-8B, Condition=Cured2026.058.62
MoEBase Model=Gemma-2-27B, Condition=Cured2026.0510.75
MoEBase Model=Gemma-2-9b, Condition=Cured2026.0511.31
MoEBase Model=Mistral-7B-v0.3, Condition=Cured2026.0513.31
MLPBase Model=Mistral-7B-v0.3, Condition=Cured2026.0513.88
MLPBase Model=Qwen2.5-7B, Condition=Cured2026.0514.12
MLPBase Model=Gemma-2-27B, Condition=Cured2026.0516.06
MoEBase Model=Qwen2.5-7B, Condition=Cured2026.0516.38
FMBase Model=Gemma-2-9b, Condition=Cured2026.0520.12
FMBase Model=Gemma-2-27B, Condition=Cured2026.0523.31
MLPBase Model=Llama-3.1-8B, Condition=Cured2026.0524.06
MoEBase Model=Llama-3.1-70B, Condition=Cured2026.0524.5
MLPBase Model=Gemma-2-9b, Condition=Cured2026.0528.75
FMBase Model=Llama-3.1-8B, Condition=Cured2026.0529.12
MoEBase Model=Qwen2.5-72B, Condition=Cured2026.0529.25
CVAEBase Model=Qwen2.5-72B, Condition=Cured2026.0529.75
FMBase Model=Qwen2.5-7B, Condition=Cured2026.0537.31
FMBase Model=Mistral-7B-v0.3, Condition=Cured2026.0545
Safe LoRABase Model=Qwen2.5-72B, Condition=Cured2026.0546.38
CVAEBase Model=Gemma-2-9b, Condition=Cured2026.0546.88
CVAEBase Model=Gemma-2-27B, Condition=Cured2026.0548.19
AEBase Model=Gemma-2-27B, Condition=Cured2026.0549.38
FMBase Model=Llama-3.1-70B, Condition=Cured2026.0555.38
Safe LoRABase Model=Gemma-2-27B, Condition=Cured2026.0556.75
CVAEBase Model=Qwen2.5-7B, Condition=Cured2026.0558.19
CVAEBase Model=Llama-3.1-70B, Condition=Cured2026.0558.56
CVAEBase Model=Mistral-7B-v0.3, Condition=Cured2026.0558.75
AEBase Model=Gemma-2-9b, Condition=Cured2026.0559.5
FMBase Model=Qwen2.5-72B, Condition=Cured2026.0559.5
Safe LoRABase Model=Gemma-2-9b, Condition=Cured2026.0559.62
AEBase Model=Qwen2.5-7B, Condition=Cured2026.0560.75
Safe LoRABase Model=Mistral-7B-v0.3, Condition=Cured2026.0564.44
Safe LoRABase Model=Qwen2.5-7B, Condition=Cured2026.0567.19
AEBase Model=Qwen2.5-72B, Condition=Cured2026.0567.75
Safe LoRABase Model=Llama-3.1-70B, Condition=Cured2026.0567.94
CVAEBase Model=Llama-3.1-8B, Condition=Cured2026.0568.19
AEBase Model=Llama-3.1-8B, Condition=Cured2026.0572.06
AEBase Model=Mistral-7B-v0.3, Condition=Cured2026.0572.31
Safe LoRABase Model=Llama-3.1-8B, Condition=Cured2026.0572.62
AEBase Model=Llama-3.1-70B, Condition=Cured2026.0575.38
Unsafe BaselineBase Model=Qwen2.5-72B, Condition=Unsafe (Finetuned)2026.0580.5
SaLoRABase Model=Qwen2.5-72B, Condition=Cured2026.0581.38
Unsafe BaselineBase Model=Qwen2.5-7B, Condition=Unsafe (Finetuned)2026.0581.69
SaLoRABase Model=Qwen2.5-7B, Condition=Cured2026.0581.94
Unsafe BaselineBase Model=Llama-3.1-70B, Condition=Unsafe (Finetuned)2026.0586.38
SaLoRABase Model=Llama-3.1-70B, Condition=Cured2026.0586.62
SaLoRABase Model=Gemma-2-27B, Condition=Cured2026.0586.62
Unsafe BaselineBase Model=Gemma-2-27B, Condition=Unsafe (Finetuned)2026.0586.94
Unsafe BaselineBase Model=Gemma-2-9b, Condition=Unsafe (Finetuned)2026.0589.88
SaLoRABase Model=Gemma-2-9b, Condition=Cured2026.0590.12
Unsafe BaselineBase Model=Llama-3.1-8B, Condition=Unsafe (Finetuned)2026.0590.44
SaLoRABase Model=Llama-3.1-8B, Condition=Cured2026.0590.88
SaLoRABase Model=Mistral-7B-v0.3, Condition=Cured2026.0592
Unsafe BaselineBase Model=Mistral-7B-v0.3, Condition=Unsafe (Finetuned)2026.0592.06