List your company

Safety evaluation on Sorry-Bench

99.09Safety ScoreIDGAF
10.22233.293556.36579.4365Jan 27, 2026

Evaluation results

MethodLinks
IDGAFModel=Llama-2-13B2026.0199.09
AOAModel=Llama-2-13B2026.0198.41
SkeletonModel=Llama-2-13B2026.0197.5
IDGAFModel=Gemma-1.1-7B2026.0196.14
FormalModel=Qwen3-4B-think2026.0193.86
IDGAFModel=Qwen3-4B-think2026.0193.64
AOAModel=Qwen3-4B-think2026.0193.18
SkeletonModel=Qwen3-4B-think2026.0193.18
AOAModel=Llama-3.1-8B2026.0192.27
BaseModel=Qwen3-4B-think2026.0190.23
SkeletonModel=Gemma-1.1-7B2026.0189.32
BaseModel=Gemma-2-9B2026.0189.09
BaseModel=Llama-3.2-1B2026.0188.86
AOAModel=Gemma-2-9B2026.0188.64
BaseModel=Llama-2-13B2026.0188.41
GCGModel=Qwen3-4B-think2026.0188.38
AOAModel=Gemma-1.1-7B2026.0187.73
SkeletonBackbone=GPT-oss-20B2026.0187.73
FormalModel=Gemma-2-9B2026.0187.05
SkeletonModel=Gemma-2-9B2026.0186.36
Refusal SuppressionModel=Gemma-2-9B2026.0186.36
IDGAFModel=Llama-3.1-8B2026.0185.91
BaseBackbone=GPT-oss-20B2026.0185.45
AOABackbone=GPT-oss-20B2026.0185.45
FormalModel=Llama-2-13B2026.0184.77
GCGModel=Llama-2-13B2026.0184.55
IDGAFModel=Gemma-2-9B2026.0184.09
BaseModel=Gemma-1.1-7B2026.0183.18
GCGModel=Gemma-2-9B2026.0183.18
IDGAFBackbone=GPT-oss-20B2026.0182.73
AOAModel=Llama-3.2-1B2026.0182.05
FormalModel=Gemma-1.1-7B2026.0181.36
BaseModel=Gemma-2-2B2026.0181.36
FormalBackbone=GPT-oss-20B2026.0180.45
FormalModel=Llama-3.1-8B2026.0179.09
BaseModel=Llama-3.1-8B2026.0178.86
GCGBackbone=GPT-oss-20B2026.0178.41
GCGModel=Llama-3.2-1B2026.0177.5
FormalModel=Gemma-2-2B2026.0177.27
GCGModel=Gemma-1.1-7B2026.0176.14
Refusal SuppressionModel=Llama-2-13B2026.0175.23
AOAModel=Gemma-2-2B2026.0175.23
SkeletonModel=Qwen3-0.6B2026.0171.82
IDGAFModel=Llama-3.2-1B2026.0171.82
AOAModel=Qwen2-7B2026.0171.82
IDGAFModel=Gemma-2-2B2026.0170
FTModel=Gemma-2-2B2026.0166.82
FTModel=Gemma-2-9B2026.0166.59
SkeletonModel=Llama-3.2-1B2026.0166.36
Refusal SuppressionModel=Gemma-1.1-7B2026.0165
SkeletonModel=Llama-3.1-8B2026.0163.41
SkeletonModel=Gemma-2-2B2026.0159.77
SkeletonModel=Qwen2-7B2026.0159.77
GCGModel=Llama-3.1-8B2026.0157.27
Refusal SuppressionModel=Qwen3-4B-think2026.0155.91
Refusal SuppressionBackbone=GPT-oss-20B2026.0155.23
FormalModel=Llama-3.2-1B2026.0154.55
BaseModel=Qwen2-7B2026.0153.64
RUModel=Qwen2-7B2026.0147.73
Refusal SuppressionModel=Qwen3-0.6B2026.0144.77
Refusal SuppressionModel=Llama-3.1-8B2026.0142.73
RU (ours)Model=Qwen3-4B-think2026.0141.82
FTModel=Llama-3.2-1B2026.0141.36
Refusal SuppressionModel=Gemma-2-2B2026.0141.14
RUModel=Qwen3-0.6B2026.0140.23
AOAModel=Qwen3-0.6B2026.0138.86
FormalModel=Qwen2-7B2026.0138.18
GCGModel=Qwen2-7B2026.0136.45
IDGAFModel=Qwen3-0.6B2026.0136.36
FTModel=Qwen2-7B2026.0135.68
FTModel=Llama-3.1-8B2026.0134.77
GCGModel=Gemma-2-2B2026.0133.18
RUModel=Gemma-1.1-7B2026.0131.59
FTModel=Qwen3-0.6B2026.0131.14
GCGModel=Qwen3-0.6B2026.0130.3
FTModel=Qwen3-4B-think2026.0130
IDGAFModel=Qwen2-7B2026.0127.27
Refusal SuppressionModel=Qwen2-7B2026.0127.05
RU (ours)Model=Gemma-2-9B2026.0124.09
RU (Refusal Unlearning)Backbone=GPT-oss-20B2026.0124.09
RU (ours)Model=Llama-3.2-1B2026.0123.64
FTModel=Llama-2-13B2026.0122.95
FormalModel=Qwen3-0.6B2026.0122.95
BaseModel=Qwen3-0.6B2026.0122.5
FTBackbone=GPT-oss-20B2026.0122.05
RUModel=Gemma-2-2B2026.0121.14
FTModel=Gemma-1.1-7B2026.0118.86
RUModel=Llama-2-13B2026.0114.55
Refusal SuppressionModel=Llama-3.2-1B2026.0114.55
RUModel=Llama-3.1-8B2026.0113.64