Safety Evaluation on Advbench
100Safety ScoreAOA
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| AOAModel=Llama-2-13B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Llama-2-13B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Llama-2-13B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| BaseModel=Gemma-1.1-7B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| AOAModel=Gemma-1.1-7B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Gemma-1.1-7B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| BaseModel=Gemma-2-9B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| AOAModel=Gemma-2-9B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| FormalModel=Gemma-2-9B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| SkeletonBackbone=GPT-oss-20B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| AOALLM Model=Gemma-2-27B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| SkeletonLLM Model=Gemma-2-27B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| FormalLLM Model=Gemma-2-27B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| IDGAFLLM Model=Gemma-2-27B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| BaseLLM Model=Qwen2.5-32B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| AOALLM Model=Qwen2.5-32B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| SkeletonLLM Model=Qwen2.5-32B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| IDGAFLLM Model=Qwen2.5-32B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| BaseModel=Gemma-2-2B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| AOAModel=Qwen2-7B2026.01 | 100 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Gemma-1.1-7B2026.01 | 99.81 | — | — | — | — | — | — | — | — | |
| BaseModel=Qwen3-4B-think2026.01 | 99.81 | — | — | — | — | — | — | — | — | |
| AOABackbone=GPT-oss-20B2026.01 | 99.81 | — | — | — | — | — | — | — | — | |
| FormalBackbone=GPT-oss-20B2026.01 | 99.81 | — | — | — | — | — | — | — | — | |
| BaseLLM Model=Gemma-2-27B2026.01 | 99.81 | — | — | — | — | — | — | — | — | |
| FormalLLM Model=Qwen2.5-32B2026.01 | 99.81 | — | — | — | — | — | — | — | — | |
| FormalModel=Llama-2-13B2026.01 | 99.62 | — | — | — | — | — | — | — | — | |
| FormalModel=Gemma-1.1-7B2026.01 | 99.62 | — | — | — | — | — | — | — | — | |
| FormalModel=Qwen3-4B-think2026.01 | 99.62 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Qwen3-4B-think2026.01 | 99.62 | — | — | — | — | — | — | — | — | |
| BaseBackbone=GPT-oss-20B2026.01 | 99.62 | — | — | — | — | — | — | — | — | |
| FormalModel=Gemma-2-2B2026.01 | 99.62 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Llama-2-13B2026.01 | 99.42 | — | — | — | — | — | — | — | — | |
| IDGAFBackbone=GPT-oss-20B2026.01 | 99.42 | — | — | — | — | — | — | — | — | |
| BaseModel=Llama-2-13B2026.01 | 99.23 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Gemma-2-9B2026.01 | 99.23 | — | — | — | — | — | — | — | — | |
| GCGLLM Model=Qwen2.5-32B2026.01 | 99.23 | — | — | — | — | — | — | — | — | |
| BaseModel=Qwen2-7B2026.01 | 99.23 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Gemma-2-9B2026.01 | 99.04 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Gemma-2-2B2026.01 | 99.04 | — | — | — | — | — | — | — | — | |
| AOAModel=Gemma-2-2B2026.01 | 98.85 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Qwen3-4B-think2026.01 | 98.65 | — | — | — | — | — | — | — | — | |
| AOAModel=Qwen3-4B-think2026.01 | 98.46 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionBackbone=GPT-oss-20B2026.01 | 98.46 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Gemma-1.1-7B2026.01 | 98.08 | — | — | — | — | — | — | — | — | |
| AOAModel=Llama-3.2-1B2026.01 | 97.5 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Gemma-2-2B2026.01 | 97.5 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Qwen2-7B2026.01 | 97.5 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Gemma-2-9B2026.01 | 97.12 | — | — | — | — | — | — | — | — | |
| GCGModel=Gemma-1.1-7B2026.01 | 96.73 | — | — | — | — | — | — | — | — | |
| AOALLM Model=Llama-3.3-70B2026.01 | 96.73 | — | — | — | — | — | — | — | — | |
| GCGModel=Gemma-2-9B2026.01 | 96.54 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Llama-3.1-8B2026.01 | 96.54 | — | — | — | — | — | — | — | — | |
| FTModel=Gemma-2-2B2026.01 | 96.54 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionLLM Model=Qwen2.5-32B2026.01 | 96.35 | — | — | — | — | — | — | — | — | |
| GCGModel=Llama-2-13B2026.01 | 96.28 | — | — | — | — | — | — | — | — | |
| FTModel=Gemma-2-9B2026.01 | 96.15 | — | — | — | — | — | — | — | — | |
| BaseLLM Model=Llama-3.3-70B2026.01 | 95.96 | — | — | — | — | — | — | — | — | |
| BaseModel=Llama-3.2-1B2026.01 | 95.77 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Qwen3-4B-think2026.01 | 95.77 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Llama-3.2-1B2026.01 | 95.19 | — | — | — | — | — | — | — | — | |
| GCGLLM Model=Gemma-2-27B2026.01 | 95.19 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Llama-3.2-1B2026.01 | 95 | — | — | — | — | — | — | — | — | |
| FormalLLM Model=Llama-3.3-70B2026.01 | 94.81 | — | — | — | — | — | — | — | — | |
| BaseModel=Llama-3.1-8B2026.01 | 94.42 | — | — | — | — | — | — | — | — | |
| GCGModel=Llama-3.2-1B2026.01 | 93.64 | — | — | — | — | — | — | — | — | |
| FormalModel=Llama-3.1-8B2026.01 | 92.88 | — | — | — | — | — | — | — | — | |
| GCGBackbone=GPT-oss-20B2026.01 | 92.69 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Llama-3.1-8B2026.01 | 92.5 | — | — | — | — | — | — | — | — | |
| GCGLLM Model=Llama-3.3-70B2026.01 | 92.29 | — | — | — | — | — | — | — | — | |
| AOAModel=Llama-3.1-8B2026.01 | 91.73 | — | — | — | — | — | — | — | — | |
| FormalModel=Qwen2-7B2026.01 | 91.73 | — | — | — | — | — | — | — | — | |
| FormalModel=Llama-3.2-1B2026.01 | 91.54 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionLLM Model=Gemma-2-27B2026.01 | 91.54 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Gemma-2-2B2026.01 | 91.54 | — | — | — | — | — | — | — | — | |
| FTModel=Llama-3.2-1B2026.01 | 91.35 | — | — | — | — | — | — | — | — | |
| GCGModel=Qwen3-4B-think2026.01 | 90.37 | — | — | — | — | — | — | — | — | |
| IDGAFLLM Model=Llama-3.3-70B2026.01 | 89.81 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Qwen2-7B2026.01 | 89.42 | — | — | — | — | — | — | — | — | |
| FTModel=Llama-3.1-8B2026.01 | 88.46 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Llama-3.1-8B2026.01 | 86.54 | — | — | — | — | — | — | — | — | |
| IDGAFModel=Qwen2-7B2026.01 | 86.35 | — | — | — | — | — | — | — | — | |
| GCGModel=Llama-3.1-8B2026.01 | 85.74 | — | — | — | — | — | — | — | — | |
| SkeletonLLM Model=Llama-3.3-70B2026.01 | 80.58 | — | — | — | — | — | — | — | — | |
| SkeletonModel=Qwen3-0.6B2026.01 | 79.42 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionLLM Model=Llama-3.3-70B2026.01 | 79.04 | — | — | — | — | — | — | — | — | |
| FTLLM Model=Gemma-2-27B2026.01 | 78.85 | — | — | — | — | — | — | — | — | |
| FTLLM Model=Qwen2.5-32B2026.01 | 76.54 | — | — | — | — | — | — | — | — | |
| FTModel=Qwen2-7B2026.01 | 74.81 | — | — | — | — | — | — | — | — | |
| GCGModel=Qwen2-7B2026.01 | 73.99 | — | — | — | — | — | — | — | — | |
| FTLLM Model=Llama-3.3-70B2026.01 | 67.31 | — | — | — | — | — | — | — | — | |
| FTBackbone=GPT-oss-20B2026.01 | 66.92 | — | — | — | — | — | — | — | — | |
| Refusal SuppressionModel=Llama-3.2-1B2026.01 | 66.15 | — | — | — | — | — | — | — | — | |
| RULLM Model=Qwen2.5-32B2026.01 | 65.77 | — | — | — | — | — | — | — | — | |
| FTModel=Qwen3-0.6B2026.01 | 63.08 | — | — | — | — | — | — | — | — | |
| FormalModel=Qwen3-0.6B2026.01 | 62.31 | — | — | — | — | — | — | — | — | |
| BaseModel=Qwen3-0.6B2026.01 | 58.08 | — | — | — | — | — | — | — | — | |
| FTModel=Llama-2-13B2026.01 | 56.92 | — | — | — | — | — | — | — | — | |
| FTModel=Gemma-1.1-7B2026.01 | 54.42 | — | — | — | — | — | — | — | — | |
| RU (ours)Model=Llama-3.2-1B2026.01 | 53.46 | — | — | — | — | — | — | — | — |