Refusal Rate Evaluation on OK (test)
0.7Refusal RateSelf-CD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Self-CDModel=Mistral-7B2024.01 | 0.7 | — | |
| NoSystemModel=Mistral-7B2024.01 | 1.7 | — | |
| Self-CDModel=Beaver2024.01 | 2 | — | |
| Self-CDModel=Vicuna-7B2024.01 | 2.7 | — | |
| Self-CDModel=InternLM-7B2024.01 | 3.3 | — | |
| CoT(zero)Model=Mistral-7B2024.01 | 3.6 | — | |
| PromptModel=Vicuna-7B2024.01 | 4.7 | — | |
| Self-CDModel=Vicuna-13B2024.01 | 4.7 | — | |
| PromptModel=Beaver2024.01 | 5 | — | |
| ICLModel=Mistral-7B2024.01 | 5 | — | |
| SelfCDBase Model=Qwen32026.04 | 5 | — | |
| AdaCDBase Model=Qwen32026.04 | 5 | — | |
| CoT(One)Model=Mistral-7B2024.01 | 5.3 | — | |
| PromptModel=Mistral-7B2024.01 | 6 | — | |
| CoT(zero)Model=Vicuna-13B2024.01 | 6.3 | — | |
| SurgicalBase Model=Qwen32026.04 | 6.67 | — | |
| NoSystemModel=Beaver2024.01 | 6.7 | — | |
| CoT(zero)Model=Beaver2024.01 | 6.7 | — | |
| ICLModel=Beaver2024.01 | 7 | — | |
| CoT(zero)Model=Vicuna-7B2024.01 | 7 | — | |
| PromptModel=Vicuna-13B2024.01 | 7.3 | — | |
| SystemModel=Mistral-7B2024.01 | 7.3 | — | |
| ICLModel=Vicuna-13B2024.01 | 8.3 | — | |
| PromptBase Model=Qwen32026.04 | 8.33 | — | |
| SSDBase Model=Qwen32026.04 | 8.67 | — | |
| Self-CDModel=LLaMA2-70B2024.01 | 8.7 | — | |
| DefaultBase Model=Qwen32026.04 | 9.67 | — | |
| ICLModel=Vicuna-7B2024.01 | 10.3 | — | |
| ICLModel=InternLM-7B2024.01 | 10.6 | — | |
| NoSystemModel=Vicuna-7B2024.01 | 11.7 | — | |
| NoSystemModel=Vicuna-13B2024.01 | 12.3 | — | |
| CoT(One)Model=Vicuna-7B2024.01 | 13 | — | |
| CoT(One)Model=Beaver2024.01 | 13.3 | — | |
| SystemModel=Beaver2024.01 | 13.6 | — | |
| CoT(One)Model=Vicuna-13B2024.01 | 13.6 | — | |
| SystemModel=Vicuna-7B2024.01 | 14.3 | — | |
| Self-CDModel=LLaMA2-7B2024.01 | 15.3 | — | |
| PromptModel=LLaMA2-70B2024.01 | 15.3 | — | |
| Self-CDModel=LLaMA2-13B2024.01 | 15.6 | — | |
| CoT(One)Model=InternLM-7B2024.01 | 16.3 | — | |
| NoSystemModel=LLaMA2-70B2024.01 | 17 | — | |
| AdaCDBase Model=Llama32026.04 | 17.33 | — | |
| SystemModel=Vicuna-13B2024.01 | 18 | — | |
| NoSystemModel=InternLM-7B2024.01 | 19 | — | |
| PromptModel=LLaMA2-13B2024.01 | 19.7 | — | |
| CoT(zero)Model=LLaMA2-70B2024.01 | 19.7 | — | |
| SurgicalBase Model=Llama32026.04 | 20.33 | — | |
| SelfCDBase Model=Llama32026.04 | 20.67 | — | |
| PromptModel=LLaMA2-7B2024.01 | 20.7 | — | |
| CoT(zero)Model=InternLM-7B2024.01 | 21.6 | — | |
| NoSystemModel=LLaMA2-7B2024.01 | 22.3 | — | |
| PromptModel=InternLM-7B2024.01 | 22.3 | — | |
| CoT(zero)Model=LLaMA2-7B2024.01 | 23.3 | — | |
| SystemModel=InternLM-7B2024.01 | 23.7 | — | |
| CoT(One)Model=LLaMA2-70B2024.01 | 24.3 | — | |
| NoSystemModel=LLaMA2-13B2024.01 | 24.7 | — | |
| DefaultBase Model=Llama32026.04 | 25.33 | — | |
| AdaCDBase Model=Gemma22026.04 | 26 | — | |
| CoT(zero)Model=LLaMA2-13B2024.01 | 26.3 | — | |
| CoT(One)Model=LLaMA2-13B2024.01 | 26.3 | — | |
| DefaultBase Model=Gemma22026.04 | 27.33 | — | |
| SelfCDBase Model=Gemma22026.04 | 29 | — | |
| SSDBase Model=Gemma22026.04 | 30.33 | — | |
| ICLModel=LLaMA2-70B2024.01 | 31 | — | |
| PromptBase Model=Llama32026.04 | 31.33 | — | |
| ICLModel=LLaMA2-7B2024.01 | 33 | — | |
| PromptBase Model=Gemma22026.04 | 33 | — | |
| SurgicalBase Model=Gemma22026.04 | 33.67 | — | |
| CoT(One)Model=LLaMA2-7B2024.01 | 36.3 | — | |
| ICLModel=LLaMA2-13B2024.01 | 39.7 | — | |
| SystemModel=LLaMA2-7B2024.01 | 45.7 | — | |
| SystemModel=LLaMA2-70B2024.01 | 45.7 | — | |
| SystemModel=LLaMA2-13B2024.01 | 57 | — | |
| SSDBase Model=Llama32026.04 | 67 | — | |
| DROBase Model=Llama-3.1-8B2026.03 | — | 68 | |
| Llama-3.1-8B-SFTBase Model=Llama-3.1-8B2026.03 | — | 58 | |
| Self-CDBase Model=Llama-3.1-8B2026.03 | — | 80 | |
| System PromptBase Model=Llama-3.1-8B2026.03 | — | 74 | |
| VCLBase Model=Llama-3.1-8B2026.03 | — | 79 | |
| Vector AblationBase Model=Llama-3.1-8B2026.03 | — | 71 |