Harmful Request Compliance on Harmful Requests PAIR
69.5Compliance RateMistral-7B-Instruct-v2
Evaluation Results
| Method | Links | |
|---|---|---|
| Mistral-7B-Instruct-v2Base Model=Mistral-7B-Instruct-v2, Training Configuration=Refusal Trained2024.06 | 69.5 | |
| R2D2Base Model=Mistral-7B-Instruct-v2, Training Configuration=Adv Trained2024.06 | 59.9 | |
| Representation ReroutingBase Model=Mistral-7B-Instruct-v2, Training Configuration=+ RR2024.06 | 23.3 | |
| Llama-3-8B-InstructBase Model=Llama-3-8B-Instruct, Training Configuration=Refusal Trained2024.06 | 18.7 | |
| Representation ReroutingBase Model=Llama-3-8B-Instruct, Training Configuration=+ RR2024.06 | 7.5 | |
| CygnetBase Model=Llama-3-8B-Instruct, Training Configuration=Circuit Breaker + RepE Fine-tune2024.06 | 0 |