Refusal Evaluation on CoCoNot Orig
96.1Refusal RateCategorical Steering
Evaluation Results
| Method | Links | |
|---|---|---|
| Categorical SteeringBackbone=REFUSE-LLAMA, Method=Categorical Steering2026.03 | 96.1 | |
| Low-Rank CombinationBackbone=REFUSE-LLAMA, Method=Low-Rank Combination2026.03 | 95.1 | |
| REFUSE-LLAMABackbone=REFUSE-LLAMA2026.03 | 94.01 | |
| Low-Rank CombinationBackbone=DEEPSEEK R1 DISTILL LLAMA, Method=Low-Rank Combination2026.03 | 51.75 | |
| DEEPSEEK R1 DISTILL LLAMABackbone=DEEPSEEK R1 DISTILL LLAMA2026.03 | 50.85 | |
| Low-Rank CombinationBackbone=LLAMA 3 8B INSTRUCT, Method=Low-Rank Combination2026.03 | 28.17 | |
| LLAMA 3 8B INSTRUCTBackbone=LLAMA 3 8B INSTRUCT2026.03 | 26.47 |