Harmful Question Forgetting on Harm-2 GPTFUZZER WildAttack
0Attack Success Rate (ASR)Unlearning
Evaluation Results
| Method | Links | |
|---|---|---|
| UnlearningModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| EraserModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| EUModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| UnlearningModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| EraserModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| EUModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| UnlearningModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| SKUModel=OLMo-2-1B-Instruct2026.04 | 0 | |
| EUModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| UnlearningModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| EraserModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| SKUModel=OLMo-2-7B-Instruct2026.04 | 0 | |
| EUModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.04 | 0 | |
| DPOModel=OLMo-2-7B-Instruct2026.04 | 0.1 | |
| DPOModel=OLMo-2-1B-Instruct2026.04 | 0.5 | |
| OLMo-2-1B-InstructModel=OLMo-2-1B-Instruct2026.04 | 0.9 | |
| OLMo-2-7B-InstructModel=OLMo-2-7B-Instruct2026.04 | 0.9 | |
| SKUModel=Llama-3.2-3B-Instruct2026.04 | 1 | |
| DPOModel=Llama-3.2-1B-Instruct2026.04 | 2.3 | |
| SKUModel=Llama-3.2-1B-Instruct2026.04 | 3.2 | |
| Llama-3.2-3B-InstructModel=Llama-3.2-3B-Instruct2026.04 | 4.6 | |
| RetainOnlyFTModel=Llama-3.2-3B-Instruct2026.04 | 4.6 | |
| DPOModel=Llama-3.2-3B-Instruct2026.04 | 4.6 | |
| RetainOnlyFTModel=OLMo-2-7B-Instruct2026.04 | 6.5 | |
| EraserModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.04 | 7.8 | |
| Llama-3.2-1B-InstructModel=Llama-3.2-1B-Instruct2026.04 | 8.8 | |
| RetainOnlyFTModel=OLMo-2-1B-Instruct2026.04 | 24.9 | |
| RetainOnlyFTModel=Llama-3.2-1B-Instruct2026.04 | 56.7 |