List your company
ResearchFormerly WizWand

Jailbreak Attempt Forgetting on JB-1 Jailbreak Harm-1

73.1ASR (%)RetainOnlyFT
-2.92416.81336.5556.287Apr 7, 2026
Updated 6mo ago

Evaluation results

MethodLinks
RetainOnlyFTModel=Llama-3.2-1B-Instruct2026.0473.1
RetainOnlyFTModel=OLMo-2-1B-Instruct2026.0455.8
RetainOnlyFTModel=OLMo-2-7B-Instruct2026.0454.5
Llama-3.2-1B-InstructModel=Llama-3.2-1B-Instruct2026.0452.5
EraserModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.0445.5
UnlearningModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.0442
RetainOnlyFTModel=Llama-3.2-3B-Instruct2026.0429.4
EraserModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.0429.3
OLMo-2-7B-InstructModel=OLMo-2-7B-Instruct2026.0429.2
DPOModel=OLMo-2-7B-Instruct2026.0428.5
DPOModel=OLMo-2-1B-Instruct2026.0419.6
Llama-3.2-3B-InstructModel=Llama-3.2-3B-Instruct2026.0418.1
DPOModel=Llama-3.2-1B-Instruct2026.0416.3
OLMo-2-1B-InstructModel=OLMo-2-1B-Instruct2026.0414.2
UnlearningModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.0410.7
DPOModel=Llama-3.2-3B-Instruct2026.048.7
SKUModel=Llama-3.2-1B-Instruct2026.048.5
SKUModel=Llama-3.2-3B-Instruct2026.048.1
SKUModel=OLMo-2-1B-Instruct2026.046.1
SKUModel=OLMo-2-7B-Instruct2026.043.6
UnlearningModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.040.6
EraserModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.040.3
EUModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.040.1
EUModel=Llama-3.2-1B-Instruct, Hyperparameter (λ)=0.62026.040
EraserModel=Llama-3.2-3B-Instruct, Hyperparameter (λ)=0.62026.040
UnlearningModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.040
EUModel=OLMo-2-1B-Instruct, Hyperparameter (λ)=0.62026.040
EUModel=OLMo-2-7B-Instruct, Hyperparameter (λ)=0.62026.040