List your company

Safety Evaluation on Harmful and Jailbreak datasets

16Harm-1 ScoreRetainOnlyFT
-0.643.68812.32Apr 7, 2026

Evaluation results

MethodLinks
RetainOnlyFTBackbone=Llama-3.2-1B-Instruct2026.041619.448.652.6
RetainOnlyFTBackbone=OLMo-2-1B-Instruct2026.041118.939.252.6
RetainOnlyFTBackbone=Llama-3.2-3B-Instruct2026.04918.960.358.9
Llama-3.2-1B-InstructBackbone=Llama-3.2-1B-Instruct2026.0478.852.551.1
RetainOnlyFTBackbone=OLMo-2-7B-Instruct2026.04613.445.945.6
SKUBackbone=Llama-3.2-1B-Instruct2026.045611.910.9
DPOBackbone=Llama-3.2-1B-Instruct2026.0422.86.75.3
Llama-3.2-3B-InstructBackbone=Llama-3.2-3B-Instruct2026.0424.618.110.2
DPOBackbone=Llama-3.2-3B-Instruct2026.0422.33.52.5
OLMo-2-1B-InstructBackbone=OLMo-2-1B-Instruct2026.0410.914.211.1
EraserBackbone=OLMo-2-1B-Instruct, Lambda (λ)=0.952026.0419.27.36.5
EUBackbone=OLMo-2-1B-Instruct, Lambda (λ)=0.952026.041000
DPOBackbone=OLMo-2-7B-Instruct2026.041028.625.8
UnlearningBackbone=Llama-3.2-1B-Instruct, Lambda (λ)=0.62026.04002122.9
EraserBackbone=Llama-3.2-1B-Instruct, Lambda (λ)=0.62026.0400.517.216
EUBackbone=Llama-3.2-1B-Instruct, Lambda (λ)=0.62026.040000.3
UnlearningBackbone=Llama-3.2-3B-Instruct, Lambda (λ)=0.62026.040025.523.3
EraserBackbone=Llama-3.2-3B-Instruct, Lambda (λ)=0.62026.04000.70.3
SKUBackbone=Llama-3.2-3B-Instruct2026.0401.37.65.7
EUBackbone=Llama-3.2-3B-Instruct, Lambda (λ)=0.62026.04000.10.2
DPOBackbone=OLMo-2-1B-Instruct2026.0400.120.820.1
UnlearningBackbone=OLMo-2-1B-Instruct, Lambda (λ)=0.952026.040022.522.9
SKUBackbone=OLMo-2-1B-Instruct2026.04006.25.7
OLMo-2-7B-InstructBackbone=OLMo-2-7B-Instruct2026.0400.929.226.6
UnlearningBackbone=OLMo-2-7B-Instruct, Lambda (λ)=0.62026.040042.141.9
EraserBackbone=OLMo-2-7B-Instruct, Lambda (λ)=0.62026.040045.140.3
SKUBackbone=OLMo-2-7B-Instruct2026.040076.5
EUBackbone=OLMo-2-7B-Instruct, Lambda (λ)=0.62026.04000.30.2