List your company

Detoxification on SafeEdit

95.78Detoxification PerformanceSTA
58.3468.0677.7887.5May 23, 2025

Evaluation results

MethodLinks
STAModel=Gemma-2-9b-it2025.0595.78—————————
CAAModel=Gemma-2-9b-it2025.0591.48—————————
SAE_AXBENCHModel=Gemma-2-9b-it2025.0590.74—————————
STAModel=Gemma-2-9b-pt2025.0589.93—————————
SAE_AXBENCHModel=Gemma-2-9b-pt2025.0586.81—————————
CAAModel=Gemma-2-9b-pt2025.0585.78—————————
Prompt_handModel=Gemma-2-9b-it2025.0578.74—————————
Prompt_autoModel=Gemma-2-9b-it2025.0575.56—————————
Prompt_handModel=Gemma-2-9b-pt2025.0572.52—————————
STAModel=Llama-3.1-8B2025.0570.81—————————
SAE_AXBENCHModel=Llama-3.1-8B2025.0570.74—————————
VanillaModel=Gemma-2-9b-it2025.0570.37—————————
CAAModel=Llama-3.1-8B2025.0568.67—————————
Prompt_autoModel=Gemma-2-9b-pt2025.0564.15—————————
Prompt_handModel=Llama-3.1-8B2025.0563.7—————————
VanillaModel=Gemma-2-9b-pt2025.0562.3—————————
Prompt_autoModel=Llama-3.1-8B2025.0561.63—————————
VanillaModel=Llama-3.1-8B2025.0559.78—————————
OrthoProjTarget Model=LLaMA-7B-Chat2024.06—50.757.950.247.56750.174.965.766.4
OrthoProjTarget Model=LLaMA2-13B-Chat2024.06—51.182.650.672.452.35851.465.175.5
PaCETarget Model=LLaMA-7B-Chat2024.06—69.646.258.275.394.262.380.872.888.3
PaCETarget Model=LLaMA2-13B-Chat2024.06—93.797.997.794.998.996.699.390.898.9
PromptingTarget Model=LLaMA-7B-Chat2024.06—82.547.357.865.275.154.87272.456.1
PromptingTarget Model=LLaMA2-13B-Chat2024.06—35.868.359.352.573.523.47871.166.5
VanillaTarget Model=LLaMA-7B-Chat2024.06—17.619.510.17.7911.317.222.611.817.2
VanillaTarget Model=LLaMA2-13B-Chat2024.06—8.0123.713.619.818.321.613.61416.7
VecAddTarget Model=LLaMA-7B-Chat2024.06—50.958.95953.966.15560.761.766.4
VecAddTarget Model=LLaMA2-13B-Chat2024.06—76.671.47064.387.266.947.474.571.1