List your company

Bias Evaluation on SexualOrientation

0.1Mean ImprovementiPASwo
0.01680.03840.060.0816Sep 25, 2025

Evaluation results

MethodLinks
iPASwoModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.090.10.070
iPASaModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.090.070.040
PASfModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.090.060.020
iPASaModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.090.060.040
iPASwoModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.090.050.020
PASfModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.090.040.020
iPASaModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.090.030.030
iPASwoModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.090.030.010
PASfModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.090.0200.02