List your company

Bias Evaluation on Race & SES

17Mean ImprovementiPASwo
-0.683.918.513.09Sep 25, 2025

Evaluation results

MethodLinks
iPASwoModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.09170.160
iPASwoModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.09100.080
iPASwoModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.09100.090
iPASwoModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.09100.090
iPASaModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.09100.090
iPASwoModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0990.080
iPASaModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.0990.080
iPASaModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0980.070
iPASwoModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0980.070
iPASaModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0970.060
iPASaModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0970.060
iPASaModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0970.070
PASfModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0950.040
PASfModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.0940.030
PASfModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.0940.030
PASfModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0930.020
PASfModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.09100
PASfModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.090-0.010.61