List your company

Bias Evaluation on Race & Gender

20Mean ImprovementiPASwo
3.367.681216.32Sep 25, 2025

Evaluation results

MethodLinks
iPASwoModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.09200.190
iPASwoModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.09170.170
iPASaModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.09170.160
iPASaModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.09140.130
iPASwoModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.09140.130
PASfModel=Llama-3.1-8B-Instruct, Steering Target=residual2025.09110.10
iPASaModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.09100.090
PASfModel=Nous-Hermes-2-Mistral-7B-DPO, Steering Target=residual2025.0970.060
PASfModel=DeepSeek-R1-Distill-8B, Steering Target=residual2025.0940.030