Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Siddique, Zara, Khalid, Irtaza, Turner, Liam D., Espinosa-Anke, Luis
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912985886228480
author Siddique, Zara
Khalid, Irtaza
Turner, Liam D.
Espinosa-Anke, Luis
author_facet Siddique, Zara
Khalid, Irtaza
Turner, Liam D.
Espinosa-Anke, Luis
contents We present a novel approach to bias mitigation in large language models (LLMs) by applying steering vectors to modify model activations in forward passes. We compute 8 steering vectors, each corresponding to a different social bias axis, such as age, gender, or race, on a training subset of the BBQ dataset and compare the effectiveness of these to 3 additional bias mitigation methods across 4 datasets. When optimized on the BBQ dataset, our individually tuned steering vectors achieve average improvements of 12.8% on BBQ, 8.3% on CLEAR-Bias, and 1% on StereoSet, and show improvements over prompting and Self-Debias in all cases, and improvements over fine-tuning in 12 out of 17 evaluations. In addition, steering vectors showed the lowest impact on MMLU scores of the four bias mitigation methods tested. The work presents the first systematic investigation of steering vectors for bias mitigation, and we demonstrate that they are a powerful and computationally efficient strategy for reducing bias in LLMs, with broader implications for enhancing AI safety.
format Preprint
id arxiv_https___arxiv_org_abs_2503_05371
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs
Siddique, Zara
Khalid, Irtaza
Turner, Liam D.
Espinosa-Anke, Luis
Machine Learning
Artificial Intelligence
Computation and Language
We present a novel approach to bias mitigation in large language models (LLMs) by applying steering vectors to modify model activations in forward passes. We compute 8 steering vectors, each corresponding to a different social bias axis, such as age, gender, or race, on a training subset of the BBQ dataset and compare the effectiveness of these to 3 additional bias mitigation methods across 4 datasets. When optimized on the BBQ dataset, our individually tuned steering vectors achieve average improvements of 12.8% on BBQ, 8.3% on CLEAR-Bias, and 1% on StereoSet, and show improvements over prompting and Self-Debias in all cases, and improvements over fine-tuning in 12 out of 17 evaluations. In addition, steering vectors showed the lowest impact on MMLU scores of the four bias mitigation methods tested. The work presents the first systematic investigation of steering vectors for bias mitigation, and we demonstrate that they are a powerful and computationally efficient strategy for reducing bias in LLMs, with broader implications for enhancing AI safety.
title Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2503.05371