Saved in:
Bibliographic Details
Main Authors: Lu, Huimin, Isonuma, Masaru, Mori, Junichiro, Sakata, Ichiro
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.16951
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914884923424768
author Lu, Huimin
Isonuma, Masaru
Mori, Junichiro
Sakata, Ichiro
author_facet Lu, Huimin
Isonuma, Masaru
Mori, Junichiro
Sakata, Ichiro
contents Large language models (LLMs) often inherit biases from vast amounts of training corpora. Traditional debiasing methods, while effective to some extent, do not completely eliminate memorized biases and toxicity in LLMs. In this paper, we study an unlearning-based approach to debiasing in LLMs by performing gradient ascent on hate speech against minority groups, i.e., minimizing the likelihood of biased or toxic content. Specifically, we propose a mask language modeling unlearning technique, which unlearns the harmful part of the text. This method enables LLMs to selectively forget and disassociate from biased and harmful content. Experimental results demonstrate the effectiveness of our approach in diminishing bias while maintaining the language modeling abilities. Surprisingly, the results also unveil an unexpected potential for cross-domain transfer unlearning: debiasing in one bias form (e.g. gender) may contribute to mitigating others (e.g. race and religion).
format Preprint
id arxiv_https___arxiv_org_abs_2407_16951
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards Transfer Unlearning: Empirical Evidence of Cross-Domain Bias Mitigation
Lu, Huimin
Isonuma, Masaru
Mori, Junichiro
Sakata, Ichiro
Computation and Language
Machine Learning
Large language models (LLMs) often inherit biases from vast amounts of training corpora. Traditional debiasing methods, while effective to some extent, do not completely eliminate memorized biases and toxicity in LLMs. In this paper, we study an unlearning-based approach to debiasing in LLMs by performing gradient ascent on hate speech against minority groups, i.e., minimizing the likelihood of biased or toxic content. Specifically, we propose a mask language modeling unlearning technique, which unlearns the harmful part of the text. This method enables LLMs to selectively forget and disassociate from biased and harmful content. Experimental results demonstrate the effectiveness of our approach in diminishing bias while maintaining the language modeling abilities. Surprisingly, the results also unveil an unexpected potential for cross-domain transfer unlearning: debiasing in one bias form (e.g. gender) may contribute to mitigating others (e.g. race and religion).
title Towards Transfer Unlearning: Empirical Evidence of Cross-Domain Bias Mitigation
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2407.16951