LIDAO: Towards Limited Interventions for Debiasing (Large) Language Models

TL;DR

This paper introduces LIDAO, a novel framework for debiasing large language models with minimal impact on fluency, using an information-theoretic approach and robustifying against adversarial prompts.

Contribution

LIDAO is the first formal, information-theoretic framework for limited intervention debiasing in LLMs that maintains fluency and resists adversarial prompts.

Findings

LIDAO outperforms previous debiasing methods in fluency preservation.

LIDAO effectively debiases models across various sizes from 0.7B to 7B parameters.

LIDAO demonstrates robustness against adversarial prompt scenarios.

Abstract

Large language models (LLMs) have achieved impressive performance on various natural language generation tasks. Nonetheless, they suffer from generating negative and harmful contents that are biased against certain demographic groups (e.g., female), raising severe fairness concerns. As remedies, prior works intervened the generation by removing attitude or demographic information, inevitably degrading the generation quality and resulting in notable \textit{fairness-fluency} trade-offs. However, it is still under-explored to what extent the fluency \textit{has to} be affected in order to achieve a desired level of fairness. In this work, we conduct the first formal study from an information-theoretic perspective. We show that previous approaches are excessive for debiasing and propose LIDAO, a general framework to debias a (L)LM at a better fluency provably. We further robustify LIDAO in adversarial scenarios, where a carefully-crafted prompt may stimulate LLMs exhibiting instruction-following abilities to generate texts with fairness issue appears only when the prompt is also taken into account. Experiments on three LMs ranging from 0.7B to 7B parameters demonstrate the superiority of our method.