Bi-directional Bias Attribution: Debiasing Large Language Models without Modifying Prompts

TL;DR

This paper introduces a novel bias mitigation framework for large language models that detects stereotype words and attributes bias to specific neurons, enabling bias reduction without fine-tuning or prompt changes.

Contribution

It presents a neuron-level bias attribution and intervention method that effectively reduces social biases in LLMs without altering their prompts or requiring additional training.

Findings

Reduces social bias in LLM outputs

Preserves overall model performance

Applicable across multiple LLMs

Abstract

Large language models (LLMs) have demonstrated impressive capabilities across a wide range of natural language processing tasks. However, their outputs often exhibit social biases, raising fairness concerns. Existing debiasing methods, such as fine-tuning on additional datasets or prompt engineering, face scalability issues or compromise user experience in multi-turn interactions. To address these challenges, we propose a framework for detecting stereotype-inducing words and attributing neuron-level bias in LLMs, without the need for fine-tuning or prompt modification. Our framework first identifies stereotype-inducing adjectives and nouns via comparative analysis across demographic groups. We then attribute biased behavior to specific neurons using two attribution strategies based on integrated gradients. Finally, we mitigate bias by directly intervening on their activations at the projection layer. Experiments on three widely used LLMs demonstrate that our method effectively reduces bias while preserving overall model performance. Code is available at the github link: https://github.com/XMUDeepLIT/Bi-directional-Bias-Attribution.