Mitigating Gender Bias in Code Large Language Models via Model Editing

TL;DR

This paper introduces a dataset and metric to evaluate gender bias in code LLMs, and proposes MG-Editing, a model editing method at multiple granularities, to mitigate bias while preserving code generation performance.

Contribution

It develops a novel model editing approach, MG-Editing, that effectively reduces gender bias in code LLMs at various parameter granularities without sacrificing performance.

Findings

MG-Editing significantly reduces gender bias in code LLMs.

Applying MG-Editing at row and neuron levels yields the best bias mitigation and performance balance.

The proposed methods generalize well across different models and bias scenarios.

Abstract

In recent years, with the maturation of large language model (LLM) technology and the emergence of high-quality programming code datasets, researchers have become increasingly confident in addressing the challenges of program synthesis automatically. However, since most of the training samples for LLMs are unscreened, it is inevitable that LLMs' performance may not align with real-world scenarios, leading to the presence of social bias. To evaluate and quantify the gender bias in code LLMs, we propose a dataset named CodeGenBias (Gender Bias in the Code Generation) and an evaluation metric called FB-Score (Factual Bias Score) based on the actual gender distribution of correlative professions. With the help of CodeGenBias and FB-Score, we evaluate and analyze the gender bias in eight mainstream Code LLMs. Previous work has demonstrated that model editing methods that perform well in knowledge editing have the potential to mitigate social bias in LLMs. Therefore, we develop a model editing approach named MG-Editing (Multi-Granularity model Editing), which includes the locating and editing phases. Our model editing method MG-Editing can be applied at five different levels of model parameter granularity: full parameters level, layer level, module level, row level, and neuron level. Extensive experiments not only demonstrate that our MG-Editing can effectively mitigate the gender bias in code LLMs while maintaining their general code generation capabilities, but also showcase its excellent generalization. At the same time, the experimental results show that, considering both the gender bias of the model and its general code generation capability, MG-Editing is most effective when applied at the row and neuron levels of granularity.