Towards Resource Efficient and Interpretable Bias Mitigation in Large Language Models
This addresses bias mitigation for marginalized communities in a resource-efficient and interpretable way, though it appears incremental as it builds on existing debiasing methods.
The paper tackles bias in large language models by using small expert models to generate a debiasing signal added at decoding-time, resulting in reduced bias on gender, race, and religion metrics while maintaining model performance.
Although large language models (LLMs) have demonstrated their effectiveness in a wide range of applications, they have also been observed to perpetuate unwanted biases present in the training data, potentially leading to harm for marginalized communities. In this paper, we mitigate bias by leveraging small biased and anti-biased expert models to obtain a debiasing signal that will be added to the LLM output at decoding-time. This approach combines resource efficiency with interpretability and can be optimized for mitigating specific types of bias, depending on the target use case. Experiments on mitigating gender, race, and religion biases show a reduction in bias on several local and global bias metrics while preserving language model performance.