In-Contextual Gender Bias Suppression for Large Language Models
This addresses bias issues in LLMs, particularly for closed models like GPT-4, but is incremental as it builds on prior debiasing approaches.
The paper tackles gender bias in large language models by proposing a method that uses textual preambles to suppress biases without accessing model parameters, showing effectiveness on the CrowsPairs dataset and acceptable impact on downstream tasks like HellaSwag and COPA.
Despite their impressive performance in a wide range of NLP tasks, Large Language Models (LLMs) have been reported to encode worrying-levels of gender biases. Prior work has proposed debiasing methods that require human labelled examples, data augmentation and fine-tuning of LLMs, which are computationally costly. Moreover, one might not even have access to the model parameters for performing debiasing such as in the case of closed LLMs such as GPT-4. To address this challenge, we propose bias suppression that prevents biased generations of LLMs by simply providing textual preambles constructed from manually designed templates and real-world statistics, without accessing to model parameters. We show that, using CrowsPairs dataset, our textual preambles covering counterfactual statements can suppress gender biases in English LLMs such as LLaMA2. Moreover, we find that gender-neutral descriptions of gender-biased objects can also suppress their gender biases. Moreover, we show that bias suppression has acceptable adverse effect on downstream task performance with HellaSwag and COPA.