BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimization
This addresses bias mitigation for more ethical and socially responsible LLMs, though it is incremental as it builds on existing DPO methods.
The paper tackled the problem of gender, racial, and religious biases in LLM-generated English text by introducing a Direct Preference Optimization framework, resulting in substantial reductions in biased outputs and outperforming baseline and other open-source models on most benchmarks.
Large Language Models (LLMs) have become pivotal in advancing natural language processing, yet their potential to perpetuate biases poses significant concerns. This paper introduces a new framework employing Direct Preference Optimization (DPO) to mitigate gender, racial, and religious biases in LLM-generated English text. By developing a loss function that favors less biased over biased completions, our approach cultivates a preference for respectful and non-discriminatory language in LLMs. We also contribute a manually designed dataset for training LLMs to recognize and correct biases. This dataset encompasses a diverse range of prompts paired with both biased and unbiased completions. Implementing this approach on the Microsoft Phi-2 model, we demonstrate substantial reductions in biased outputs as our model outperforms the baseline model on almost all bias benchmarks. Our model also achieves better performance compared to other open-source models on most benchmarks. By reducing biases in the language generated by the model, our study marks a significant step towards developing more ethical and socially responsible LLMs. We publicly release BiasDPO dataset on HuggingFace.