LFTF: Locating First and Then Fine-Tuning for Mitigating Gender Bias in Large Language Models
It addresses social bias issues in AI systems, particularly gender bias in LLMs, which is a critical problem for fairness and ethics in AI applications.
The paper tackles gender bias in large language models by proposing the LFTF algorithm, which locates and fine-tunes specific model blocks, resulting in significant bias mitigation while preserving general capabilities.
Nowadays, Large Language Models (LLMs) have attracted widespread attention due to their powerful performance. However, due to the unavoidable exposure to socially biased data during training, LLMs tend to exhibit social biases, particularly gender bias. To better explore and quantifying the degree of gender bias in LLMs, we propose a pair of datasets named GenBiasEval and GenHintEval, respectively. The GenBiasEval is responsible for evaluating the degree of gender bias in LLMs, accompanied by an evaluation metric named AFGB-Score (Absolutely Fair Gender Bias Score). Meanwhile, the GenHintEval is used to assess whether LLMs can provide responses consistent with prompts that contain gender hints, along with the accompanying evaluation metric UB-Score (UnBias Score). Besides, in order to mitigate gender bias in LLMs more effectively, we present the LFTF (Locating First and Then Fine-Tuning) algorithm.The algorithm first ranks specific LLM blocks by their relevance to gender bias in descending order using a metric called BMI (Block Mitigating Importance Score). Based on this ranking, the block most strongly associated with gender bias is then fine-tuned using a carefully designed loss function. Numerous experiments have shown that our proposed LFTF algorithm can significantly mitigate gender bias in LLMs while maintaining their general capabilities.