Tsz Fung Yau
2024
Can Machine Unlearning Reduce Social Bias in Language Models?
Omkar Dige
|
Diljot Arneja
|
Tsz Fung Yau
|
Qixuan Zhang
|
Mohammad Bolandraftar
|
Xiaodan Zhu
|
Faiza Khan Khattak
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track
Mitigating bias in language models (LMs) has become a critical problem due to the widespread deployment of LMs in the industry and customer-facing applications. Numerous approaches revolve around data pre-processing and subsequent fine-tuning of language models, tasks that can be both time-consuming and computationally demanding. As alternatives, machine unlearning techniques are being explored, yet there is a notable lack of comparative studies evaluating the effectiveness of these methods. In this work, we explore the effectiveness of two machine unlearning methods: Partitioned Contrastive Gradient Unlearning (PCGU) applied on decoder models, and Negation via Task Vector, and compare them with Direct Preference Optimization (DPO) to reduce social biases in open-source LMs such as LLaMA-2 and OPT. We also implement distributed PCGU for large models. It is empirically shown, through quantitative and qualitative analyses, that negation via Task Vector method outperforms PCGU and is comparable to DPO in debiasing models with minimum deterioration in model performance and perplexity. Negation via Task Vector reduces the bias score by 25.5% for LLaMA-2 and achieves bias reduction of up to 40% for OPT models. Moreover, it can be easily tuned to balance the trade-off between bias reduction and generation quality, unlike DPO.
Search
Co-authors
- Omkar Dige 1
- Diljot Arneja 1
- Qixuan Zhang 1
- Mohammad Bolandraftar 1
- Xiaodan Zhu 1
- show all...