تقرير
LMO-DP: Optimizing the Randomization Mechanism for Differentially Private Fine-Tuning (Large) Language Models
العنوان: | LMO-DP: Optimizing the Randomization Mechanism for Differentially Private Fine-Tuning (Large) Language Models |
---|---|
المؤلفون: | Yang, Qin, Mohammad, Meisam, Wang, Han, Payani, Ali, Kundu, Ashish, Shu, Kai, Yan, Yan, Hong, Yuan |
سنة النشر: | 2024 |
المجموعة: | Computer Science |
مصطلحات موضوعية: | Computer Science - Cryptography and Security, Computer Science - Computation and Language, Computer Science - Machine Learning |
الوصف: | Differentially Private Stochastic Gradient Descent (DP-SGD) and its variants have been proposed to ensure rigorous privacy for fine-tuning large-scale pre-trained language models. However, they rely heavily on the Gaussian mechanism, which may overly perturb the gradients and degrade the accuracy, especially in stronger privacy regimes (e.g., the privacy budget $\epsilon < 3$). To address such limitations, we propose a novel Language Model-based Optimal Differential Privacy (LMO-DP) mechanism, which takes the first step to enable the tight composition of accurately fine-tuning (large) language models with a sub-optimal DP mechanism, even in strong privacy regimes (e.g., $0.1\leq \epsilon<3$). Furthermore, we propose a novel offline optimal noise search method to efficiently derive the sub-optimal DP that significantly reduces the noise magnitude. For instance, fine-tuning RoBERTa-large (with 300M parameters) on the SST-2 dataset can achieve an accuracy of 92.20% (given $\epsilon=0.3$, $\delta=10^{-10}$) by drastically outperforming the Gaussian mechanism (e.g., $\sim 50\%$ for small $\epsilon$ and $\delta$). We also draw similar findings on the text generation tasks on GPT-2. Finally, to our best knowledge, LMO-DP is also the first solution to accurately fine-tune Llama-2 with strong differential privacy guarantees. The code will be released soon and available upon request. Comment: 18 pages, 15 figures |
نوع الوثيقة: | Working Paper |
الوصول الحر: | http://arxiv.org/abs/2405.18776Test |
رقم الانضمام: | edsarx.2405.18776 |
قاعدة البيانات: | arXiv |
الوصف غير متاح. |