paper-with-me

Papers

Differentially Private Zeroth-Order Methods for Scalable Large Language Model Finetuning

2024-02-12 · Z Liu, J Lou, W Bao, Y Hu, B Li, Z Qin, K Ren

Fine-tuning on task-specific datasets is a widely-embraced paradigm of harnessing the powerful capability of pretrained LLMs for various downstream tasks. Due to the popularity of LLMs fine-tuning and its accompanying privacy concerns, differentially private (DP) fine-tuning of pretrained LLMs has been widely used to safeguarding the privacy of task-specific datasets. Lying at the design core of DP LLM fine-tuning methods is the satisfactory tradeoff among privacy, utility, and scalability. Most existing methods build upon the seminal work of DP-SGD. Despite pushing the scalability of DP-SGD to its limit, DP-SGD-based fine-tuning methods are unfortunately limited by the inherent inefficiency of SGD. In this paper, we investigate the potential of DP zeroth-order methods for LLM pretraining, which avoids the scalability bottleneck of SGD by approximating the gradient with the more efficient zeroth-order gradient. Rather than treating the zeroth-order method as a drop-in replacement for SGD, this paper presents a comprehensive study both theoretically and empirically. First, we propose the stagewise DP zeroth-order method (DP-ZOSO) that dynamically schedules key hyperparameters. This design is grounded on the synergy between DP random perturbation and the gradient approximation error of the zeroth-order method, and its effect on fine-tuning trajectory. We provide theoretical analysis for both proposed methods. We conduct extensive empirical analysis on both encoder-only masked language model and decoder-only autoregressive language model, achieving impressive results in terms of scalability and utility regardless of the class of tasks (compared with DPZero, DP-ZOPO improves $4.5\%$ on SST-5, $5.5\%$ on MNLI with RoBERTa-Large and 9.2\% on CB, 3.9\% on BoolQ with OPT-2.7b when $\epsilon=4$, demonstrates more significant enhancement in performance on more complicated tasks).

📄 PDF Abstract BibTeX arXiv:2402.07818

Code (0)

등록된 구현이 없습니다.

Tasks

Language ModelingLanguage ModellingLarge Language Model

Methods 이 논문이 사용한 방법론

Pruning 설명 없음
SGD Stochastic Gradient Descent is an iterative optimization technique that uses minibatches of data to form an expectation of the gradient, rather than the full gradient using…

Similar Papers 제목 키워드 기반

On the Inherent Privacy of Zeroth Order Projected Gradient Descent

2025-07-08 · Devansh Gupta, Meisam Razaviyayn, Vatsal Sharan arxiv

Differentially private zeroth-order optimization methods have recently gained popularity in private fine tuning of machine learning models due to their reduced memory requirements. Current approaches for privatizing zero…

Private Zeroth-Order Optimization with Public Data

2025-11-13 · Xuchen Gong, Tian Li arxiv

One of the major bottlenecks for deploying popular first-order differentially private (DP) machine learning algorithms (e.g., DP-SGD) lies in their high computation and memory cost, despite the existence of optimized imp…

Private Fine-tuning of Large Language Models with Zeroth-order Optimization

2024-01-09 · Xinyu Tang, Ashwinee Panda, Milad Nasr, Saeed Mahloujifar 외

Differentially private stochastic gradient descent (DP-SGD) allows models to be trained in a privacy-preserving manner, but has proven difficult to scale to the era of foundation models. We introduce DP-ZO, a private fin…

Privacy Preserving

DPZero: Private Fine-Tuning of Language Models without Backpropagation

2023-10-14 · Liang Zhang, Bingcong Li, Kiran Koshy Thekumparampil, Sewoong Oh 외

The widespread practice of fine-tuning large language models (LLMs) on domain-specific data faces two major challenges in memory and privacy. First, as the size of LLMs continues to grow, the memory demands of gradient-b…

Noise-Aware Shrinkage for Differentially Private Zeroth-Order Fine-Tuning of Large Language Models

2026-08-04 · Lele Zheng, Weifeng Kong, Xinyi Zhang, Ke Cheng 외 arxiv

Differentially private zeroth-order optimization (DP-ZO) enables memory-efficient private fine-tuning of large language models using only forward evaluations. Existing aggregation-based DP-ZO methods reconstruct model up…