RewardDS: Privacy-Preserving Fine-Tuning for Large Language Models via Reward Driven Data Synthesis

TL;DR

RewardDS is a novel framework that enhances privacy-preserving data synthesis for fine-tuning large language models by using reward signals to filter and refine synthetic data, improving quality in sensitive domains.

Contribution

It introduces a reward-guided filtering and self-optimizing refinement approach to generate high-quality synthetic data with differential privacy guarantees.

Findings

Effective noise mitigation in synthetic data

Improved fine-tuning performance in sensitive domains

Demonstrated success across multiple application areas

Abstract

The success of large language models (LLMs) has attracted many individuals to fine-tune them for domain-specific tasks by uploading their data. However, in sensitive areas like healthcare and finance, privacy concerns often arise. One promising solution is to generate synthetic data with Differential Privacy (DP) guarantees to replace private data. However, these synthetic data contain significant flawed data, which are considered as noise. Existing solutions typically rely on naive filtering by comparing ROUGE-L scores or embedding similarities, which are ineffective in addressing the noise. To address this issue, we propose \textit{RewardDS}, a novel privacy-preserving framework that fine-tunes a reward proxy model and uses reward signals to guide the synthetic data generation. Our \textit{RewardDS} introduces two key modules, Reward Guided Filtering and Self-Optimizing Refinement, to both filter and refine the synthetic data, effectively mitigating the noise. Extensive experiments across medical, financial, and code generation domains demonstrate the effectiveness of our method.