Contrastive Private Data Synthesis via Weighted Multi-PLM Fusion

TL;DR

This paper introduces WASP, a novel framework for private data synthesis that combines multiple pre-trained language models with contrastive learning to generate high-quality, privacy-preserving synthetic data, especially effective in data-scarce scenarios.

Contribution

WASP is the first method to utilize weighted multi-PLM fusion with contrastive learning for private data synthesis, addressing data scarcity and bias issues in existing approaches.

Findings

WASP outperforms existing methods on 6 datasets across various tasks.

The Top-Q voting mechanism improves private data distribution estimation.

Dynamic weighting of PLMs enhances synthetic data quality.

Abstract

Substantial quantity and high quality are the golden rules of making a good training dataset with sample privacy protection equally important. Generating synthetic samples that resemble high-quality private data while ensuring Differential Privacy (DP), a formal privacy guarantee, promises scalability and practicality. However, existing methods relying on pre-trained models for data synthesis %that avoid fine-tuning large pre-trained generative models often struggle in data-deficient scenarios, suffering from limited sample size, inevitable generation noise and existing pre-trained model bias. To address these challenges, we propose a novel contrAstive private data Synthesis via Weighted multiple Pre-trained language models (PLM) framework, named as WASP. WASP utilizes limited private samples for more accurate private data distribution estimation via a Top-Q voting mechanism, and leverages low-quality synthetic samples for contrastive generation via collaboration among dynamically weighted multiple pre-trained models.Extensive experiments on 6 well-developed datasets with 6 open-source and 3 closed-source PLMs demonstrate the superiority of WASP in improving model performance over diverse downstream tasks. Code is available at https://anonymous.4open.science/r/WASP.