EPSVec: Efficient and Private Synthetic Data Generation via Dataset Vectors

2026-01-31

arXiv.org

AI Summary

This work addresses the inefficiency and high computational cost of existing privacy-preserving text generation methods in low-data regimes, which often fail to produce high-quality synthetic data. The authors propose a novel approach that constructs a differentially private “dataset vector” to capture the distributional discrepancy between private data and public priors in the activation space, leveraging this vector to guide large language models during text generation. By integrating dataset vectors with differential privacy for the first time, the method decouples the privacy budget from the generation process, enabling unlimited text synthesis without incurring additional privacy loss. Experiments demonstrate that the proposed method significantly outperforms current baselines under low-data conditions, achieving high fidelity in both distribution alignment and downstream task utility while substantially reducing computational overhead.

Abstract

High-quality data is essential for modern machine learning, yet many valuable corpora are sensitive and cannot be freely shared. Synthetic data offers a practical substitute for downstream development, and large language models (LLMs) have emerged as powerful engines for generating it. However, existing private text generation methods are severely inefficient: they are data-intensive, computationally slow, and often require large private corpora or batch sizes to achieve usable quality. We introduce EPSVec, a differentially-private lightweight alternative that steers LLM generation using *dataset vectors*--directions in activation space that capture the distributional gap between private data and public priors. EPSVec extracts and sanitizes steering vectors just once and then performs standard decoding. This decouples the privacy budget from generation, enabling arbitrarily many synthetic samples without additional privacy cost and yielding strong fidelity even in low-data regimes. Furthermore, we enhance our method by utilizing pretrained (base) models and introducing fixed-shot prompting to boost generation diversity and fidelity. Our experiments demonstrate that EPSVec outperforms existing baselines in distributional alignment and downstream utility, particularly in low-data regimes, while significantly reducing computational overhead.

Problem

• synthetic data generation
• differential privacy
• large language models
• data efficiency
• privacy-preserving machine learning

Innovation

• dataset vectors
• differentially private
• synthetic data generation
• low-data regime
• LLM steering