![[2602.21218] EPSVec: Efficient and Private Synthetic Data Generation via Dataset Vectors](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.21218] EPSVec: Efficient and Private Synthetic Data Generation via Dataset Vectors
Summary
The paper introduces EPSVec, a novel method for generating synthetic data using dataset vectors, enhancing privacy and efficiency in machine learning applications.
Why It Matters
As data privacy concerns grow, EPSVec offers a solution for generating high-quality synthetic data without compromising sensitive information. This method significantly reduces computational costs and improves data utility, making it crucial for researchers and practitioners in AI and machine learning.
Key Takeaways
- EPSVec utilizes dataset vectors to enhance synthetic data generation while maintaining privacy.
- The method decouples privacy budget from data generation, allowing for unlimited synthetic samples without additional privacy costs.
- EPSVec demonstrates superior performance in low-data scenarios compared to existing methods.
- The approach reduces computational overhead, making it more efficient for practical applications.
- Utilizing pretrained models and fixed-shot prompting boosts generation diversity and fidelity.
Computer Science > Computation and Language arXiv:2602.21218 (cs) [Submitted on 31 Jan 2026] Title:EPSVec: Efficient and Private Synthetic Data Generation via Dataset Vectors Authors:Amin Banayeeanzade, Qingchuan Yang, Deqing Fu, Spencer Hong, Erin Babinsky, Alfy Samuel, Anoop Kumar, Robin Jia, Sai Praneeth Karimireddy View a PDF of the paper titled EPSVec: Efficient and Private Synthetic Data Generation via Dataset Vectors, by Amin Banayeeanzade and 8 other authors View PDF HTML (experimental) 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 ...
