Privacy-Preserving In-Context Learning with Differentially Private Few-Shot Generation
Xinyu Tang, Richard Shin, Huseyin A Inan, Andre Manoel, Fatemehsadat Mireshghallah, Zinan Lin, Sivakanth Gopi, Janardhan Kulkarni
Abstract
We study the problem of in-context learning (ICL) with large language models (LLMs) on private datasets. This scenario poses privacy risks, as LLMs may leak or regurgitate the private examples demonstrated in the prompt. We propose a novel algorithm that generates synthetic few-shot demonstrations from the private dataset with formal differential privacy (DP) guarantees, and show empirically that it can achieve effective ICL. We conduct extensive experiments on standard benchmarks and compare our algorithm with non-private ICL and zero-shot solutions. Our results demonstrate that our algorithm can achieve competitive performance with strong privacy levels. These results open up new possibilities for ICL with privacy protection for a broad range of applications.
BibTeX
@inproceedings{
tang2024privacypreserving,
title={Privacy-Preserving In-Context Learning with Differentially Private Few-Shot Generation},
author={Xinyu Tang and Richard Shin and Huseyin A Inan and Andre Manoel and Fatemehsadat Mireshghallah and Zinan Lin and Sivakanth Gopi and Janardhan Kulkarni and Robert Sim},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=oZtt0pRnOl}
}