Year: 2024
Type: preprint
Abstract: Large Language Models (LLMs) rely on the contextual information embedded in examples/demonstrations to perform in-context learning (ICL). To mitigate the risk of LLMs potentially leaking private information contained in examples in the prompt, we introduce a novel data-adaptive differentially private algorithm called AdaDPSyn to generate synthetic examples from the private dataset and then use these synthetic examples to perform ICL. The objective of AdaDPSyn is to adaptively adjust the noise level in the data synthesis mechanism according to the inherent statistical properties of the data, thereby preserving high ICL accuracy while maintaining formal differential privacy guarantees. A key innovation in AdaDPSyn is the Precision-Focused Iterative Radius Reduction technique, which dynamically refines the aggregation radius - the scope of data grouping for noise addition - based on patterns observed in data clustering, thereby minimizing the amount of additive noise. We conduct extensive experiments on standard benchmarks and compare AdaDPSyn with DP few-shot generation algorithm (Tang et al., 2023). The experiments demonstrate that AdaDPSyn not only outperforms DP few-shot generation, but also maintains high accuracy levels close to those of non-private baselines, providing an effective solution for ICL with privacy protection.
Source: arXiv (Cornell University)
Authors: Gao, Fengyu, Zhou, Ruida, Wang, Tianhao, Shen, Cong, Yang, Jing
Institutions: University of Virginia, University of California, Los Angeles
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: Collaborative Research: SWIFT: SMALL: Learning-Efficient Spectrum Access for No-Sensing Devices in Shared Spectrum, Collaborative Research: CPS Medium: Learning through the Air: Cross-Layer UAV Orchestration for Online Federated Optimization, Collaborative Research: Optimized Testing Strategies for Fighting Pandemics: Fundamental Limits and Efficient Algorithms, Collaborative Research: IMR: MM-1B: Foundations for Differentially Private Internet Measurement, CAREER: Towards a Communication Foundation for Distributed and Decentralized Machine Learning, Collaborative Research: MLWiNS: Dino-RL: A Domain Knowledge Enriched Reinforcement Learning Framework for Wireless Network Optimization, Integrating Federated Split Neural Network with Artificial Stereoscopic Compound Eyes for Optical Flow Sensing in 3D Space with Precision
OpenAlex ID: W4403578567