Differentially Private In-Context Learning

05/02/2023

∙

An important question in deploying large language models (LLMs) is how to augment LLMs with private data. We propose Differentially Private In-context Learning (DP-ICL) to enable LLMs to adapt to new tasks while maintaining privacy guarantees. DP-ICL performs private inference by establishing noisy consensus over an ensemble of exemplars using the Report-Noisy-Max mechanism. We evaluate DP-ICL on four benchmarks and find that it achieves comparable performance (<2% degradation) with non-private ICL.

READ FULL TEXT

Differentially Private In-Context Learning

Classical-Quantum Differentially Private Mechanisms Beyond Classical Ones

DP-InstaHide: Provably Defusing Poisoning and Backdoor Attacks with Differentially Private Data Augmentations

Privacy-Preserving In-Context Learning with Differentially Private Few-Shot Generation

Differentially Private Graph Learning via Sensitivity-Bounded Personalized PageRank

"Private Prediction Strikes Back!” Private Kernelized Nearest Neighbors with Individual Renyi Filter

Differentially Private Decentralized Optimization with Relay Communication

Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models

Differentially Private In-Context Learning

Related Research

Classical-Quantum Differentially Private Mechanisms Beyond Classical Ones

DP-InstaHide: Provably Defusing Poisoning and Backdoor Attacks with Differentially Private Data Augmentations

Privacy-Preserving In-Context Learning with Differentially Private Few-Shot Generation

Differentially Private Graph Learning via Sensitivity-Bounded Personalized PageRank

"Private Prediction Strikes Back!” Private Kernelized Nearest Neighbors with Individual Renyi Filter

Differentially Private Decentralized Optimization with Relay Communication

Privacy-Preserving Recommender Systems with Synthetic Query Generation using Differentially Private Large Language Models