Differentially Private In-Context Learning with Nearest Neighbor Search

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Koskela, Antti, Kulkarni, Tejas, Zumot, Laith
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence Cryptography and Security
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866915602260557824
author	Koskela, Antti Kulkarni, Tejas Zumot, Laith
author_facet	Koskela, Antti Kulkarni, Tejas Zumot, Laith
contents	Differentially private in-context learning (DP-ICL) has recently become an active research topic due to the inherent privacy risks of in-context learning. However, existing approaches overlook a critical component of modern large language model (LLM) pipelines: the similarity search used to retrieve relevant context data. In this work, we introduce a DP framework for in-context learning that integrates nearest neighbor search of relevant examples in a privacy-aware manner. Our method outperforms existing baselines by a substantial margin across all evaluated benchmarks, achieving more favorable privacy-utility trade-offs. To achieve this, we employ nearest neighbor retrieval from a database of context data, combined with a privacy filter that tracks the cumulative privacy cost of selected samples to ensure adherence to a central differential privacy budget. Experimental results on text classification and document question answering show a clear advantage of the proposed method over existing baselines.
format	Preprint
id	arxiv_https___arxiv_org_abs_2511_04332
institution	arXiv
publishDate	2025
record_format	arxiv
spellingShingle	Differentially Private In-Context Learning with Nearest Neighbor Search Koskela, Antti Kulkarni, Tejas Zumot, Laith Machine Learning Artificial Intelligence Cryptography and Security Differentially private in-context learning (DP-ICL) has recently become an active research topic due to the inherent privacy risks of in-context learning. However, existing approaches overlook a critical component of modern large language model (LLM) pipelines: the similarity search used to retrieve relevant context data. In this work, we introduce a DP framework for in-context learning that integrates nearest neighbor search of relevant examples in a privacy-aware manner. Our method outperforms existing baselines by a substantial margin across all evaluated benchmarks, achieving more favorable privacy-utility trade-offs. To achieve this, we employ nearest neighbor retrieval from a database of context data, combined with a privacy filter that tracks the cumulative privacy cost of selected samples to ensure adherence to a central differential privacy budget. Experimental results on text classification and document question answering show a clear advantage of the proposed method over existing baselines.
title	Differentially Private In-Context Learning with Nearest Neighbor Search
topic	Machine Learning Artificial Intelligence Cryptography and Security
url	https://arxiv.org/abs/2511.04332