Saved in:
Bibliographic Details
Main Authors: Cho, Young Hyun, Sun, Will Wei
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.22563
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908908626378752
author Cho, Young Hyun
Sun, Will Wei
author_facet Cho, Young Hyun
Sun, Will Wei
contents Preference-based fine-tuning has become an important component in training large language models, and the data used at this stage may contain sensitive user information. A central question is how to design a differentially private pipeline that is well suited to the distinct structure of reinforcement learning from human feedback. We propose a privacy-preserving framework that imposes differential privacy only on reward learning and derives the final policy from the resulting private reward model. Theoretically, we study the suboptimality gap and show that privacy contributes an additional additive term beyond the usual non-private statistical error. We also establish a minimax lower bound and show that the dominant term changes with sample size and privacy level, which in turn characterizes regimes in which the upper bound is rate-optimal up to logarithmic factors. Empirically, synthetic experiments confirm the scaling predicted by the theory, and experiments on the Anthropic HH-RLHF dataset using the Gemma-2B-IT model show stronger private alignment performance than existing differentially private baseline methods across privacy budgets.
format Preprint
id arxiv_https___arxiv_org_abs_2603_22563
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Privacy-Preserving Reinforcement Learning from Human Feedback via Decoupled Reward Modeling
Cho, Young Hyun
Sun, Will Wei
Machine Learning
Preference-based fine-tuning has become an important component in training large language models, and the data used at this stage may contain sensitive user information. A central question is how to design a differentially private pipeline that is well suited to the distinct structure of reinforcement learning from human feedback. We propose a privacy-preserving framework that imposes differential privacy only on reward learning and derives the final policy from the resulting private reward model. Theoretically, we study the suboptimality gap and show that privacy contributes an additional additive term beyond the usual non-private statistical error. We also establish a minimax lower bound and show that the dominant term changes with sample size and privacy level, which in turn characterizes regimes in which the upper bound is rate-optimal up to logarithmic factors. Empirically, synthetic experiments confirm the scaling predicted by the theory, and experiments on the Anthropic HH-RLHF dataset using the Gemma-2B-IT model show stronger private alignment performance than existing differentially private baseline methods across privacy budgets.
title Privacy-Preserving Reinforcement Learning from Human Feedback via Decoupled Reward Modeling
topic Machine Learning
url https://arxiv.org/abs/2603.22563