LoRe: Personalizing LLMs via Low-Rank Reward Modeling

Avinandan Bose, Zhihan Xiong, Yuejie Chi, Simon Shaolei Du, Lin Xiao, Maryam Fazel

Keywords: Personalization, Pluralistic Alignment

TL;DR: We introduce low-rank personalized preference modeling for LLMs, enabling scalable and efficient user-specific reward learning with superior generalization and few-shot adaptation.

Abstract: Personalizing large language models (LLMs) to accommodate diverse user preferences is essential for enhancing alignment and user satisfaction. Traditional reinforcement learning from human feedback (RLHF) approaches often rely on monolithic value representations, limiting their ability to adapt to individual preferences. We introduce a novel framework that leverages \textit{low-rank preference modeling} to efficiently learn and generalize user-specific reward functions. By representing reward functions in a low-dimensional subspace and modeling individual preferences as weighted combinations of shared basis functions, our approach avoids rigid user categorization while enabling scalability and few-shot adaptation. We validate our method on multiple preference datasets, demonstrating superior generalization to unseen users and improved accuracy in preference prediction tasks.

Submission Number: 10