Optimization Methods for Personalizing Large Language Models through Retrieval Augmentation

TL;DR

This paper introduces novel optimization algorithms for retrieval-augmented personalization of large language models, significantly improving task-specific performance by optimizing retrieval models based on feedback from downstream tasks.

Contribution

It presents the first optimization methods for retrieval models in personalized LLMs, including reinforcement learning and knowledge distillation techniques, along with a retriever selection model.

Findings

Significant performance improvements on six out of seven datasets.

Effective retrieval optimization enhances personalized generation quality.

Proposed methods outperform baseline retrieval approaches.

Abstract

This paper studies retrieval-augmented approaches for personalizing large language models (LLMs), which potentially have a substantial impact on various applications and domains. We propose the first attempt to optimize the retrieval models that deliver a limited number of personal documents to large language models for the purpose of personalized generation. We develop two optimization algorithms that solicit feedback from the downstream personalized generation tasks for retrieval optimization -- one based on reinforcement learning whose reward function is defined using any arbitrary metric for personalized generation and another based on knowledge distillation from the downstream LLM to the retrieval model. This paper also introduces a pre- and post-generation retriever selection model that decides what retriever to choose for each LLM input. Extensive experiments on diverse tasks from the language model personalization (LaMP) benchmark reveal statistically significant improvements in six out of seven datasets.