This paper presents a novel approach to aligning large language models (LLMs) with individual human preferences, sometimes referred to as Reinforcement Learning from \textit{Personalized} Human Feedback (RLPHF). Given stated preferences along multiple dimensions, such as helpfulness, conciseness, or humor, the goal is to create an LLM without re-training that best adheres to this specification. Starting from specialized expert LLMs, each trained for one such particular preference dimension, we propose a black-box method that merges their outputs on a per-token level. We train a lightweight Preference Control Model (PCM) that dynamically translates the preference description and current context into next-token prediction weights. By combining the expert models' outputs at the token level, our approach dynamically generates text that optimizes the given preference. Empirical tests show that our method matches or surpasses existing preference merging techniques, providing a scalable, efficient alternative to fine-tuning LLMs for individual personalization.
@article{arxiv.2407.04181,
title = {Orchestrating LLMs with Different Personalizations},
author = {Jin Peng Zhou and Katie Z Luo and Jingwen Gu and Jason Yuan and Kilian Q. Weinberger and Wen Sun},
journal= {arXiv preprint arXiv:2407.04181},
year = {2024}
}