Multi-objective Large Language Model Alignment with Hierarchical Experts

TL;DR

This paper introduces HoE, a lightweight, plug-and-play hierarchical mixture-of-experts method that enables large language models to efficiently balance multiple conflicting objectives without retraining, improving alignment across diverse preferences.

Contribution

HoE is a novel hierarchical mixture-of-experts approach that allows LLMs to adapt across the entire Pareto frontier without retraining, balancing multiple objectives efficiently.

Findings

Outperforms 15 recent baselines on 14 objectives and 200 preferences.

Achieves optimal Pareto frontiers with reduced training cost.

Demonstrates versatility across various tasks and benchmarks.

Abstract

Aligning large language models (LLMs) to simultaneously satisfy multiple objectives remains a significant challenge, especially given the diverse and often conflicting nature of human preferences. Existing alignment methods struggle to balance trade-offs effectively, often requiring costly retraining or yielding suboptimal results across the Pareto frontier of preferences. In this paper, we introduce \textit{HoE}(Hierarchical Mixture-of-Experts), a \textit{lightweight}, \textit{parameter-efficient}, and \textit{plug-and-play} approach that eliminates the need for model training, while enabling LLMs to adapt across the entire Pareto frontier and accommodate diverse user preferences. In particular, \textit{HoE} consists of three hierarchical components: LoRA Experts, Router Experts and Preference Routing, reaching optimal Pareto frontiers and achieving a trade-off between parameter size, training cost, and performance. We evaluate \textit{HoE} across various tasks on 14 objectives and 200 different preferences among 6 benchmarks, demonstrating superior performance over 15 recent baselines. Code is available in the supplementary materials.