HAPS: Hierarchical LLM Routing with Joint Architecture and Parameter Search

TL;DR

HAPS introduces a hierarchical framework for LLM routing that jointly searches for optimal architectures and parameters, significantly improving task performance over existing methods.

Contribution

It is the first to jointly optimize LLM architectures and parameters using a hierarchical routing framework with shared parameter generation.

Findings

HAPS outperforms strong routing baselines on benchmark tasks.

The shared parameter generation enhances routing capabilities.

Joint architecture and parameter search improves task-specific performance.

Abstract

Large language model (LLM) routing aims to exploit the specialized strengths of different LLMs for diverse tasks. However, existing approaches typically focus on selecting LLM architectures while overlooking parameter settings, which are critical for task performance. In this paper, we introduce HAPS, a hierarchical LLM routing framework that jointly searches over model architectures and parameters. Specifically, we use a high-level router to select among candidate LLM architectures, and then search for the optimal parameters for the selected architectures based on a low-level router. We design a parameter generation network to share parameters between the two routers to mutually enhance their capabilities. In the training process, we design a reward-augmented objective to effectively optimize our framework. Experiments on two commonly used benchmarks show that HAPS consistently outperforms strong routing baselines. We have released our code at https://github.com/zihangtian/HAPS.