RouterDC: Query-Based Router by Dual Contrastive Learning for Assembling Large Language Models

TL;DR

RouterDC introduces a dual contrastive learning approach to improve query-based routing among multiple large language models, effectively assembling their strengths and outperforming existing methods on various tasks.

Contribution

The paper proposes a novel contrastive learning-based router for assembling LLMs, addressing limitations of existing routing models in multi-LLM scenarios.

Findings

Outperforms existing routing methods on in-distribution tasks (+2.76%)

Achieves better out-of-distribution performance (+1.90%)

Effectively assembles multiple LLMs for improved task performance

Abstract

Recent works show that assembling multiple off-the-shelf large language models (LLMs) can harness their complementary abilities. To achieve this, routing is a promising method, which learns a router to select the most suitable LLM for each query. However, existing routing models are ineffective when multiple LLMs perform well for a query. To address this problem, in this paper, we propose a method called query-based Router by Dual Contrastive learning (RouterDC). The RouterDC model consists of an encoder and LLM embeddings, and we propose two contrastive learning losses to train the RouterDC model. Experimental results show that RouterDC is effective in assembling LLMs and largely outperforms individual top-performing LLMs as well as existing routing methods on both in-distribution (+2.76\%) and out-of-distribution (+1.90\%) tasks. Source code is available at https://github.com/shuhao02/RouterDC.