Language Model Networks: Supervision-Efficient Learning through Dense Communication

TL;DR

This paper introduces LMNet, a dense, differentiable language model network that enhances communication efficiency and end-to-end training in multi-model systems by replacing natural language communication with dense vector exchange.

Contribution

The paper proposes LMNet, a novel architecture that enables dense, trainable communication between language models, improving efficiency and optimization over traditional natural language interfaces.

Findings

LMNet achieves comparable performance with less training cost.

Dense communication improves information transfer efficiency.

Effective adaptation under limited supervision is demonstrated.

Abstract

Language models are increasingly used not only as standalone predictors but also as components in larger inference systems, from test-time reasoning to multi-model collaboration. We study language model networks, where pre-trained language models serve as reusable nodes and intelligence emerges from their topology, communication, and optimization. Existing systems mostly communicate through natural language: easy to deploy, but discrete, inefficient, and hard to optimize from end-task supervision. We propose LMNet, a dense and differentiable realization of this paradigm. LMNet uses stripped LLMs as vertex modules and trainable seq2seq modules as communication edges, enabling intermediate nodes to exchange dense vectors while preserving natural-language input and output at the system boundary. By bypassing intermediate embedding and de-embedding, LMNet enables efficient information transfer, end-to-end gradient optimization, and learned communication beyond hand-designed protocols. Experiments show performance with small additional training cost and effective adaptation under limited supervision.