TelcoLM: collecting data, adapting, and benchmarking language models for the telecommunication domain

TL;DR

This paper explores how to adapt and benchmark large language models for the telecommunications domain by collecting domain-specific data, applying adaptation techniques, and evaluating their performance against generalist models.

Contribution

It introduces a large telco-specific corpus, compares various adaptation methods, and demonstrates that a single instruction-tuning step can effectively adapt models without prior fine-tuning.

Findings

Domain-adapted models challenge larger generalist models in telco tasks.

A single instruction-tuning step suffices for effective adaptation.

The collected corpus contains 800 million tokens and 80,000 instructions.

Abstract

Despite outstanding processes in many tasks, Large Language Models (LLMs) still lack accuracy when dealing with highly technical domains. Especially, telecommunications (telco) is a particularly challenging domain due the large amount of lexical, semantic and conceptual peculiarities. Yet, this domain holds many valuable use cases, directly linked to industrial needs. Hence, this paper studies how LLMs can be adapted to the telco domain. It reports our effort to (i) collect a massive corpus of domain-specific data (800M tokens, 80K instructions), (ii) perform adaptation using various methodologies, and (iii) benchmark them against larger generalist models in downstream tasks that require extensive knowledge of telecommunications. Our experiments on Llama-2-7b show that domain-adapted models can challenge the large generalist models. They also suggest that adaptation can be restricted to a unique instruction-tuning step, dicarding the need for any fine-tuning on raw texts beforehand.