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

TL;DR

TelcoLM demonstrates a data-centric approach to tailoring LLMs to the telecommunications domain by collecting a large telco-specific corpus and instruction set, then systematically comparing domain-adaptation strategies. The results show that instruction-tuning (IAPT) can achieve strong performance, even when pretraining on raw telco data (DAPT) is omitted, and that a blend of domain- and general instructions yields the best telco-task performance. A dedicated telco benchmark suite spanning MCQ, Open QA, and abstract generation enables nuanced evaluation of domain knowledge and generalization. Overall, the work provides a cost-efficient path to effective telco LLMs and highlights the importance of data quality and instruction design for domain-specific NLP deployments.

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.

Figures (1)

• Figure 1: Steps for LM training and adaptation.