This research presents QMOS, a novel technique that significantly improves Large Language Model performance for telecommunications question-answering systems.

• Combines Question Masked Loss and Option Shuffling to enhance domain-specific accuracy
• Addresses challenges of specialized telecom vocabulary and complex technical concepts
• Builds upon RAG (Retrieval-Augmented Generation) pipelines with telecom-specific optimizations
• Offers practical engineering solutions for creating more accurate and reliable telecom support systems

This advancement matters for engineering teams implementing AI-powered customer support in telecommunications, enabling more precise technical responses in this domain-specific context.

QMOS: Enhancing LLMs for Telecommunication with Question Masked loss and Option Shuffling