This paper explores the use of Large Language Models (LLMs) as Question Answering (QA) assistants for 3GPP standards. The authors evaluate the performance of state-of-the-art LLMs in answering questions based on 3GPP documents. They introduce TeleRoBERTa, an extractive QA model that performs as well as top foundation models but with significantly fewer parameters. They also provide data preprocessing and fine-tuning methods to improve model accuracy for 3GPP standards. The results show that LLMs can be used as credible reference tools for telecom technical documents, with potential applications in troubleshooting, maintenance, network operations, and software development. The paper discusses the architecture of LLMs, including the Transformer model, and the use of prompt engineering and fine-tuning to adapt LLMs to specific domains. They also describe the use of Retrieval Augmented Generation (RAG) to enhance model performance by incorporating external data. The authors evaluate the performance of various LLMs using metrics such as BERTScore and GPT-4 Ref. They find that TeleRoBERTa performs on-par with larger foundation models, and that fine-tuning and context engineering can significantly improve accuracy. The paper also addresses challenges such as misinterpretation of technical jargon, difficulty in identifying information in tables, and issues with cross-referencing in documents. They propose solutions such as replacing tables with natural language and fine-tuning models on domain-specific data to reduce hallucinations and improve accuracy. The results show that the fine-tuned model outperforms the baseline model by approximately 16%. The authors conclude that LLMs have significant potential for use in telecom domains, and that further research is needed to explore their applications in various areas such as field service operations, customer incident management, and software development. The paper also highlights the importance of using appropriate metrics and methods to evaluate the performance of LLMs in QA tasks.This paper explores the use of Large Language Models (LLMs) as Question Answering (QA) assistants for 3GPP standards. The authors evaluate the performance of state-of-the-art LLMs in answering questions based on 3GPP documents. They introduce TeleRoBERTa, an extractive QA model that performs as well as top foundation models but with significantly fewer parameters. They also provide data preprocessing and fine-tuning methods to improve model accuracy for 3GPP standards. The results show that LLMs can be used as credible reference tools for telecom technical documents, with potential applications in troubleshooting, maintenance, network operations, and software development. The paper discusses the architecture of LLMs, including the Transformer model, and the use of prompt engineering and fine-tuning to adapt LLMs to specific domains. They also describe the use of Retrieval Augmented Generation (RAG) to enhance model performance by incorporating external data. The authors evaluate the performance of various LLMs using metrics such as BERTScore and GPT-4 Ref. They find that TeleRoBERTa performs on-par with larger foundation models, and that fine-tuning and context engineering can significantly improve accuracy. The paper also addresses challenges such as misinterpretation of technical jargon, difficulty in identifying information in tables, and issues with cross-referencing in documents. They propose solutions such as replacing tables with natural language and fine-tuning models on domain-specific data to reduce hallucinations and improve accuracy. The results show that the fine-tuned model outperforms the baseline model by approximately 16%. The authors conclude that LLMs have significant potential for use in telecom domains, and that further research is needed to explore their applications in various areas such as field service operations, customer incident management, and software development. The paper also highlights the importance of using appropriate metrics and methods to evaluate the performance of LLMs in QA tasks.