Improving Industrial Question Answering Chatbots with Domain-Specific LLMs Fine-Tuning

The industrial landscape is experiencing a significant transformation and there is a growing emphasis on human-machine collaboration, personalization, and sustainable manufacturing. Within this context, large language models (LLMs) have emerged as pivotal Artificial Intelligence (AI) tools that enhance communication, streamline operations, and support complex decision-making tasks through advanced natural language processing capabilities. This paper explores the process of fine-tuning a foundational general-purpose LLM, specifically the open-source LLaMA 3 model, to cater to industrial applications. We delve into the methodology of fine-tuning LLaMA on domain-specific data from two distinct industrial scenarios (i.e., product management and production line operations). Our results demonstrate significant improvements in the model’s performance, accuracy, and relevance in generating responses, thereby enhancing decision-making and operational processes in industrial settings. Additionally, we propose the integration of the fine-tuned model into a chatbot system designed to assist human operators by providing targeted, accurate responses in real-time. This paper aims to provide a comprehensive guide for industry practitioners and researchers to harness the power of LLMs for specialized industrial tasks, ultimately contributing to the advancement of intelligent industrial automation and decision-making systems.