Interaction Analysis Study on Response Delay of the sLLM Question-Answering System for Industrial Safety Response

This study aims to analyze the performance of an AI agent-based question-answering system in the industrial safety domain. In the industrial safety field, providing rapid and accurate information plays a critical role in preventing and responding to industrial accidents. Therefore, this study compared and analyzed the response latency performance of a question-answering system utilizing a large language model (LLM). To evaluate the performance, two models, LLaMA 3.1 Instruct and Google Gemma 3 4B-IT, were tested under identical conditions. The Google Gemma 3 model measured an average response latency of 88,583.71 ms, while the LLaMA 3.1 model measured 47,534.75 ms. The analysis revealed that the LLaMA 3.1 model exhibited a relatively low response latency, demonstrating faster response times. This study is significant in that it identifies the performance characteristics of large language models in an industrial safety-related question-answering environment and provides useful baseline data for the future development of industrial safety knowledge-based systems.