Managing knowledge efficiently is crucial for organizational success. In
manufacturing, operating factories has become increasing knowledge-intensive
putting strain on the factory's capacity to train and support new operators. In
this paper, we introduce a Large Language Model (LLM)-based system designed to
use the extensive knowledge contained in factory documentation. The system aims
to efficiently answer queries from operators and facilitate the sharing of new
knowledge. To assess its effectiveness, we conducted an evaluation in a factory
setting. The results of this evaluation demonstrated the system's benefits;
namely, in enabling quicker information retrieval and more efficient resolution
of issues. However, the study also highlighted a preference for learning from a
human expert when such an option is available. Furthermore, we benchmarked
several closed and open-sourced LLMs for this system. GPT-4 consistently
outperformed its counterparts, with open-source models like StableBeluga2
trailing closely, presenting an attractive option given its data privacy and
customization benefits. Overall, this work offers preliminary insights for
factories considering using LLM-tools for knowledge management.

工厂知识管理的关键是高效地处理知识。本研究使用基于大型语言模型的系统，通过利用工厂文档中的广泛知识，旨在有效回答操作员的查询并促进新知识的分享。在工厂环境中进行的评估结果显示该系统具有快速信息检索和更高效问题解决的优势，然而研究还表明在有人类专家可用时，人们更倾向于向其学习。此外，我们还对该系统进行了几个闭源和开源大型语言模型进行基准测试，GPT-4 始终优于其他模型，而开源模型如 StableBeluga2 则紧随其后，提供了数据隐私和定制化方面的吸引力。总体而言，该研究为考虑采用大型语言模型工具进行知识管理的工厂提供了初步见解。