Scholarlink

Large language models (LLMs) have demonstrated potential in knowledge representation and natural language processing, with the ability to understand context, analyse data, and provide step-by-step guidance across diverse topics. Learning factories, which integrate theoretical knowledge with hands-on industrial training, could benefit from LLMs by leveraging the capabilities of LLMs to offer tailored guidance. However, the extent to which LLMs can provide guidance in manufacturing environments remains largely unexplored. This study assesses the capabilities of multiple LLMs in manufacturing-related tasks within a simulated manufacturing process environment. While LLMs excel at solving general problems, the LLMs considered fall short in providing accurate answers to manufacturing-specific questions where precision is critical. Given simulation data, the LLMs demonstrated an understanding of context and can generate code, but cannot produce correct and executable simulation code. These results highlight the limitations of using LLMs in manufacturing environments and suggest the need for developing manufacturing-specific LLMs.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluating the Capabilities and Assessing the Inference of Large Language Models in a Learning Factory

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Refiloe Motsoeneng,
  • Leon Eldon Burger

摘要

Large language models (LLMs) have demonstrated potential in knowledge representation and natural language processing, with the ability to understand context, analyse data, and provide step-by-step guidance across diverse topics. Learning factories, which integrate theoretical knowledge with hands-on industrial training, could benefit from LLMs by leveraging the capabilities of LLMs to offer tailored guidance. However, the extent to which LLMs can provide guidance in manufacturing environments remains largely unexplored. This study assesses the capabilities of multiple LLMs in manufacturing-related tasks within a simulated manufacturing process environment. While LLMs excel at solving general problems, the LLMs considered fall short in providing accurate answers to manufacturing-specific questions where precision is critical. Given simulation data, the LLMs demonstrated an understanding of context and can generate code, but cannot produce correct and executable simulation code. These results highlight the limitations of using LLMs in manufacturing environments and suggest the need for developing manufacturing-specific LLMs.