(How) Do Large Language Models Understand High-Level Message Sequence Charts?

Large Language Models' (LLMs) ability to comprehend High-Level Message Sequence Charts (MSCs) is a crucial aspect of their application in software development, particularly in architectural design specification. Researchers have investigated this capability, seeking to understand whether LLMs can consistently interpret the semantics of MSCs. The study aimed to address the knowledge gap in this area, as the use of LLMs in automating tasks across the software development life-cycle continues to grow. The findings of this research have significant implications for the reliability and consistency of LLMs in handling complex design specifications¹. As LLMs become increasingly integrated into various aspects of software development, their ability to accurately understand and interpret MSCs is essential for ensuring the correctness and validity of the resulting systems. This understanding is critical for practitioners, as it directly impacts the trustworthiness and security of the systems they develop.