Using LLMs in Software Design: An Empirical Study of GitHub and A Practitioner Survey
An evidence-based characterization of current LLM use in software design from both open-source and practitioner perspectives is provided, showing that developers in the mined conversations primarily use ChatGPT for knowledge acquisition and code generation, with most tasks situated at the detailed design level.
Yi-Fei Wang, Rui-Yin Li, Peng Liang et al.
· arXiv.org · 0 citations