Answers are grounded in this paper and cite their sources.
The rapid advancement of large language models (LLMs) has redefined artificial intelligence (AI), pushing the boundaries of AI research and enabling unbounded possibilities for both academia and industry. However, LLM development faces increasingly complex challenges throughout its lifecycle, while existing research has not systematically investigated these challenges and solutions from the perspective of software engineering (SE) approaches. To fill the gap, we systematically analyze the research status throughout the LLM development lifecycle, which is divided into six phases: requirements engineering, dataset construction, model development and enhancement, testing and evaluation, deployment and operations, and maintenance and evolution. We conclude by identifying the key challenges in each phase and presenting potential research directions to address them. In general, we provide valuable insights from an SE perspective to facilitate future advances in LLM development.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: