This research evaluates how effectively Large Language Models can convert natural language software requirements into formal, machine-readable specifications.

• LLMs demonstrated strong capability in generating accurate software specifications
• GPT-4 significantly outperformed previous approaches, achieving up to 78% accuracy
• Models showed impressive understanding of complex programming concepts and contextual requirements
• Results indicate potential for reducing manual effort in specification engineering

For Engineering teams, this breakthrough could streamline documentation processes, improve bug detection, and enhance automated testing by removing barriers between natural language requirements and formal specifications.

How Effective are Large Language Models in Generating Software Specifications?