From Specifications to Prompts: On the Future of Generative LLMs in Requirements Engineering

TL;DR

Generative LLMs can transform Requirements Engineering by enabling new tasks across elicitation, specification, and validation through natural-language prompts. The paper argues that prompts are themselves requirements artifacts and that human-in-the-loop remains essential for evaluating and guiding model outputs. It outlines the decoder-only LLM architecture, how to formulate RE tasks as prompts, and provides examples such as trace link recovery, rationales, and dialog. It further discusses prompting frameworks like CRISPE and RICE, evaluation challenges with generative and multimodal outputs, and the need for systematic prompt documentation and lifecycle management in RE.

Abstract

Generative LLMs, such as GPT, have the potential to revolutionize Requirements Engineering (RE) by automating tasks in new ways. This column explores the novelties and introduces the importance of precise prompts for effective interactions. Human evaluation and prompt engineering are essential in leveraging LLM capabilities.