Requirements Satisfiability Research Honored With Challenge Award at IEEE International Requirements Engineering Conference

Researchers in the Software and Societal Systems Department (S3D) received a Challenge Award for their work at the recent IEEE International Requirements Engineering Conference in Reykjavik, Iceland.

The paper, “Requirements Satisfiability With In-Context Learning,” builds on advances in large language models (LLMs) to show how authoritative design guidance and indicative requirements can be used to generate and evaluate informal software specifications while collecting design rationale using few-shot learning. Sarah Santos, a Ph.D. student in software engineering, presented the paper at the conference on June 26. Additional authors on the paper include Travis Breaux, an associate professor in S3D; Thomas Norton, the executive director of the Center on Law and Information Policy at Fordham University’s School of Law; and Sara Haghighi and Sepideh Ghanavati from the School of Computing and Information Science at the University of Maine.

The IEEE International Requirements Engineering Conference is among the largest software engineering conferences. The Challenge Award goes to one paper at the conference voted most likely to reshape the frontiers of the field. The paper was also selected for publication in a journal based on the conference. Only 5% of all submissions to the conference are selected for publication.

The researchers believe their experimental designs are adaptable to other requirements engineering problems that use LLMs. Their tools and results are publicly available and built on open-source frameworks using low-cost, cloud-based APIs. They believe companies can adopt these techniques similar to how many companies are using LLMs for a variety of other knowledge- and reasoning-intensive tasks.