The Artificial Intelligence and Information Retrieval (AIIR) Lab has been named Best of Lab in the SimpleText Track at the Conference and Labs of the Evaluation Forum (CLEF 2025), highlighting the team’s groundbreaking work in using artificial intelligence to simplify and improve access to scientific literature.
The lab’s paper addressed one of science’s most persistent challenges: the difficulty of navigating dense, jargon-heavy research papers. Leveraging large language models (LLMs) such as LLaMA3 and Mistral, the team developed methods to make scientific information more understandable for broader audiences while preserving technical accuracy.
Their study investigated three key tasks in scientific text simplification. For Task 1, the team focused on retrieving passages for inclusion in simplified summaries, using TF-IDF with query expansion from LLaMA3, followed by a bi-encoder, cross-encoder, and LLaMA3-based re-ranking. This system delivered the most effective results across all participating teams. For Task 2, they explored approaches to identifying and explaining difficult concepts, again using LLaMA3 and Mistral. Finally, for Task 3, the researchers applied prompting and fine-tuning strategies to simplify full passages of scientific text.
This year’s recognition is especially significant: the AIIR Lab was named the top lab among 20 international teams, marking their third year of participation in the SimpleText Track.
The achievement reflects the dedication of AIIR Lab’s student researchers, in particular Nick Largey, Reihaneh Maarefdoust, Shea Durgin, and Deiby Wu, whose innovation and hard work were central to the project.
“Being named Best of Lab at CLEF is a tremendous honor,” said the AIIR Lab’s faculty leadership. “It shows that our students are helping lead the way in rethinking how AI can expand access to knowledge and make science more inclusive.”
The full paper can be accessed through Springer: Read here