Umar Malhi
Senior Front-end Developer
Stony Brook University
Academic library discovery remains fragmented across catalogs, licensed databases, research guides, repositories, special collections, and other systems. Stony Brook University Libraries is exploring whether large language models (LLMs) can provide an intelligent layer that determines user intent and decides where a query should go before retrieval begins: what the user is asking for, the details needed to construct a search, and which systems should receive the query. Using real patron queries from the Libraries’ discovery logs and reference transcripts, with subject librarians providing the final annotations, the team is developing a benchmark to compare commercial and open-weight models. This briefing will share the approach, results, and learnings about the potential and limitations of LLMs for academic library discovery.
