Improving Queries to Retrieval-Augmented Generation AI Tools by Inferring the Sources of Bad Results
Document Type
Presentation
Original Presentation Date
2025
Date of Submission
April 2025
Abstract
Since ChatGPT’s wide release, businesses have rushed into expanding their use of generative artificial intelligence. Information providers, including suppliers of library databases and integrated library systems, are joining the trend, with many products combining the verifiability of direct source access and citation with the conversational answers produced through large language models. Retrieval-augmented generation is a common technique that overlays search results with text extracted and processed through generative artificial intelligence. This poster provides an overview of retrieval-augmented generation and summarizes ways that it can produce errors. Strategies for improving results, based on those error modes are suggested.
Is Part Of
VCU Libraries Faculty and Staff Presentations
Recommended Citation
Arendt, J. (2025). Improving queries to retrieval-augmented generation AI Tools by inferring the sources of bad results [Poster presentation]. Association of College and Research Libraries Conference, Minneapolis, MN, United States.
Comments
Poster presented at the Association of College and Research Libraries Conference, 2025 in Minneapolis, MN