Jack Cheverton and Oliwia Majtyka

Abstract:

The TREC 2025 DRAGUN Track is an open data competition where participants develop a system to analyze a given a news article and generate a list of questions that a reader should investigate if trying to evaluate the credibility of the article. Additionally, the system should generate a report that gives more background and context for the article, allowing for readers to assess how trustworthy the article is. Our approach to this task used a combination of Large Language Models (LLMs) and query expansion to generate questions and retrieve text from a large data collection to help answer these questions.