On June 30, 2025, Professor Preslav Nakov will lead a seminar at the University of Massachusetts Amherst´s Manning College of Information & Computer Sciences, focusing on the complex landscape of factuality in the era of Large Language Models (LLMs). The event will examine the risks, challenges, and wide-ranging opportunities that LLMs present regarding factual reliability, both as creators of content and as tools for combating misinformation.
Nakov, who chairs the Natural Language Processing department at the Mohamed bin Zayed University of Artificial Intelligence, will present recent advancements in applying LLMs for fact-checking and for detecting machine-generated text. He will introduce a suite of new tools developed at MBZUAI that tackle key issues in factuality: LM-Polygraph for predicting LLM output uncertainty, Factcheck-Bench as a comprehensive evaluation and benchmarking framework, Loki as a speed- and quality-optimized open-source fact-checker, OpenFactCheck for both customized fact-checking system building and benchmarking, LLM-DetectAIve for identifying machine-generated text, and FRAPPE, an explorer for framing, persuasion, and propaganda techniques. These instruments collectively aim to detect and mitigate the spread of misinformation and empower critical evaluation of LLM outputs.
The seminar will also highlight Nakov´s extensive background in Artificial Intelligence, including his work leading the development of top open-source LLMs in Arabic, Hindi, and Kazakh, and contributing to the Tanbih project with MIT, which is designed to counteract ´fake news´ and media bias. Nakov´s distinguished career features leadership roles in numerous computational linguistics organizations and editorial boards, as well as significant recognitions like the John Atanasoff award and several best paper prizes. Open to attendees in a hybrid format, the seminar offers a timely exploration of the intersection of factuality and rapidly evolving Artificial Intelligence language technologies, providing participants with insights into state-of-the-art solutions for ensuring reliable, fact-based content in the digital age. The session will be hosted by Marco Serafini at the Computer Science Building, room 150.
