KInIT presents the capabilities of Large Language Models (LLMs) in the context of tackling online disinformation in EBU webinar series no. 4

 

On 20 January 2025, the European Broadcasting Union (EBU) hosted the 4th edition of the webinar series in which selected vera.ai work and outcomes are presented. This time, it was the turn of researchers based at Kempelen Institute of Intelligent Technologies (KInIT).

During this first webinar of 2025, Ivan SrbaAneta Žugecová and Domnik Macko from the Kempelen Institute of Intelligent Technologies (KInIT) discussed their investigations of the capabilities of Large Language Models (LLMs) in the context of tackling online disinformation.

LLMs have recently revolutionized many areas, and tackling online disinformation is no exception. In this webinar, following their own research results, KInIT dived into both positive and negative aspects of LLMs.

First, it was presented how the emergence of LLMs has heightened concerns about automatic generation and spread of disinformation, and how KinIT has identified and analysed vulnerabilities of LLMs that can be potentially misused to generate false news articles. A specific focus was given to their ability to personalize generated content towards specific target audiences. KInIT then focused on how such generated text can be detected in highly multilingual settings.

Second, LLMs exhibit a great potential to automate complex time-consuming steps of the disinformation combatting process. KInIT demonstrated this potential on cross-lingual retrieval of fact-checking articles, as well as data augmentation.

These are just some snapshots of what was presented. Below, you can view a recording of the entire talk, including corresponding slides.

Upcoming webinar no. 5

Did this raise appetite for more? You can view previous recordings here, on the vera.ai website, and also look forward to the next session in the series. It will take place on 11 February at 11:00 CET when the series continues with its 5th webinar. Then, we will showcase CERTH’s work on an AI-powered keyframe selection and enhancement service for media verification.

Register here (compulsory) and mark your calendars accordingly - we would be delighted if you joined us in February!

Author: Lalya Gaye (EBU)

Editor: Jochen Spangenberg (DW)

vera.ai is co-funded by the European Commission under grant agreement ID 101070093, and the UK and Swiss authorities. This website reflects the views of the vera.ai consortium and respective contributors. The EU cannot be held responsible for any use which may be made of the information contained herein.