I am a first year PhD student based in Brno, focusing on explainable NLP and information retrieval. I work on extending the ColBERT model to highlight relevant parts of a document while retrieving it, so users can immediately see why a result was selected. I also developed a constrained decoding method that inserts tags directly into the text to extract fine grained evidence. In addition, I am involved in research on factual claim extraction for fact checking.
In my free time, I enjoy adrenaline sports, bouldering, and cooking.
project
An extension of ColBERT that scores each document token for relevance during retrieval, using span supervision distilled from Gemma 2 on MS MARCO, so you get evidence-style highlights without a second LLM call.
event
I've always wanted to organize a hackathon at our university, so I finally made it happen. Students built baseline solutions for our fact-checking pipeline.
They had fun, food, and opportunity. At the end, everything went off without major hiccups.
note
We hosted a group from the University of Taipei as part of our collaboration on the FactDeMice project, focusing on fact-checking, disinformation detection, and fake review identification.
We also introduced them to researchers working on fact-checking across the Czech Republic. Beyond research, it was great getting to know each other's cultures; they seemed to enjoy our beer!