What's
Oct 10. 2025
KInIT team at the 2025 ACL conference
During the summer, the 63rd Annual Meeting of the Association of Computational Linguistics (ACL) took place in Vienna. From July 27 to August 1, the natural language processing community gathered for a leading conference full of paper presentations, workshops and tutorials, alongside a programme packed with interesting discussions, networking, and social events.
We are very excited that the KInIT team members were a part of this year’s edition and enriched the conference by presenting their work and projects. Now, we want to share with you the highlights of the conference and KInIT’s experience of this amazing event!
KInIT papers
At the conference, four of KInIT’s papers were accepted and presented by our team. They come from the many inspiring minds of our institute and our partnered institutions, and investigate the edge-cutting topics from the field of natural language processing. Delve into the research and download the papers, and check out the details from our team about how they were received at the conference.
“The massive attendance at the conference resulted in full hands of work to present the results of our research. It was very interesting to observe that even three co-authors needed to present the work in parallel to satisfy the interest.”
Dominik Macko, Researcher at KInIT
“It was wonderful to see the interest in the paper presented by Daniil Gurgurov, a student from DFKI, where I contributed as a co-author. The engaging discussions around low-resource settings and methods to adapt models to support languages beyond English were both inspiring and motivating. They reinforced the importance of developing multilingual approaches and encouraged me to continue exploring this direction further.”
Ivan Vykopal, PhD Student at KInIT
Slavic NLP workshop
“I must say, it was a very cool experience being surrounded by so many influential figures in NLP – seeing people like Prof. Christopher Manning, Sarah Hooker from Cohere, or Daniel Jurafsky all in one place. And honestly, I’m still amazed at how they managed to pack such an incredible amount of content into just three days. With countless sessions happening simultaneously, choosing where to go next became genuinely challenging – it was always one room against another. A good strategy seemed like focusing on talks that either introduced something novel, sparked genuine excitement, or connected with my own research interests – of course.”
“Another interesting observation was that initially, based on the program and session titles, most research was simply leveraging popular LLMs to build application agents. However, once I started attending the actual talks and delving into the details, I discovered that there was plenty of work that went far beyond that, and it was a refreshing discovery for me.
The papers that resonate with me most tend to be the creative ones presented with fun and personality. Don’t take this the wrong way, research is a serious job, and a very important one, but there’s something special about having a brilliant insight, enjoying the process of exploring it, and then entertainingly engaging your audience while sharing your findings (whether groundbreaking or not). That’s what it is all about! And I found papers like that as well.”
Jaroslav Kopčan, Researcher at KInIT
Highlights
“The highlight for me was the generalisation panel hosted by Eduard Hovy, with panellists Mirella Lapata, Dan Roth, and Yue Zhang. The panel was about ensuring that models behave robustly, reliably, and fairly when making predictions on data different from their training data. Achieving such a good generalisation is critically important for models used in real-world applications, as they should emulate human-like behaviour. We really need more people working on understanding the LLMs, how they work, and why they do not work, to improve them for the future and research. Without that, the NLP field as a whole will stagnate, and the currently already existing replicability crisis will get only worse.”
Branislav Pecher, Researcher at KInIT
“One of the highlights of the conference for me was the opportunity to network with new researchers and finally meet in person the collaborators and project partners we had only met online until now. Conferences play an important role not only in building new connections but also in strengthening existing collaborations, which can often lead to new joint projects and papers between our institutes. It was a pleasure to meet colleagues from DFKI, TU Berlin and to connect in person with Antonia Karamolegkou from the University of Copenhagen.”
Ivan Vykopal, Researcher at KInIT
“I would single out tutorials taking place before the conference proper, particularly the one on measuring uncertainty in LLMs. When was the last time an LLM replied to you something along the lines of “I don’t know”? In such situations, LLMs instead tend to generate incorrect answers to questions or text that does not make sense to humans (in literature, this is referred to as hallucination). The tutorial provided an extensive overview of how to measure the uncertainty of LLMs. Knowing uncertainty can be useful for improving the robustness of LLMs and reducing occurrences of hallucination.”
Kamil Burda, Researcher at KInIT
And that’s a wrap on this year’s ACL. For our institute, it was an opportunity to truly establish ourselves among the best research institutions in the field of natural language processing. We are proud to be a part of such a prestigious event, and we look forward to the next edition!
