NLP Summer School 2026: Three Days Exploring Language, AI and the Future of Intelligent Systems

How quickly is the field of natural language processing changing? In just a few years, we have moved from traditional NLP tasks towards large language models, multimodal AI, AI agents, and systems capable of working with text, images, audio, and external tools.

These developments were at the centre of the NLP Summer School 2026 {https://kinit.sk/event/nlp-school/}, organised by the Kempelen Institute of Intelligent Technologies (KInIT) in Bratislava in September. Over three days, students, PhD candidates, researchers, and AI practitioners came together to discuss the current state of NLP, where the field is heading, and questions that go beyond technology itself.

From large language models to AI agents

The programme opened with one of the most significant developments in modern AI: the shift from standalone models towards agentic systems.

Participants explored large language models, their evaluation, and their use in increasingly complex systems. Marek Šuppa from Comenius University gave a talk, offering a practical perspective on building AI agents and systems capable of planning, using tools, and performing multiple steps to solve a task.

The programme also moved beyond text. Other sessions covered vision-language models, speech and audio processing, LLM fine-tuning, mechanistic interpretability, and model robustness.

For students, this provided an opportunity to see NLP from a much broader perspective than the traditional view of individual tasks and models.

And what about Slovak?

One question naturally emerges with the rise of large language models: How well do they actually understand smaller languages?

Slovak may be a small language in the global AI landscape, but that does not mean it should be left out of technological progress. High-quality datasets, benchmarks, language resources, and models are essential for measuring and improving AI systems' capabilities in Slovak.

This was also the focus of the “Whose model speaks your language?” panel, which brought together different perspectives on Slovak NLP. The discussion featured Radovan Garabík from the Ľ. Štúr Institute of Linguistics of the Slovak Academy of Sciences, Peter Bednár from the Technical University of Košice, and Matúš Pikuliak from KInIT, moderated by Marián Šimko.

Bringing these communities together is important for the development of Slovak language technologies. NLP does not advance through new models alone. It also requires high-quality linguistic data, linguistic expertise, technical knowledge, and people who can critically evaluate what AI systems actually do.

That is also why this discussion matters for the wider Slovak NLP ecosystem, which in recent years has been producing new datasets and benchmarks for Slovak. It was followed up by Andrej Ridzik's (KInIT) talk "Evaluating LLMs" on how to tell whether a new model is actually better. Using the Slovak benchmarks skLEP, SkMTEB and the newly created sk-bench, the talk showed that different types of models need to be evaluated differently, and that a leaderboard number shouldn't be taken at face value.

AI is not only a technical problem

Another part of the programme approached AI from the other direction – from the perspective of people.

As interactions with AI systems become increasingly natural, anthropomorphism is becoming an important topic. People may attribute intentions, emotions, or human-like understanding to AI systems even though the way these models operate is fundamentally different from how humans think.

The panel “AI as a reflection of our minds? Anthropomorphism and ethics of human-AI interaction” therefore brought together technological, cognitive, psychological, and societal perspectives.

Among the participants were Igor Farkaš from Comenius University, whose work includes neural networks, cognitive science, and explainable AI, and Nora Hargaš, who works in consulting, coaching, and education and collaborates with psychologists and psychotherapists.

The discussion highlighted that when thinking about the future of AI, it is no longer enough to ask only what models can do. We also need to consider how people interpret their capabilities, how we interact with them, and what impact their use may have on humans.

More than lectures

The NLP Summer School was also an opportunity for the Slovak AI and NLP community to meet in person.

Alongside lectures and panel discussions, the programme offered opportunities to present projects, talk to researchers, network, and meet informally. A poster session gave PhD students and other early-career researchers the opportunity to present their work and discuss it with fellow participants and experienced researchers.

This community aspect can be just as important as the lectures themselves. For a student only beginning to explore NLP, meeting a researcher or PhD student can be the first step towards discovering a research topic, finding a mentor, or identifying a future opportunity.

What can we take away?

Three days of NLP Summer School showed that modern NLP is much more than text processing.

It is a combination of linguistics, machine learning, multimodal AI, cognitive science, and societal questions. Connecting these perspectives will become increasingly important – especially if we want to build technologies that work well not only for the world's largest languages, but also for smaller languages such as Slovak.

The NLP Summer School 2026 created a space where people at different stages of this journey could meet – from students taking their first steps in NLP to experienced researchers and industry professionals.

And perhaps these kinds of encounters are one of the best ways to keep Slovak NLP moving forward.