"Recommendation systems are technologies that use data analysis to help users discover content that may interest them—such as movies on streaming platforms, products in e-shops, or articles on news websites," says Pavel Kordík, Ph.D., from FIT CTU, co-chair of the conference.
The fact that RecSys 2025 was held in Prague is mainly thanks to a team of researchers from FIT CTU and colleagues from the technology startup Recombee. Both institutions jointly operate the RecombeeLab research laboratory at FIT CTU, which connects academic research with practical applications. "Bringing the world's largest conference on recommendation systems to Prague was a huge challenge and an even greater honor," says doc. Kordík. "I am proud that we have managed to showcase not only the beauty of our city, but also the strength of Czech research and innovation in the field of artificial intelligence," he adds.
Teams of researchers and doctoral students from FIT CTU and RecombeeLab contributed to the conference with several scientific publications devoted to, among other things, scalable search in recommendation systems, probabilistic modeling and uncertainty estimation, shopping cart prediction, and fair group recommendation. In addition, they also participated in the INRA 2025 workshop with an interesting contribution focused on segment-oriented analytics for media houses, based on the practical experience of The Telegraph newspaper.
"RecSys 2025 in Prague reached a historic milestone—the number of contributions in the industrial part of the program more than doubled compared to previous years and was the highest ever for the entire duration of the conference," says Dr. rer. nat. Rodrigo Augusto da Silva Alves. "The strong connection between academic and applied research in the field of recommendation systems confirms that we are on the right track at RecombeeLab and provide our students with unique opportunities for growth," he adds.
As part of the industrial symposia that were part of the conference, experts from leading global companies, including Google DeepMind, discussed topics such as scaling inference of large language models and the future impact of quantum computers on recommendation systems.
The conference was co-chaired by doc. Ing. Pavel Kordík, Ph.D. (FIT CTU) together with prof. Mária Bieliková (Kempelen Institute of Intelligent Technologies) and prof. Markus Schedl (Johannes Kepler Universität Linz). The organizing team also included Dr. rer. nat. Rodrigo Augusto da Silva Alves (assistant professor at the Department of Applied Mathematics at FIT CTU), who led the industrial program at the conference, and Local Chairs Ing. Vojtěch Vančura, Ph.D. (recent graduate of doctoral studies at FIT CTU) and Ing. Petr Kasalický (doctoral student at FIT CTU).
Selected presentations, lectures, and tutorials from the conference are now publicly available online.
Scientific contributions
The FIT CTU team presented a series of original scientific publications covering key topics in current research at the conference:
• The Future is Sparse: Embedding Compression for Scalable Retrieval in Recommender Systems
• Probabilistic Modeling, Learnability and Uncertainty Estimation for Interaction Prediction in Movie Rating Datasets
• Recurrent Autoregressive Linear Model for Next-Basket Recommendation
• SAGEA: Sparse Autoencoder-based Group Embeddings Aggregation for Fairness-Preserving Group Recommendations
• Full-Page Recommender: A Modular Framework for Multi-Carousel Recommendations
FIT CTU and RecombeeLab also participated in the INRA 2025 workshop with a contribution entitled Segment-Aware Analytics for Real-Time Editorial Support in Media Groups: Lessons from The Telegraph, which reflected practical experience from the media industry.
(Translated with DeepL.com)
