Jitka Mertlová, a first-year master's student working in the G²OAT research group, presented her scientific paper at the prestigious international conference AAAI 2025 (AAAI Conference on Artificial Intelligence). The conference took place in Philadelphia, USA, and is considered one of the most significant conferences in the field of artificial intelligence.
Jitka Mertlová's publication, along with other co-authors, stood out among nearly 17,000 scientific papers, with only around 3,000 submissions selected for the main part of the conference after a rigorous peer-review process.
The presented paper, "Distances Between Top-Truncated Elections of Different Sizes," explores the relationships and measures the differences between instances of elections of varying sizes. The paper presents the results in the form of electoral maps, which serve as a guide for assembling datasets for experimental verification of hypotheses in the field of electoral theory. This work significantly expands the understanding of both real and synthetic electoral data.
"It was ten days packed with experiences to the brim. Of course, it was great to present my research to people from a similar field I hadn’t known before. Hearing the perspective and questions from someone with a fresh outlook is incredibly valuable and pushes you forward. It was almost unbelievable to meet people I had only (often) quoted before, and at AAAI, I had the opportunity to talk to them or attend their lectures for the first time,” says Jitka.
The paper builds on Jitka's bachelor's thesis titled "Maps of Elections" for which she was awarded the Dean's Award in the winter semester of 2024/2025. Her work is the result of international collaboration among researchers from AGH University (Piotr Faliszewski), University of Oxford (Tomasz Wąs), PSL Dauphine (Stanisław Szufa), and ENS Rennes (Pierre Nunn).
In addition to Jitka Mertlová's paper, several other papers from the G²OAT research group were presented at the conference. In the two papers "Exact Algorithms for Multiagent Path Finding with Communication Constraints on Tree-Like Structures" and "Solving Multiagent Path Finding on Highly Centralized Networks," the authors Foivos Fioravantes, Dušan Knop, Jan Matyáš Křišťan, Nikolaos Melissinos, Michal Opler, and Tung Anh Vu (MFF UK) expanded last year's paper on parameterized algorithms for multi-agent pathfinding by adding further restrictions on the structure of the graph on which agents can move.
Another paper, "Balanced and Fair Partitioning of Friends," with authors Argyrios Deligkas (RHUL), Eduard Eiben (RHUL), Stavros D. Ioannidis (RHUL), Dušan Knop, and Šimon Schierreich, studies algorithms for the fair partitioning of agents into groups of equal size. This problem underlies many real-world applications, from assigning students to team projects, to allocating employees into educational groups, to data clustering.
The final presented paper, titled "Exact Algorithms and Lower Bounds for Forming Coalitions of Constrained Maximum Size," with authors Foivos Fioravantes, Harmender Gahlawat (Grenoble INP), and Nikolaos Melissinos, investigates the conditions under which agents can be divided into groups of constrained sizes such that no agent can improve their satisfaction by changing their group.
