Welcome
I am Samuele Bortolotti, a Ph.D. candidate in Computer Science at the University of Trento, working with the Structured Machine Learning Group under the supervision of Stefano Teso and Andrea Passerini.
I work on machine learning, trustworthy AI, and neuro-symbolic integration. I study how machine learning models learn from data, and where and why they fail — for example, reasoning shortcuts, where a model reaches the right answer for the wrong reasons. I want AI to support people and to be trustworthy, grounded in explicit, inspectable rules written in logic.
In spring 2026, I spent three months at the APRIL Lab, working with Antonio Vergari on constraint acquisition with Large Language Models.
I earned my B.Sc. and M.Sc. in Computer Science, both cum laude, from the University of Trento in 2021 and 2023, respectively.
I am always open to collaboration or a chat, whether you are interested in my research, curious about what I do, or simply want to say hello. Feel free to reach out in English, Italian, or German. My German is still a work in progress, so please bear with me if something gets lost in translation :)
News
- Our paper “Concise and Logically Consistent Conformal Sets for Neuro-Symbolic Concept-Based Models” has been accepted at the HLDM workshop at ECML-PKDD 2026!
- New preprint on arXiv: “Concise and Logically Consistent Conformal Sets for Neuro-Symbolic Concept-Based Models”!
- I gave an invited talk, “How to Learn and Not Learn Concepts, Symbols, and Representations”, at the University of Edinburgh, together with Emanuele Marconato.
- I presented “Reasoning Shortcuts in Neuro-Symbolic AI” at a workshop at the University of Edinburgh.
- I started a three-month research visit to the APRIL Lab at the University of Edinburgh, working with Antonio Vergari on constraint acquisition with Large Language Models.