Hi there! I’m Mathias.
I am currently pursuing a PhD in Machine Learning at the University of Oxford as part of the EPSRC CDT in Autonomous Intelligent Machines and Systems and the Department of Computer Science. I am grateful to be supervised by Professor Alessandro Abate.
My research focuses on instruction following and safety in reinforcement learning (RL). That is, I am interested in how RL agents can learn to follow high-level instructions (e.g. natural or formal language commands) while ensuring that they operate safely and effectively in complex environments. This is a crucial aspect of developing autonomous systems that can interact with humans and the world in a reliable manner.
Previously, I have worked on learning decision trees for explainable and effective controller representation. I also have an interest in logic and knowledge representation, and have worked on integrating formal logical knowledge with learning-based reasoning approaches (e.g. knowledge graph embeddings or graph neural networks). For more details, see my publications.
Education
PhD in Machine Learning (AIMS CDT), 2022-2026, University of Oxford
MSc in Computer Science, 2020-2022, University of Oxford
BSc in Informatics, 2016-2020, Technical University of Munich
News
- [06/2025] Our paper Zero-Shot Instruction Following in RL via Structured LTL Representations has been accepted at the ICML 2025 Workshop on Programmatic Representations for Agent Learning! Many thanks to my co-authors Mattia Giuri and Alessandro Abate.
- [06/2025] I joined QuantCo as a Deep Learning Research Intern for the summer of 2025. I am excited to work on improving LLM agents for complex workflows.
- [02/2025] Our paper DeepLTL: Learning to Efficiently Satisfy Complex LTL Specifications for Multi-Task RL has been accepted as an oral presentation at ICLR 2025! Check out the website here.
- [01/2025] I am excited to join St Hugh’s College, Oxford as a Stipendiary Lecturer in Computer Science for Hilary Term 2025.