Trial lecture
Methods in LLM mechanistic interpretability
Time and place: April 17, 2026 11:15 AM. Lille Auditorium, Kristen Nygaards hus
Main research findings
Our ability to understand and use language is not limited to expressions we are already familiar with. Instead, we continually create novel expressions by systematically combining words and phrases from a finite vocabulary. This ability is often described by the principle of semantic compositionality, which defines the meaning of a compound expression as determined by the meaning of the expression’s parts and their syntactic composition.
Today, a large portion of all text is generated by language models built with artificial neural networks. Unlike the principle of compositionality, these networks do not rely on an explicit definition of meaning composition, but on general-purpose machine learning techniques.
In this thesis, we evaluate several types of neural networks for both graph classification and sequence modeling, targeting productive and systematic properties of generalization. We achieve this through enforcing different constraints on the data distribution, introducing several evaluation frameworks. Our results show that most networks lack productive and systematic generalization. We further demonstrate the connection between systematic generalization and the information entropy of the training distribution, and we present novel ways of grounding the evaluation of neural networks in formal definitions of compositionality. Our findings advocate for a more principled approach to evaluation in NLP.
Adjudication committee
- Professor Marco Baroni, Pompeu Fabra University, Barcelona, Spain
- Professor Paola Merlo, University of Geneva, Geneva, Switzerland
- Associate Professor Kai Olav Ellefsen, Department of Informatics, University of Oslo, Norway
Supervisors
Professor Lilja Øvrelid, Department of Informatics, University of Oslo, Norway
Professor Eik Velldal, Department of Informatics, University of Oslo, Norway
Dr. Étienne Simon, Department of Informatics, University of Oslo, Norway
Chair of defence
- Professor Ingrid Chieh Yu, Department of Informatics, University of Oslo
Contact information at Department: Pernille Adine Nordby