Mario Giulianelli

Mario Giulianelli is an Associate Professor of Computational Linguistics at University College London (UCL). Prior to this position, he was a senior research scientist at the UK AI Security Institute working on AI evaluation science, and a postdoctoral fellow at ETH Zurich in the Institute for Machine Learning, Department of Computer Science. He completed his PhD at the Institute for Logic, Language and Computation of the University of Amsterdam under Raquel Fernández's supervision. He maintains affiliations as an associated researcher at the ETH AI Center and as a member of the ELLIS Society.

His educational background includes a PhD from the University of Amsterdam, Master's degree in Artificial Intelligence, and undergraduate studies in Computational Linguistics at the University of Tübingen. His research explores computational principles underlying language understanding, production, learning, and use in both natural and artificial cognitive systems. He has made significant contributions to computational psycholinguistics, semantics and pragmatics, language variation and change modeling, and AI evaluation science. Notably, his 2018 paper has been described as introducing the first mechanistic interpretability method for language models.

Giulianelli's recent publications demonstrate a strong focus on advancing AI evaluation methodologies while maintaining deep connections to computational linguistics foundations. His work bridges theoretical linguistics with practical AI development, particularly in language model evaluation, information theory applications to discourse analysis, and cognitive modeling of language processing. The consistent publication of high-impact work at top venues (including ICML Spotlight and ACL panel selections) reflects the significance of his contributions to both NLP and AI safety communities.

Notable achievements include:

• ICML Spotlight paper (top 2.6%) on token-to-character language model conversion
• ACL panel discussion selection (top 0.8%) for research on inductive biases in language models
• Best Paper Award at EMNLP 2018 workshop on neural network interpretability
• Competitive ETH Zürich postdoctoral fellowship
• ELLIS Society membership nomination

Giulianelli actively contributes to the research community through mentoring (including SPAR program involvement), organizing workshops like GenBench, and extensive collaboration across institutions. His invited talks at leading universities worldwide demonstrate the international recognition of his work at the intersection of computational linguistics and AI evaluation.