The Center for Reasoning, Normativity and AI (CERNAI) at the University of Pavia is pleased to announce the next talk in the CERNAI Research Seminars, GENERIC Series. Speaker: Vittoria Dentella (University of Pavia)

Title: The shape of AI: From grammar to cognition

Date and time: October 21, 2026, 15:30-17:00

Venue: Sala Lauree, Department of Law, University of Pavia, Corso Strada Nuova 65, Pavia, Italy

Online: a Zoom link is available upon request

Abstract LLMs’ progressive gains in performance across Natural Language Processing applications and tasks has sparked an interest towards characterizing the underlying knowledge that they encode and which allows for such performance to surface. Such knowledge is often described using vocabulary that refers to human-like linguistic capabilities. More specifically, LLMs have been attributed with human-like linguistic competence both over grammar and over form-meaning mappings. However, a closer look at LLMs’ performance in language tasks reveals fundamental shortcomings which cast doubt on the human-likeness of the knowledge that they encode. In this talk, I present evidence that challenges the attribution of human-like competence to LLMs, and I discuss the importance of adopting high standards for LLM evaluations.

Everyone is welcome. For more information, or to request the Zoom link, visit generic.unipv.it or write to generic@unipv.it.

Luogo

Pavia