Beyond surprisal: GPT-derived attention metrics offer additional explanatory power in predicting the N400 during naturalistic reading

Sven Terpstra, Willem Zuidema, Marianne De Heer Kloots

Published: 03 Oct 2025, Last Modified: 03 Oct 2025CPL 2025 SpotlightPosterEveryoneRevisionsBibTeXCC BY 4.0
Keywords: language models, surprisal, attention, N400, EEG, reading
TL;DR: GPT-derived attention metrics offer additional explanatory power in predicting the N400 during naturalistic reading
Submission Number: 30