P. Llorens Pena

The paper investigates how the brain encodes novel task instructions using fMRI and finds that the MDN uses both abstract and high-dimensional coding to balance generalization and expressivity.

The paper uses EEG and deep neural networks to show that high-level visual predictions, not low-level features, drive neural responses to unexpected stimuli around 200ms after stimulus onset.

This study uses EEG to examine how neural patterns dynamically represent and reconfigure information during novel verbal instruction tasks, revealing changes in dimensionality and abstraction over time.