NeurosciencebioRxivPreprint
Brain-like dynamics in speech representations can emerge through self-supervised learning
Oli Danyi Liu, Hao Tang, Naomi H. Feldman, Sharon Goldwater
This study shows self-supervised speech models can develop brain-like temporal dynamics and abstraction of speech variability without lexical feedback, even when trained on non-speech audio.
DOI 10.64898/2026.01.16.700011