Ablating Masking Geometry for EEG Foundation Models
September 26, 2026
This study evaluates 58 pre-trained models to determine how spatio-temporal masking strategies affect EEG signal decoding. The research compares masking configurations across MAE and JEPA frameworks using the OpenEEGBench dataset.
HOW THIS AFFECTS YOU
●
researcherYou can select more effective masking strategies based on empirical performance on brain-signal data.
●
healthThis informs the development of more robust foundation models for clinical neuroscience applications.