Understanding EEG Foundation Model Representations

motivation

Foundation models for EEG are getting better at downstream tasks — but we have very little understanding of why. What structure do they actually encode? Do their internal representations correspond to anything cognitively or clinically meaningful, or are they just useful statistical features with no interpretable structure?

This project is an attempt to open up that black box.

approach

[Write about how you’re probing the representations — what probing tasks, what models, what you’re looking for.]

early findings

[Write about early results or interesting observations here.]

open questions

[What are the key open questions you’re still working through?]