Let EEG Models Learn EEG

Researchers introduced JET, a conditional flow-matching model for synthetic EEG generation.

Unlike denoising-based methods, JET models EEG as continuous raw signal trajectories.

It adds constraints to preserve spectral, temporal, and statistical properties of neural signals.

The paper reports state-of-the-art results on three large benchmarks, with major TS-FID gains.

Better synthetic EEG could help alleviate data scarcity and privacy limits in neuroscience.