general
fact
bullish
A multimodal EEG foundation model combining Mamba-based raw signal encoder, ViT-style encoder for time-frequency data, and text encoder can create rich seizure-relevant representations without labeled data
Artificial Intelligence28 Jul 2026