Phoneme- vs. Character-Level Targets and Selective State-Space Models for Intracortical Brain-to-Text | ArxivCSExplorer