Back to bookshelf

Multi-user Speech BCI with Rapid New-User Adaptation | bioRxiv 

science•research paper•View Source
Appreciation
--
Importance
--
Date Added
8.24.26
TLDR
Multi-participant data using 6 BrainGate participants.
2 Cents
--
Tags
  • No pretraining/encoder, this is a decoder that is trained supervised on multiple labeled datasets.
  • Potential leakage in evaluations because sentences can be repeated in train and eval (e.g., cross-participant sentence leakage for BrainGate data). Unsure how much to trust numbers
  • Basically none of the data is publicly available (T12 and T15 have small slices published for Brain-to-Text ‘24 and ‘25)