Stacked LoRA, combining global and subject-specific low-rank adapters on frozen EEG foundation models, yields the best motor-imagery accuracy in most backbone–dataset pairs and shows when each path helps.
Title resolution pending
1 Pith paper cite this work, alongside 2 external citations. Polarity classification is still indexing.
1
Pith paper citing it
2
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Stacked LoRA for Subject-Adaptive EEG Foundation Models in Motor Imagery Decoding
Stacked LoRA, combining global and subject-specific low-rank adapters on frozen EEG foundation models, yields the best motor-imagery accuracy in most backbone–dataset pairs and shows when each path helps.