Sebastian Stober

This paper benchmarks five positional encoding strategies for transformer-based EEG foundation models, concluding that the optimal encoding is task-dependent and no single strategy is universally superior.