Cross-sensory combined EEG training effectiveness under data scarcity

Establish whether training Brain Passage Retrieval models on combined electroencephalography datasets spanning auditory and visual modalities improves retrieval performance relative to modality-specific training in the presence of limited EEG data availability.

Background

EEG data collection is resource-intensive and scarce, which limits the training of neural retrieval models. Combining datasets across different sensory modalities (auditory and visual) could potentially enhance model performance and address data scarcity, but this has not been empirically validated.

Demonstrating the effectiveness of cross-sensory training would inform how to leverage diverse EEG corpora collected under different protocols to improve BPR and broaden its applicability.

References

Second, whether training on diverse EEG datasets from different sensory modalities improves retrieval performance remains unexplored, which would be valuable given the severe scarcity of EEG training data.

— Auditory Brain Passage Retrieval: Cross-Sensory EEG Training for Neural Information Retrieval  (2601.14001 - McGuire et al., 20 Jan 2026) in Introduction, Section 1 (page 2)