Robust ErrP Decoding Under Multisensory Feedback with Varying Congruency

A study on arXiv (2607.24806) investigates error-related potentials (ErrPs) under multisensory feedback with controlled congruency. Researchers used a multi-branch EEGNet-based architecture with auxiliary supervision to improve decoding robustness across visual, auditory, and tactile modalities. Experiments employed a maze-observation task with unimodal, bimodal, and trimodal feedback configurations. Incongruent feedback was found to increase decoding difficulty and reduce classification performance. The proposed approach avoids explicit modality-specific assumptions, aiming to enhance human-machine interaction in realistic settings.