cross-modality

This research paper explores emotion-sensitive neurons in multimodal foundation models, revealing shared affective mechanisms between speech and facial emotion recognition through causal interventions and cross-modal analysis.