Brain-Aligned Multimodal LLMs for Social and Emotional Understanding of Cartoon Movies

Explain Like I'm Normal

Researchers used fMRI data to compare how human brains and Multimodal LLMs process social and emotional scenes in cartoon movies. The study found that while models show similar temporal patterns to human 'Theory of Mind' responses, their internal rating distributions still differ significantly from biological neural activity. This helps bridge the gap between artificial neural networks and actual human affective processing.