Overall, these zero shot methods do not perform nearly as well humans nor models tuned on the training data from EMOTIC (Emotions in Context). But there appears to be promise in linking vision with language to account for both affective and cognitive empathy

→ View original post on X — @petitegeek,