In my HRI lab, the main issue that we have been trying to tackle is precision over many categories and contextual emotion recognition. Gpt-4o needs to be benchmarked properly to compare it to current research (paper forthcoming).

→ View original post on X — @petitegeek