Madhav Agarwal

Vision-language models struggle with emotion recognition due to data bias and temporal limitations

A new paper explores why vision-language models (VLMs) struggle with recognizing human emotions, despite advancements in AI. The research identifies two key vulnerabilities: the long-tailed nature of emotion datasets, w…