Rethinking Emotion Annotations in the Era of Large Language Models
{{output}}
Modern affective computing systems rely heavily on datasets with human-annotated emotion labels for both training and evaluation. However, human annotations are expensive to obtain, sensitive to study design, and difficult to quality control, because of the su... ...
