Fig. 3: Task and subtask distributions of models used in model selection.
From: Using large language models as a scalable mental status evaluation technique
BERT models were pretrained on an emotion recognition task before the current study training.
