notesum.ai

Published at November 26

Synthetic Data Generation with LLM for Improved Depression Prediction

cs.LG

Released Date: November 26, 2024

Authors: Andrea Kang¹, Jun Yu Chen¹, Zoe Lee-Youngzie¹, Shuhao Fu¹

Aff.: ¹University of California, Los Angeles

Arxiv: http://arxiv.org/abs/2411.17672v1

Refer to caption

Model	Training Data	Input Format	# of Params	Test RMSE	Test MAE
GPT-4o (OpenAI 2024)	None	Synopsis	-	5.79	4.50
Random Forest (Sun et al. 2017)	Train	Selected-Text	-	4.98	3.87
Dual Encoder (Lau, Zhu, and Chan 2023)	Train	Transcript	234.6M	4.67	3.80
BERT	Train	Synopsis	109.5M	5.59	4.71
BERT	Synthetic	Synopsis	109.5M	4.80	4.06
BERT	Train + Synthetic	Synopsis	109.5M	4.64	3.66