Table 3.

Zero-shot and few-shot learning performance comparison of GPT-4 versus BERT and Lattice-LSTM models on Chinese medical entity extraction dataset.

Model	CMeEEa
	Precision	Recall	F1-score
BERTb-base	63.08	64.08	62.11
BERT-wwm	61.5	61.29	61.72
Lattice-LSTMc	46.34	43.60	49.44
Med-BERT	53.33	47.58	60.66
GPT-4 and Zero shot	64.07	68.97	59.82
GPT-4 and Few shot	65.31	64.89	65.73

^aCMeEE: Chinese Medical Entity Extraction.

^bBERT: Bidirectional Encoder Representations from Transformers.

^cLSTM: long short-term memory.