Table 2.

Accuracy of different prompt techniques against the osteoarthritis benchmark.

	GPT-3.5			GPT-4
	IOa	COTb	P value	IO	COT	P value
Osteoarthritis benchmark	0.16	0.17	.41	0.24	0.23	.52
Guideline item QAc	0.26	0.28	.03	0.38	0.38	.80
Management option QA	0.22	0.20	.004	0.30	0.27	.002
Treatment strategy QA	0.02	0.03	<.001	0.07	0.07	.79
Real-case QA	0.03	0.01	<.001	0.01	0.01	.20

^aInput-output prompt technique.

^bZero-shot chain of thought prompt technique.

^cQA: question-answer.