Table 4.

Topic-level correct matching responses and accuracy consensus across compared models.

Topic	GPTa-3.5 vs Bard			Bard vs GPT-4			GPT-3.5 vs GPT-4			Bard, GPT-3.5, and GPT-4
	Total correct matching responses, n	Accuracy consensus	Total correct matching responses, n		Accuracy consensus	Total correct matching responses, n		Accuracy consensus	Total correct matching responses, n		Accuracy consensus
Biotechnology	3	0.75	3		0.60	4		0.80b	3		0.75
Evolution and health	3	0.75b	3		0.50	3		0.75b	3		0.75b
Population and ecology	2	0.67	2		0.67	3		1.00b	2		1.00
Biomolecules	0	N/Ac	0		N/A	0		N/A	0		N/A
Cell biology and genetics	3	0.30	3		0.43b	4		0.36b	3		0.43b
Ecosystem and environmental issues	1	0.33	2		0.67b	1		0.33	1		0.50
Plant kingdom	2	0.22	4		0.31b	3		0.30	2		0.33
Animal kingdom	3	0.38	5		0.50b	5		0.50b	3		0.43
Physical chemistry	2	0.67	3		0.50	4		0.67b	2		0.67
Organic chemistry	1	0.50	3		0.75b	1		0.33	1		1.00
Inorganic chemistry	1	0.13	1		0.25	3		0.43b	1		0.25
Mechanics	2	0.50	3		1.00	4		1.00b	2		1.00
Heat and thermodynamics	0	N/A	0		N/A	0		N/A	0		N/A
Electrostatics and electricity	3	0.60	5		1.00	6		1.00b	3		1.00
Optics	0	N/A	0		N/A	1		1.00b	0		N/A
Simple harmonic motion and waves	0	N/A	0		N/A	0		N/A	0		N/A
Magnetism	2	0.50	2		1.00b	2		1.00b	2		1.00b
Modern physics and electronics	1	1.00	3		1.00b	1		1.00	1		1.00

^aGPT: Generative Pre-trained Transformers.

^bHighest combination of accurate responses and accuracy consensus in a topic.

^cN/A: not applicable.