Table 2 Extraction Accuracy by LLM (n = 50; prompt C).

From: Structured dataset of reported cloud seeding activities in the United States (2000–2025) using an LLM

LLM

Accuracy (%)

95% Confidence Interval

gpt-4.1

93.33%

[86.11%, 97.48%]

gpt-4.1-mini

93.33%

[86.11%, 97.48%]

gpt-4o-mini

91.67%

[84.03%, 96.34%]

o3

96.33%

[88.08%, 98.67%]

o4-mini

95.00%

[88.08%, 98.67%]

Back to article page