Table 2 Benchmarks for sentiment... | Emerald Publishing

Table 2

Benchmarks for sentiment classification of different GPT models

(a) GPT-3.5 using title and first paragraph
	Bond	Stock	Crypto
weighted-F1	0.920	0.364	0.977
micro-F1	0.899	0.299	0.980
macro-F1	0.340	0.269	0.745

(b) GPT-4 using title and first paragraph
	Bond	Stock	Crypto
weighted-F1	0.944	0.724	0.994
micro-F1	0.916	0.586	0.996
macro-F1	0.364	0.389	0.749

(c) GPT-4o using title and first paragraph
	Bond	Stock	Crypto
weighted-F1	0.967	0.823	0.995
micro-F1	0.962	0.721	0.999
macro-F1	0.422	0.424	0.750

(d) GPT-4o using full text
	Bond	Stock	Crypto
weighted-F1	0.950	0.766	0.995
micro-F1	0.938	0.667	0.999
macro-F1	0.355	0.401	0.750