Correlation with human evaluation on SEEDA Fluency (E) benchmark leaderboard · SOTA2 Research

Correlation with human evaluation on SEEDA Fluency (E)

0.981Pearson Correlation (r)GPT-4 S (fluency)

Evaluation Results

Method	Links
GPT-4 S (fluency)2025.02		0.981	0.982
GPT-4 S (meaning)2025.02		0.976	0.974
IMPARA2025.02		0.969	0.965
SOME2025.02		0.965	0.965
SOME2025.02		0.943	0.969
IMPARA2025.02		0.9	0.978
GPT-4 E (fluency)2025.02		0.848	0.987
Scribendi2025.02		0.826	0.912
Scribendi2025.02		0.715	0.842
GREEN2025.02		0.252	0.618
GLEU+2025.02		0.232	0.569
GLEU+2025.02		0.102	0.486
GREEN2025.02		-0.135	0.42
PTERRANT2025.02		-0.288	0.279
ERRANT2025.02		-0.374	0.231
PTERRANT2025.02		-0.47	0.231
ERRANT2025.02		-0.507	0.033