Correlation with human evaluation on SEEDA-S Base benchmark leaderboard · SOTA2 Research

Correlation with human evaluation on SEEDA-S Base

0.958Pearson Correlation (r)GPT-4 S (meaning)

Evaluation Results

Method	Links
GPT-4 S (meaning)2025.02		0.958	0.881
IMPARA2025.02		0.939	0.923
SOME2025.02		0.932	0.881
GREEN2025.02		0.925	0.881
IMPARA2025.02		0.916	0.902
GPT-4 S (fluency)2025.02		0.913	0.874
SOME2025.02		0.892	0.867
GLEU+2025.02		0.886	0.902
PTERRANT2025.02		0.87	0.797
GLEU+2025.02		0.863	0.846
GREEN2025.02		0.855	0.846
GPT-4 E (fluency)2025.02		0.844	0.86
ERRANT2025.02		0.763	0.706
PTERRANT2025.02		0.7	0.629
Scribendi2025.02		0.674	0.762
Scribendi2025.02		0.62	0.636
ERRANT2025.02		0.545	0.343