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

Correlation with human evaluation on SEEDA E Base

0.974Pearson Correlation (r)GPT-4 S (fluency)

Evaluation Results

Method	Links
GPT-4 S (fluency)2025.02		0.974	0.979
PT-ERRANT2026.02		0.957	0.986
ERRANT2026.02		0.952	0.951
UOT-ERRANT2026.02		0.95	0.979
GLEU2026.02		0.949	0.958
UOT-ERRANT2026.02		0.942	0.993
GREEN2025.02		0.932	0.965
CLEME2026.02		0.932	0.937
PTERRANT2025.02		0.924	0.951
PT-ERRANT2026.02		0.924	0.951
UOT-ERRANT2026.02		0.915	0.93
GLEU+2025.02		0.912	0.944
GREEN2025.02		0.912	0.965
GREEN2026.02		0.912	0.965
GPT-4 S (meaning)2025.02		0.911	0.96
GREEN2026.02		0.91	0.979
CLEME2026.02		0.91	0.93
CLEME2026.02		0.91	0.965
GLEU+2025.02		0.909	0.965
GLEU2026.02		0.909	0.965
GPT-4 E (fluency)2025.02		0.905	0.986
GPT-4-E2026.02		0.905	0.986
IMPARA2025.02		0.902	0.965
SOME2025.02		0.901	0.951
IMPARA2025.02		0.901	0.944
UOT-ERRANT2026.02		0.901	0.909
IMPARA2026.02		0.901	0.944
PT-ERRANT2026.02		0.894	0.888
SOME2025.02		0.893	0.944
SOME2026.02		0.893	0.944
Qwen3-8B-S2026.02		0.886	0.965
ERRANT2025.02		0.881	0.895
ERRANT2026.02		0.881	0.895
ERRANT2026.02		0.864	0.804
GREEN2026.02		0.858	0.93
GLEU2026.02		0.848	0.916
Scribendi2025.02		0.837	0.888
Scribendi2026.02		0.837	0.888
Scribendi2025.02		0.825	0.839
GLEU2026.02		0.808	0.895
PT-ERRANT2026.02		0.802	0.832
PTERRANT2025.02		0.788	0.874
ERRANT2026.02		0.74	0.72
CLEME2026.02		0.716	0.706
GREEN2026.02		0.7	0.825
ERRANT2025.02		0.689	0.643