FactScore
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
FActScore
59.3FActScore
32
FActScore
84.5Pairwise Score
24
FActScore v1.0
47.3Score
20
FactScore
68.1FactScore
20
FactScore English (test)
71ROC-AUC
20
FactScore English
0.34PR-AUC
20
FActScore
84FActScore
16
FactScore
65.41Precision@1
15
FactScore
0.735AURC
14
FActScore (test)
0.858AUROC
12
FActScore (test)
20.4Number of Facts
12
FactScore (unlabeled)
76.4US (%)
10
FactScore (labeled)
64.8LS Score (%)
10
FactScore
96.1FactScore
9
FactScore
0.076ECE
8
FactScore LLM-based evaluation
86.36Score
6
FactScore 200-word response length constraint
98.9Response Coverage (%)
5
FActScore Context and General truth
80.2FActScore
2
FActScore* Context only
76.9FActScore
2