WebNLG
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
WebNLG-en (test)
56.3ROUGE-2
5
WebNLG (Human Evaluation)
6.71Fluency
5
WebNLG Constrained 2.0 (test)
61.01BLEU
5
WebNLG (test)
52.8F1 Score
5
WebNLG (val)
52.8F1 Score
5
WebNLG whole span annotation (test)
91.9Precision
5
WebNLG
84.8Precision
4
WebNLG partial match (test)
0.98Micro-F1
4
WebNLG (val)
39.1BLEU
4
WebNLG
0.963Micro-F1
3
WebNLG (seen)
5.1Avg Worker Score
3
WebNLG (test)
52.27BLEU Score
2
WebNLG en (test)
55.4ROUGE-2
2
WebNLG Unseen
2.33Naturalness
2
WebNLG Seen
2.39Naturalness
2
WebNLG U (test)
29Fluency Win Rate
2
WebNLG ru GEM (test)
—ROUGE-2
0