Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Natural language generation research benchmarks – Page 3 · SOTA2 Research

Natural language generation

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

2.46Naturalness

4

Feb 26, 2026

SGD Seen domains

2.48Naturalness

4

Feb 26, 2026

SGD Unseen domains

2.46Naturalness

4

Feb 26, 2026

SGD 1.0 (overall)

28.6BLEU

4

Feb 26, 2026

SGD 1.0 (unseen domains)

22.2BLEU Score

4

Feb 26, 2026

SGD seen domains 1.0

29.4BLEU

4

Feb 26, 2026

Laptop RNNLG benchmark (test)

Deep Ensemble Model

52.38BLEU

4

Feb 26, 2026

TV RNNLG benchmark (test)

0.529BLEU Score

4

Feb 26, 2026

MSMARCO NLG (val)

PEGASUS-2B (calibrated)

71.06ROUGE-L

3

Feb 26, 2026

3.34Performance Score

2

Jun 15, 2026

52.27BLEU Score

2

Feb 26, 2026

Page 3 of 3

Popular datasets

WebNLG SGD E2E Xsum MATH

Follow for updates

Join SOTA2 Contact