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

Open-ended Text Generation research benchmarks · SOTA2 Research

Open-ended Text Generation

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

0.811COH Score

112

Feb 26, 2026

67.7COH Score

64

Feb 26, 2026

ϵ-sampling + VCM

54.83Rep-2

18

Jun 23, 2026

Law-MT Out of Domain (test)

32.17MAUVE

16

Feb 26, 2026

Scaling Data Store

33.79MAUVE

12

Feb 26, 2026

Typical Sampling

0.96Diversity (DIV)

12

Feb 26, 2026

Wikinews (test)

Typical Sampling

0.95Diversity (DIV)

12

Feb 26, 2026

Wikitext (test)

Typical Sampling

95Diversity (DIV)

12

Feb 26, 2026

80Sentiment Score

12

Feb 26, 2026

98.9Diversity

11

Mar 19, 2026

InferDPT + RANTEXT

94.8Diversity

11

Mar 19, 2026

Wikitext-103 v1

98.7Diversity

11

Mar 19, 2026

98.3Diversity

11

Mar 19, 2026

1.76PPL

10

Feb 26, 2026

2.55PPL

10

Feb 26, 2026

4.88MT-1 Score

8

Jun 4, 2026

Wikitext-103 (test)

84Win Rate

8

Feb 26, 2026

92MAUVE

5

Jun 12, 2026

CTC-trained MDLM

-1.7Score

4

Feb 26, 2026

Creative-Writing-Bench v3

CTC-trained MDLM

27.4Score

4

Feb 26, 2026

Arena-hard Creative-Writing

CTC-trained MDLM

80.2Pairwise Win Rate

4

Feb 26, 2026

Arena-hard Hard-Prompt

58.5Pairwise Win Rate

4

Feb 26, 2026

Chatbot Arena inspired qualitative prompts (val)

1,150.78ELO

4

Feb 26, 2026

76.81BERTScore

3

Feb 26, 2026

BookCorpus Story Project Gutenberg

0.91MAUVE

2

Jun 12, 2026

Page 1 of 2

PreviousNext

Popular datasets

WikiText-103 WikiText Arena-Hard WritingPrompts WikiNews

Follow for updates

Join SOTA2 Contact