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 generation research benchmarks – Page 2 · SOTA2 Research

Open-ended generation

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

LLaVA-Bench COCO

85.76Reference Score

11

Apr 10, 2026

TruthfulQA open-ended

51.2BLEU

10

Feb 26, 2026

0.44FDR

9

Feb 26, 2026

0.75FDR

9

Feb 26, 2026

0.35FDR

9

Feb 26, 2026

2.5FDR

9

Feb 26, 2026

4.22FDR (%)

9

Feb 26, 2026

87.8BERTScore

8

Jun 17, 2026

Arena-Hard v2.0

47.8Score

8

May 29, 2026

29.19ROUGE-Lsum

8

Apr 16, 2026

COCO 2014 (val)

8.58Accuracy

8

Feb 26, 2026

NoveltyBench, WildChat, and Narrative-Discourse Average (test)

49Lexical Dominance

7

Jun 2, 2026

11.83Average Score

7

Apr 28, 2026

Open-ended generation tasks (Human Evaluation)

4.4Quality Score

7

Feb 26, 2026

TruthfulQA double info 1.0 (test)

68.4True Score

7

Feb 26, 2026

TruthfulQA single info 1.0 (test)

66.6Truthfulness Score

7

Feb 26, 2026

Social Bias Tasks (test)

Nous-Hermes-2-Mistral-7B-DPO

0.47Disability Status Score

6

May 18, 2026

Depression 500-sample held-out (test)

57Win Ratio

6

May 7, 2026

Political Leaning (Pol.) 500-sample (test)

60Win Ratio

6

May 7, 2026

Affect (Sentiment) 500-sample (test)

72Win Ratio

6

May 7, 2026

mArenaHard 66 languages

76.7Mean Score

6

Mar 13, 2026

mDolly 66 languages

Tiny Aya Global

86.9Mean Performance

6

Mar 13, 2026

TruthfulQA (test)

0.031Steering Effect

4

May 18, 2026

3.4Perplexity

4

Feb 27, 2026

44.91EM

4

Feb 26, 2026

Page 2 of 3

Popular datasets

TruthfulQA LLaVA-Bench NQ AlpacaEval MATH

Follow for updates

Join SOTA2 Contact