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 Inference research benchmarks – Page 3 · SOTA2 Research

Natural Language Inference

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

ChaosNLI S_amb (test)

Model A (projection target)

41.8Top-1 Accuracy

27

Apr 3, 2026

ChaosNLI (test)

Model C (vote-proportion target)

50.5Top-1 Accuracy

27

Apr 3, 2026

73.1ANLI R1 Accuracy

27

Feb 26, 2026

Fine-tuning (full)

93.3Accuracy

27

Feb 26, 2026

SNLI hard 1.0 (test)

BERT-base + PoE w/ Seq-Z

84.48Accuracy

27

Feb 26, 2026

41Accuracy

26

Feb 26, 2026

0.61Accuracy

26

Feb 26, 2026

0.964Accuracy

26

Apr 28, 2026

70.6Accuracy

26

Apr 28, 2026

76.91Accuracy

26

Apr 28, 2026

92.13Accuracy

26

Feb 26, 2026

Individual Task

96.52Accuracy

26

May 29, 2026

44.7Accuracy

26

Feb 26, 2026

44.3Accuracy

26

Feb 26, 2026

89Accuracy

25

Feb 26, 2026

MultiNLI mismatched (cross-domain) RepEval 2017 (test)

75.8Accuracy

25

Feb 26, 2026

0.918Accuracy

24

Feb 26, 2026

82.7Accuracy

24

Feb 26, 2026

87.4MNLI Accuracy

23

May 19, 2026

82.4Accuracy

23

May 27, 2026

100Accuracy

23

Feb 27, 2026

MultiNLI matched (dev)

88.4Accuracy

23

Feb 26, 2026

DeBERTa-V3 base + PiFi

87.98Accuracy

22

Feb 26, 2026

GNLI Human (test)

84.69Accuracy

21

Feb 26, 2026

96.84Accuracy

21

Feb 26, 2026

Page 3 of 15

Popular datasets

MNLI ANLI SNLI XNLI MultiNLI

Follow for updates

Join SOTA2 Contact