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

Hallucination Detection research benchmarks – Page 4 · SOTA2 Research

Hallucination Detection

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

RAGTruth & HalluRAG Overall Average 1.0 (test)

0.728Average F1 Score

30

Feb 26, 2026

HalluRAG 1.0 (test)

0.727F1 Score

30

Feb 26, 2026

RAGTruth RT-Summ 1.0 (test)

0.697F1 Score

30

Feb 26, 2026

RAGTruth RT-D2T 1.0 (test)

0.738F1 Score

30

Feb 26, 2026

COCO-Caption (val)

77.11A-ROC

30

Feb 26, 2026

M-HalDetect (val)

89.98A-ROC

30

Feb 26, 2026

AMBER sampled 5k

85.99A-ROC

30

Feb 26, 2026

75Accuracy

30

Feb 26, 2026

94.76ROC-AUC

28

Jun 23, 2026

TriviaQA (40% test)

91AUROC

28

May 7, 2026

89.6ROC-AUC

28

Apr 14, 2026

0.798ROC-AUC

28

Apr 14, 2026

Combined: PC + SE

78.95AUROC

28

Mar 24, 2026

69.9AUCs

28

May 12, 2026

0.777ROC

27

Feb 26, 2026

0.834ROC

27

Feb 26, 2026

77.9ROC

27

Feb 26, 2026

85.74AUROC

26

Apr 30, 2026

HaluEval (in-distribution)

99.1Accuracy

25

Jun 30, 2026

Hallucination-detection dataset

neuron-based linear probe

86.3AUROC

24

Jul 2, 2026

F: Ans. Expect.

97AUROC

24

May 19, 2026

TriviaQA n=1000 (test)

0.794AUROC

24

May 7, 2026

PopQA n=1000 (test)

0.895AUROC

24

May 7, 2026

1AUROC

24

May 19, 2026

Summedits (test)

Separate Guardrail 9B

93AUC

24

May 4, 2026

Page 4 of 17

Popular datasets

RAGTruth HaluEval TriviaQA SQuAD CoQA

Follow for updates

Join SOTA2 Contact