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

AI-generated text detection research benchmarks – Page 8 · SOTA2 Research

AI-generated text detection

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

IntelLabs peer-review withheld pre-ChatGPT (test)

100TPR @ 0.1% FPR (GPT-4o)

2

Jul 7, 2026

M4GT-Bench Russian (test)

99.5Precision

2

Feb 26, 2026

M4GT-Bench Italian (test)

98.86Precision

2

Feb 26, 2026

M4GT-Bench German (test)

98.27Precision

2

Feb 26, 2026

M4GT-Bench English (test)

98.02Precision

2

Feb 26, 2026

M4GT-Bench Chinese (test)

93.48Precision

2

Feb 26, 2026

Paraphrasing attack (test)

Detector + Defender

88.98Human Recall

2

Feb 27, 2026

Cross-lingual and adversarial robustness benchmark Perturbation attack

92.9Accuracy

1

May 26, 2026

Cross-lingual and adversarial robustness benchmark Paraphrase attack

91.8Accuracy

1

May 26, 2026

Cross-lingual and adversarial robustness benchmark Mixed attack

92.1Accuracy

1

May 26, 2026

Cross-lingual and adversarial robustness benchmark Normal

93.4Accuracy

1

May 26, 2026

Long-form QA 46K ShareGPT-augmented corpus 1.0 (test)

—

—Detection Accuracy (1% FPR)

0

Feb 26, 2026

Long-form QA 9K pooled generations corpus 1.0 (test)

—

—Accuracy (1% FPR)

0

Feb 26, 2026

Long-form QA 3K generations corpus 1.0 (test)

—

—Detection Acc (1% FPR)

0

Feb 26, 2026

Wikipedia OPT-13B generations

—

—Accuracy (1% FPR)

0

Feb 26, 2026

Wikipedia GPT2-XL generations

—

—Accuracy (1% FPR)

0

Feb 26, 2026

Page 8 of 8

Popular datasets

XSum, SQuAD, WritingPrompts M4GT-Bench Wikipedia MAGE RAID

Follow for updates

Join SOTA2 Contact