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

LLM-generated text detection research benchmarks – Page 5 · SOTA2 Research

LLM-generated text detection

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

99AUROC

4

Apr 3, 2026

DetectRL Word-level perturbation (OOD)

99.59AUROC

3

May 12, 2026

DetectRL Character-level perturbation (OOD)

99.8AUROC

3

May 12, 2026

DetectRL Yelp_review domain

90.18AUROC

3

May 12, 2026

DetectRL WritingPrompts domain

81.32AUROC

3

May 12, 2026

DetectRL Mixed-domain Source: GPT-3.5-turbo (test)

80.57AUROC

3

May 12, 2026

DetectRL Mixed-domain Source: Claude-instant (test)

80.92AUROC

3

May 12, 2026

DetectRL Mixed-domain Source: Llama-2-70B (test)

94.7AUROC

3

May 12, 2026

Kaggle DAIGT V2 (train)

0.962AUROC

3

May 12, 2026

90TPR @ FPR=1%

3

May 12, 2026

WritingPrompts Llama3-8B

100TPR @ FPR=1%

3

May 12, 2026

SQuAD GPT-Neo-2.7B

53.33TPR @ FPR=1%

3

May 12, 2026

XSum Llama2-13B

80.67TPR@FPR=1%

3

May 12, 2026

WritingPrompts GPT-4.1-mini

31.33TPR @ FPR=1%

3

May 12, 2026

XSum GPT-5-Chat

31.33TPR @ FPR=1%

3

May 12, 2026

HART Writing domain (Leave-One-Domain-Out)

86.2AUROC

3

Apr 7, 2026

HART News domain (Leave-One-Domain-Out)

92.69AUROC

3

Apr 7, 2026

HART Essay domain (Leave-One-Domain-Out)

95.88AUROC

3

Apr 7, 2026

HART Arxiv domain (Leave-One-Domain-Out)

96.61AUROC

3

Apr 7, 2026

Fast-DetectGPT PubMed, XSum, WritingPrompts (test)

98.7AUROC (GPT-3.5T)

2

Feb 26, 2026

Page 5 of 5

Popular datasets

Fast-DetectGPT RAID Xsum DetectRL HC3

Follow for updates

Join SOTA2 Contact