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

Harmfulness Detection research benchmarks · SOTA2 Research

Harmfulness Detection

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

BeaverTails bottom 30% uncertainty slice (test)

85.1AUROC

70

May 21, 2026

BeaverTails, ToxicChat, PKU-SafeRLHF Hard (test)

MultiLayer-Linear

90.2AUROC

63

May 21, 2026

Full in-distribution (test)

MultiLayer-Linear

0.964AUROC

63

May 21, 2026

OpenAI Moderation

92.9Macro F1 Score

59

Jun 23, 2026

90.6Macro F1 Score

47

Jun 1, 2026

89.78Macro F1

25

Jun 2, 2026

83.4Macro F1

8

Apr 21, 2026

Popular datasets

Aegis WildGuard BeaverTails In-distribution OpenAI Moderation

Follow for updates

Join SOTA2 Contact