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

Prompt Harmfulness Detection research benchmarks · SOTA2 Research

Prompt Harmfulness Detection

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

AegisSafety (test)

99.5F1 Score

41

Jun 1, 2026

Text & Image Benchmarks Average

GuardReasoner-Omni 2B

83.84F1 Score

19

Feb 26, 2026

Combined Prompt Harmfulness Suite (ToxicChat, HarmBench, OpenAI Mod, Aegis SafetyTest, WildGuard Test)

84.4Macro Avg Harmfulness Detection Rate

17

May 29, 2026

GuardReasoner-Omni 4B

95.5F1 Score

7

Feb 26, 2026

GuardReasoner-Omni 4B

67.86F1 Score

7

Feb 26, 2026

GuardReasoner-Omni 2B

96.82F1 Score

7

Feb 26, 2026

GuardReasoner-Omni 4B

91.67F1 Score

7

Feb 26, 2026

Video Benchmarks Average

GuardReasoner-Omni 4B

94.84F1 Score

5

Feb 26, 2026

GuardReasoner-Omni 4B

99.47F1 Score

5

Feb 26, 2026

Popular datasets

UCF-Crime XD-Violence AegisSafety HarmTextVideo Text & Image Benchmarks

Follow for updates

Join SOTA2 Contact