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

Safety Alignment research benchmarks – Page 2 · SOTA2 Research

Safety Alignment

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

Safety Alignment

0.013Multiplicative Gap (Epsilon)

12

Apr 7, 2026

Safety Generation Open-ended

8.63Refusal Score

11

Jun 9, 2026

Multiple-choice safety evaluation

84.48Refusal Rate

11

Jun 9, 2026

98.2Score

10

Mar 25, 2026

Safety Alignment Dataset 4-order (test)

100DSR

10

Mar 18, 2026

Safety Alignment Dataset 3-order (test)

100DSR

10

Mar 18, 2026

Safety Alignment Dataset 2-order (test)

99.8DSR

10

Mar 18, 2026

Safety Alignment Dataset 1-order (test)

100DSR

10

Mar 18, 2026

PKU-SafeRLHF in-distribution (test)

99.44Accuracy (EN)

10

Feb 26, 2026

Safety BeaverTails, HEx-PHI (test)

95.67BeaverTails Score

10

Feb 26, 2026

PKU-SafeRLHF 10K

261.26HV

8

Jul 3, 2026

2.2ASR

8

Mar 30, 2026

0.4ASR

8

Mar 30, 2026

Chain-of-Thought

83Win Rate

8

Feb 26, 2026

Alpaca-Llama Safety Sets 3

97I-Mali

7

May 22, 2026

CoSApien Public Prosecutor

Llama-3.1-8B-Inst

0.941Safety

6

Jun 1, 2026

Llama-3.1-8B-Inst

99.5Safety Score

6

Jun 1, 2026

CoSApien Game Development

0.975Safety

6

Jun 1, 2026

CoSApien Film Production

Llama-3.1-8B-Inst

0.866Safety Score

6

Jun 1, 2026

CoSApien Arab Publisher

Llama-3.1-8B-Inst

91.6Safety

6

Jun 1, 2026

81.2DSR

6

May 21, 2026

99.06Accuracy

6

May 20, 2026

Jailbreak Bench (JBB)

31JBB Score

6

Apr 7, 2026

Harmful Benchmarks Aggregated (test)

3.63Harmful Response Rate

5

Jun 16, 2026

Alpaca 7B (test)

1.292HV Score

5

Feb 26, 2026

Page 2 of 3

Popular datasets

PKU-SafeRLHF CoSApien Safety Alignment Dataset HarmBench AdvBench

Follow for updates

Join SOTA2 Contact