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

Jailbreak Defense research benchmarks – Page 5 · SOTA2 Research

Jailbreak Defense

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

8.4ASR (Noise, T)

6

Feb 26, 2026

Jailbreak Attacks ReNeLLM, DeepInception, GPTFuzzer, CodeAttack (test)

1Harmful Score (ReNeLLM)

6

Feb 26, 2026

GCG adversarial prompts (test)

Backtranslation w/o early termination

98DSR

6

Feb 26, 2026

Adversarial Jailbreak Attacks Cipher, Instructional Constraint, Prefix Injection, Psychological Coercion (Alternative)

Claude Opus 4.5

100Safety Score (Cipher)

5

Apr 27, 2026

100 dialogue samples (test)

2ASR (Low Difficulty)

5

Feb 26, 2026

95Acrostic DSR

5

Feb 26, 2026

ClearHarm Jailbreak (held-out)

47Attack Success Rate (ASR)

4

Jun 5, 2026

100LlamaGuard Score

4

Mar 9, 2026

Adaptive Jailbreak Evaluation target: Qwen2.5-72B (test)

4ASR

3

Jun 16, 2026

Emoji Attack Average

0.36ASR

3

Apr 3, 2026

Emoji Attack ActorAttack

0Attack Success Rate (ASR)

3

Apr 3, 2026

Emoji Attack MultiJail

0Attack Success Rate (ASR)

3

Apr 3, 2026

Emoji Attack DrAttack

0ASR

3

Apr 3, 2026

Emoji Attack AutoDAN

2ASR

3

Apr 3, 2026

Emoji Attack GCG

1Attack Success Rate (ASR)

3

Apr 3, 2026

StrongREJECT AutoDAN-Turbo attack (test)

87LlamaGuard Score

3

Mar 9, 2026

Jailbreak Attack Vectors

0GCG ASR

1

Apr 10, 2026

Page 5 of 5

Popular datasets

AdvBench JailbreakBench Emoji Attack GCG AutoDAN

Follow for updates

Join SOTA2 Contact