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

Adversarial Attack research benchmarks – Page 12 · SOTA2 Research

Adversarial Attack

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

GF-counterattack

0.12Perturbation Magnitude (L-inf)

4

Apr 29, 2026

SMAC 27m_vs_30m

0.19Perturbation Magnitude (L-inf)

4

Apr 29, 2026

SMAC bane_vs_bane

0.16Perturbation Magnitude (L-inf)

4

Apr 29, 2026

0.12Perturbation Magnitude (L-inf)

4

Apr 29, 2026

SecRiskBench real physical environments

56.2OS

4

Apr 28, 2026

67ASR (2.0m)

4

Mar 23, 2026

OR-Bench unsafe inputs

88ASR

4

Feb 26, 2026

AdvBench unsafe inputs

94ASR

4

Feb 26, 2026

76.72Attack Success Rate

4

Feb 26, 2026

SafeMT ATTACK 600

71.29Attack Success Rate

4

Feb 26, 2026

LibrAI Do-Not-Answer and IndicAlign (145 prompts)

GCG (Brute-Force)

59ASR (%)

4

Feb 26, 2026

CIFAR100 (test)

89.54ASR (e=2/255)

4

Feb 26, 2026

ImageNet epsilon=4/255 (val)

74.52Attack Success Rate

4

Feb 26, 2026

0Clean Accuracy

4

Feb 27, 2026

37.98Attack Success Rate

4

Feb 26, 2026

Microsoft Azure

18.7Attack Success Rate

4

Feb 26, 2026

Claude Opus 4.1 (test)

46.7Bin-ASR

3

Jun 9, 2026

OpenAI o3 (test)

66.2Bin-ASR

3

Jun 9, 2026

31.9Drop

3

May 25, 2026

0.446Performance Drop

3

May 25, 2026

Llama-3.3-70B successful attacks

2,240Unique Queries

3

Apr 24, 2026

Llama-3-70B successful attacks

1,321Unique Queries Count

3

Apr 24, 2026

Mistral-7B (successful attacks)

3,021Unique Queries

3

Apr 24, 2026

ImageNet Various Robust Models

59.7Average tSuc

3

Apr 6, 2026

ImageNet Ensemble AT

70.8tSuc (INv3ens3)

3

Apr 6, 2026

Page 12 of 14

Popular datasets

ImageNet CIFAR AdvBench Robotic Visuomotor Policy SMAC

Follow for updates

Join SOTA2 Contact