# Image Classification on CIFAR-10 (Certified Robustness vs Perturbation Level)

Source: [Image Classification on CIFAR-10 (Certified Robustness vs Perturbation Level)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-certified-robustness-vs-perturbation-level>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:39:18.604Z

Evaluation of Image Classification performance on CIFAR-10, reporting standard clean accuracy alongside Certified Robust Accuracy (CRA) measured at varying levels of input perturbation (36/255, 72/255, and 108/255).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Image Classification |
| Dataset (including any stated split) | CIFAR-10 |
| Metric | Clean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Clean Accuracy](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-certified-robustness-vs-perturbation-level.md?metric=Clean%20Accuracy>): Higher is better; unit: PERCENT.

- [CRA (36/255)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-certified-robustness-vs-perturbation-level.md?metric=CRA%20(36%2F255)>): Higher is better; unit: PERCENT.

- [CRA (72/255)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-certified-robustness-vs-perturbation-level.md?metric=CRA%20(72%2F255)>): Higher is better; unit: PERCENT.

- [CRA (108/255)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-certified-robustness-vs-perturbation-level.md?metric=CRA%20(108%2F255)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Clean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| LipNeXt L32W2896 | 92.7 | Not recorded | 697846dfa2b28e761d87c821 |
| LipNeXt L32W2048 | 88.2 | Not recorded | 697846dfa2b28e761d87c821 |
| BRO | 87.2 | Not recorded | 697846dfa2b28e761d87c821 |
| Cayley | 86.7 | Not recorded | 697846dfa2b28e761d87c821 |
| Lip-reg | 86.7 | Not recorded | 697846dfa2b28e761d87c821 |
| SOC | 86.6 | Not recorded | 697846dfa2b28e761d87c821 |
| LOT | 85.7 | Not recorded | 697846dfa2b28e761d87c821 |
| SLL | 85.6 | Not recorded | 697846dfa2b28e761d87c821 |
| Cholesky | 85.4 | Not recorded | 697846dfa2b28e761d87c821 |

### LipNeXt L32W2896

Clean Accuracy: 92.7. Result ID: 6999158b5698665ec47bae8f.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

Recorded evaluation configuration: {"#Param.":"512M"}. This record does not assert independent reproduction.

### LipNeXt L32W2048

Clean Accuracy: 88.2. Result ID: 6999158b5698665ec47bae8e.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

Recorded evaluation configuration: {"#Param.":"256M"}. This record does not assert independent reproduction.

### BRO

Clean Accuracy: 87.2. Result ID: 6999158b5698665ec47bae8d.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a611e209131cea82bb3.md>)

Recorded evaluation configuration: {"#Param.":"68M"}. This record does not assert independent reproduction.

### Cayley

Clean Accuracy: 86.7. Result ID: 6999158b5698665ec47bae88.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69cb1e209131cea24142.md>)

Recorded evaluation configuration: {"#Param.":"68M"}. This record does not assert independent reproduction.

### Lip-reg

Clean Accuracy: 86.7. Result ID: 6999158b5698665ec47bae8c.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a281e209131cea266ee.md>)

Recorded evaluation configuration: {"#Param.":"83M"}. This record does not assert independent reproduction.

### SOC

Clean Accuracy: 86.6. Result ID: 6999158b5698665ec47bae8b.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a291e209131cea2676b.md>)

Recorded evaluation configuration: {"#Param.":"83M"}. This record does not assert independent reproduction.

### LOT

Clean Accuracy: 85.7. Result ID: 6999158b5698665ec47bae87.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ee71e209131cea42cbd.md>)

[Implementation](<https://github.com/AI-secure/Layerwise-Orthogonal-Training>)

Recorded evaluation configuration: {"#Param.":"59M"}. This record does not assert independent reproduction.

### SLL

Clean Accuracy: 85.6. Result ID: 6999158b5698665ec47bae8a.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71091e209131cea4ecb1.md>)

[Implementation](<https://www.sota2.com/github.com/araujoalexandre/Lipschitz-SLL-Networks>)

Recorded evaluation configuration: {"#Param.":"83M"}. This record does not assert independent reproduction.

### Cholesky

Clean Accuracy: 85.4. Result ID: 6999158b5698665ec47bae89.

[Reporting paper](<https://www.sota2.com/research/paper/697846dfa2b28e761d87c821.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b61e209131cea5e0f9.md>)

Recorded evaluation configuration: {"#Param.":"68M"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.