# Certified Robust Classification on CIFAR-10 official (test)

Source: [Certified Robust Classification on CIFAR-10 official (test)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test>)

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

Data updated: 2026-02-27T07:00:17.315Z

Evaluation of Certified Robust Classification performance on the official test set of CIFAR-10, reporting certified accuracy across various perturbation radii (r=0.00 to r=1.75) and the overall ACR score.

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 | Certified Robust Classification |
| Dataset (including any stated split) | CIFAR-10 official (test) |
| Metric | ACR |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [ACR](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=ACR>): Higher is better; unit: SCALAR.

- [Certified Accuracy (r=0.00)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D0.00)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=0.25)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D0.25)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=0.50)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D0.50)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=0.75)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D0.75)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=1.00)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D1.00)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=1.25)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D1.25)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=1.50)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D1.50)>): Higher is better; unit: PERCENT.

- [Certified Accuracy (r=1.75)](<https://www.sota2.com/research/sota/certified-robust-classification-on-cifar-10-official-test.md?metric=Certified%20Accuracy%20(r%3D1.75)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ACR | Result date | Report paper |
| --- | --- | --- | --- |
| SmoothMix + One-step adversary | 0.737 | Not recorded | 68fa6b4a1e209131cea2d586 |
| Consistency | 0.72 | Not recorded | 68fa6b4a1e209131cea2d586 |
| SmoothMix | 0.715 | Not recorded | 68fa6b4a1e209131cea2d586 |
| MACER | 0.691 | Not recorded | 68fa6b4a1e209131cea2d586 |
| SmoothAdv | 0.684 | Not recorded | 68fa6b4a1e209131cea2d586 |
| SmoothMix | 0.553 | Not recorded | 68fa6b4a1e209131cea2d586 |
| Consistency | 0.552 | Not recorded | 68fa6b4a1e209131cea2d586 |
| SmoothMix + One-step adversary | 0.548 | Not recorded | 68fa6b4a1e209131cea2d586 |
| SmoothAdv | 0.544 | Not recorded | 68fa6b4a1e209131cea2d586 |
| MACER | 0.531 | Not recorded | 68fa6b4a1e209131cea2d586 |
| Gaussian | 0.525 | Not recorded | 68fa6b4a1e209131cea2d586 |
| Stability training | 0.521 | Not recorded | 68fa6b4a1e209131cea2d586 |
| Gaussian | 0.424 | Not recorded | 68fa6b4a1e209131cea2d586 |
| Stability training | 0.421 | Not recorded | 68fa6b4a1e209131cea2d586 |

### SmoothMix + One-step adversary

ACR: 0.737. Result ID: 69a140e335dfcabb00e5ce81.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Implementation](<https://github.com/jh-jeong/smoothmix>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110","One-step adversary":"true"}. This record does not assert independent reproduction.

### Consistency

ACR: 0.72. Result ID: 69a140e335dfcabb00e5ce7f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690b1e209131cea1fae9.md>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110"}. This record does not assert independent reproduction.

### SmoothMix

ACR: 0.715. Result ID: 69a140e335dfcabb00e5ce80.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Implementation](<https://github.com/jh-jeong/smoothmix>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110","One-step adversary":"false"}. This record does not assert independent reproduction.

### MACER

ACR: 0.691. Result ID: 69a140e335dfcabb00e5ce7e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea317f0.md>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110","Evaluated from official pre-trained models":"true"}. This record does not assert independent reproduction.

### SmoothAdv

ACR: 0.684. Result ID: 69a140e335dfcabb00e5ce7d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66e41e209131cea12df7.md>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110","Evaluated from official pre-trained models":"true"}. This record does not assert independent reproduction.

### SmoothMix

ACR: 0.553. Result ID: 69a140e335dfcabb00e5ce79.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Implementation](<https://github.com/jh-jeong/smoothmix>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110","One-step adversary":"false"}. This record does not assert independent reproduction.

### Consistency

ACR: 0.552. Result ID: 69a140e335dfcabb00e5ce78.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690b1e209131cea1fae9.md>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110"}. This record does not assert independent reproduction.

### SmoothMix + One-step adversary

ACR: 0.548. Result ID: 69a140e335dfcabb00e5ce7a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Implementation](<https://github.com/jh-jeong/smoothmix>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110","One-step adversary":"true"}. This record does not assert independent reproduction.

### SmoothAdv

ACR: 0.544. Result ID: 69a140e335dfcabb00e5ce76.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66e41e209131cea12df7.md>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110","Evaluated from official pre-trained models":"true"}. This record does not assert independent reproduction.

### MACER

ACR: 0.531. Result ID: 69a140e335dfcabb00e5ce77.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfe1e209131cea317f0.md>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110","Evaluated from official pre-trained models":"true"}. This record does not assert independent reproduction.

### Gaussian

ACR: 0.525. Result ID: 69a140e335dfcabb00e5ce7b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65de1e209131cea0cf72.md>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110"}. This record does not assert independent reproduction.

### Stability training

ACR: 0.521. Result ID: 69a140e335dfcabb00e5ce7c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66ae1e209131cea11af1.md>)

Recorded evaluation configuration: {"sigma":"0.50","Backbone":"ResNet-110"}. This record does not assert independent reproduction.

### Gaussian

ACR: 0.424. Result ID: 69a140e335dfcabb00e5ce74.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65de1e209131cea0cf72.md>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110"}. This record does not assert independent reproduction.

### Stability training

ACR: 0.421. Result ID: 69a140e335dfcabb00e5ce75.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b4a1e209131cea2d586.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66ae1e209131cea11af1.md>)

Recorded evaluation configuration: {"sigma":"0.25","Backbone":"ResNet-110"}. This record does not assert independent reproduction.

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