# Image Classification on SVHN (test) (Corruption and Adversarial Robustness)

Source: [Image Classification on SVHN (test) (Corruption and Adversarial Robustness)](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness>)

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

Data updated: 2026-02-27T08:03:07.495Z

Evaluation of Image Classification performance on the SVHN test set, reporting standard accuracy alongside robustness metrics covering common corruptions (Hue, Saturation, Brightness, Contrast) and L-infinity adversarial attacks.

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) | SVHN (test) |
| Metric | Clean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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-svhn-test-corruption-and-adversarial-robustness.md?metric=Clean%20Accuracy>): Higher is better; unit: PERCENT.

- [Hue Robustness](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=Hue%20Robustness>): Higher is better; unit: PERCENT.

- [Saturation Robustness](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=Saturation%20Robustness>): Higher is better; unit: PERCENT.

- [Rotation Robustness](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=Rotation%20Robustness>): Higher is better; unit: PERCENT.

- [Brightness Robustness](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=Brightness%20Robustness>): Higher is better; unit: PERCENT.

- [Contrast Robustness](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=Contrast%20Robustness>): Higher is better; unit: PERCENT.

- [L-inf Robustness (20-step)](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=L-inf%20Robustness%20(20-step)>): Higher is better; unit: PERCENT.

- [Auto L-inf Robustness](<https://www.sota2.com/research/sota/image-classification-on-svhn-test-corruption-and-adversarial-robustness.md?metric=Auto%20L-inf%20Robustness>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Clean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Generalist | 94.11 | Not recorded | 68fa6ee61e209131cea42be3 |
| FAT | 93.87 | Not recorded | 68fa6ee61e209131cea42be3 |
| TRADES | 93.1 | Not recorded | 68fa6ee61e209131cea42be3 |

### Generalist

Clean Accuracy: 94.11. Result ID: 69a14fab35dfcabb00e62e94.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ee61e209131cea42be3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ee61e209131cea42be3.md>)

[Implementation](<https://github.com/PKU-ML/Generalist>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Maximum perturbation (ε)":"8/255"}. This record does not assert independent reproduction.

### FAT

Clean Accuracy: 93.87. Result ID: 69a14fab35dfcabb00e62e93.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ee61e209131cea42be3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68461e209131cea1af64.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Maximum perturbation (ε)":"8/255"}. This record does not assert independent reproduction.

### TRADES

Clean Accuracy: 93.1. Result ID: 69a14fab35dfcabb00e62e92.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6ee61e209131cea42be3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65f71e209131cea0d897.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Maximum perturbation (ε)":"8/255"}. This record does not assert independent reproduction.

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