# Image Classification on MNIST (test) (Adversarial Robustness: NAT, PGD, AA)

Source: [Image Classification on MNIST (test) (Adversarial Robustness: NAT, PGD, AA)](<https://www.sota2.com/research/sota/image-classification-on-mnist-test-adversarial-robustness-nat-pgd-aa>)

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

Data updated: 2026-02-27T08:03:07.540Z

Evaluation of Image Classification performance on the MNIST test set, reporting natural accuracy alongside robustness metrics against Projected Gradient Descent (PGD) and AutoAttack (AA).

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) | MNIST (test) |
| Metric | NAT (Natural 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

- [NAT (Natural Accuracy)](<https://www.sota2.com/research/sota/image-classification-on-mnist-test-adversarial-robustness-nat-pgd-aa.md?metric=NAT%20(Natural%20Accuracy)>): Higher is better; unit: PERCENT.

- [PGD-20 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-mnist-test-adversarial-robustness-nat-pgd-aa.md?metric=PGD-20%20Accuracy>): Higher is better; unit: PERCENT.

- [AA (AutoAttack Accuracy)](<https://www.sota2.com/research/sota/image-classification-on-mnist-test-adversarial-robustness-nat-pgd-aa.md?metric=AA%20(AutoAttack%20Accuracy)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | NAT (Natural Accuracy) | Result date | Report paper |
| --- | --- | --- | --- |
| Generalist | 99.24 | Not recorded | 68fa6ee61e209131cea42be3 |
| FAT | 99.18 | Not recorded | 68fa6ee61e209131cea42be3 |
| TRADES | 99.07 | Not recorded | 68fa6ee61e209131cea42be3 |

### Generalist

NAT (Natural Accuracy): 99.24. Result ID: 69a14fab35dfcabb00e62e91.

[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)":"0.3"}. This record does not assert independent reproduction.

### FAT

NAT (Natural Accuracy): 99.18. Result ID: 69a14fab35dfcabb00e62e90.

[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)":"0.3"}. This record does not assert independent reproduction.

### TRADES

NAT (Natural Accuracy): 99.07. Result ID: 69a14fab35dfcabb00e62e8f.

[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)":"0.3"}. This record does not assert independent reproduction.

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