# Image Classification on CIFAR-10 (test) (Adversarial Robustness Metrics: FGSM, PGD)

Source: [Image Classification on CIFAR-10 (test) (Adversarial Robustness Metrics: FGSM, PGD)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-adversarial-robustness-metrics-fgsm-pgd>)

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

Data updated: 2026-06-02T09:08:09.655Z

Evaluation of Image Classification performance on the CIFAR-10 test set, reporting clean accuracy alongside robustness metrics against specific adversarial attacks (FGSM, PGD-10).

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 (test) |
| Metric | Accuracy (Clean) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

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

- [Accuracy (FGSM)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-adversarial-robustness-metrics-fgsm-pgd.md?metric=Accuracy%20(FGSM)>): Higher is better; unit: PERCENT.

- [Accuracy (PGD-10)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-adversarial-robustness-metrics-fgsm-pgd.md?metric=Accuracy%20(PGD-10)>): Higher is better; unit: PERCENT.

- [Adversarial Accuracy (AA)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-adversarial-robustness-metrics-fgsm-pgd.md?metric=Adversarial%20Accuracy%20(AA)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (Clean) | Result date | Report paper |
| --- | --- | --- | --- |
| Benign | 93.99 | Not recorded | 6a1e6385dbd7150506f8516d |
| ATAS | 89.07 | Not recorded | 6a1e6385dbd7150506f8516d |
| PGD-2 | 87.27 | Not recorded | 6a1e6385dbd7150506f8516d |
| NuAT | 86.53 | Not recorded | 6a1e6385dbd7150506f8516d |
| FGSM | 86.07 | Not recorded | 6a1e6385dbd7150506f8516d |
| GradAlign | 85.72 | Not recorded | 6a1e6385dbd7150506f8516d |
| MultiGrad | 84.92 | Not recorded | 6a1e6385dbd7150506f8516d |
| FGSM-RS | 84.12 | Not recorded | 6a1e6385dbd7150506f8516d |
| SORA | 83.44 | Not recorded | 6a1e6385dbd7150506f8516d |
| PGD-10 | 83.34 | Not recorded | 6a1e6385dbd7150506f8516d |
| AAER | 82.67 | Not recorded | 6a1e6385dbd7150506f8516d |
| N-FGSM | 82.54 | Not recorded | 6a1e6385dbd7150506f8516d |
| ZeroGrad | 81.66 | Not recorded | 6a1e6385dbd7150506f8516d |
| Free AT | 80.62 | Not recorded | 6a1e6385dbd7150506f8516d |
| TRADES | 80.43 | Not recorded | 6a1e6385dbd7150506f8516d |
| ELLE | 54.71 | Not recorded | 6a1e6385dbd7150506f8516d |

### Benign

Accuracy (Clean): 93.99. Result ID: 6a1e9d69fbc373592c18f6a9.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### ATAS

Accuracy (Clean): 89.07. Result ID: 6a1e9d69fbc373592c18f6b2.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cbc1e209131cea35b29.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### PGD-2

Accuracy (Clean): 87.27. Result ID: 6a1e9d69fbc373592c18f6b6.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa665a1e209131cea0fbae.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### NuAT

Accuracy (Clean): 86.53. Result ID: 6a1e9d69fbc373592c18f6ae.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### FGSM

Accuracy (Clean): 86.07. Result ID: 6a1e9d69fbc373592c18f6aa.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa61c91e209131ce9f5e06.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### GradAlign

Accuracy (Clean): 85.72. Result ID: 6a1e9d69fbc373592c18f6ad.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68a21e209131cea1d4e1.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### MultiGrad

Accuracy (Clean): 84.92. Result ID: 6a1e9d69fbc373592c18f6b0.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69a11e209131cea23383.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### FGSM-RS

Accuracy (Clean): 84.12. Result ID: 6a1e9d69fbc373592c18f6ac.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66e51e209131cea12efc.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### SORA

Accuracy (Clean): 83.44. Result ID: 6a1e9d69fbc373592c18f6b5.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### PGD-10

Accuracy (Clean): 83.34. Result ID: 6a1e9d69fbc373592c18f6b7.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa665a1e209131cea0fbae.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### AAER

Accuracy (Clean): 82.67. Result ID: 6a1e9d69fbc373592c18f6b3.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63c91.md>)

[Implementation](<https://github.com/tmllab/2023_NeurIPS_AAER>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### N-FGSM

Accuracy (Clean): 82.54. Result ID: 6a1e9d69fbc373592c18f6b1.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6da71e209131cea3b2ef.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### ZeroGrad

Accuracy (Clean): 81.66. Result ID: 6a1e9d69fbc373592c18f6af.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69a11e209131cea23383.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### Free AT

Accuracy (Clean): 80.62. Result ID: 6a1e9d69fbc373592c18f6ab.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66ad1e209131cea11a3f.md>)

[Implementation](<https://github.com/ashafahi/free_adv_train>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### TRADES

Accuracy (Clean): 80.43. Result ID: 6a1e9d69fbc373592c18f6b8.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

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

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

### ELLE

Accuracy (Clean): 54.71. Result ID: 6a1e9d69fbc373592c18f6b4.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6385dbd7150506f8516d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72541e209131cea55aa2.md>)

Recorded evaluation configuration: {"Architecture":"WideResNet-28"}. This record does not assert independent reproduction.

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