# Image Classification on ImageNet (val) (Adversarial Robustness - FGSM/PGD)

Source: [Image Classification on ImageNet (val) (Adversarial Robustness - FGSM/PGD)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd>)

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

Data updated: 2026-04-02T08:43:04.644Z

Evaluation of Image Classification performance on the validation split of the ImageNet dataset, specifically measuring robustness against adversarial attacks such as FGSM and PGD, and reporting associated accuracy and negative log-likelihood scores.

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) | ImageNet (val) |
| Metric | FGSM Top-1 Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [FGSM Top-1 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd.md?metric=FGSM%20Top-1%20Accuracy>): Higher is better; unit: PERCENT.

- [FGSM NLL](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd.md?metric=FGSM%20NLL>): Lower is better; unit: SCALAR.

- [PGD Top-1 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd.md?metric=PGD%20Top-1%20Accuracy>): Higher is better; unit: PERCENT.

- [PGD NLL](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd.md?metric=PGD%20NLL>): Lower is better; unit: SCALAR.

- [Auto Attack Top-1 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd.md?metric=Auto%20Attack%20Top-1%20Accuracy>): Higher is better; unit: PERCENT.

- [Auto Attack NLL](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-adversarial-robustness-fgsm-pgd.md?metric=Auto%20Attack%20NLL>): Lower is better; unit: SCALAR.

## Recorded results

| Method | FGSM Top-1 Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| ViT-B/16† | 70.67 | Not recorded | 69cdf7f20636e5bf4afcc99a |
| ViT-B/16† | 69.41 | Not recorded | 69cdf7f20636e5bf4afcc99a |
| ViT-S/16 | 67.02 | Not recorded | 69cdf7f20636e5bf4afcc99a |
| ViT-S/16 | 65.9 | Not recorded | 69cdf7f20636e5bf4afcc99a |
| ViT-L/16 | 62.85 | Not recorded | 69cdf7f20636e5bf4afcc99a |
| ViT-L/16 | 58.95 | Not recorded | 69cdf7f20636e5bf4afcc99a |

### ViT-B/16†

FGSM Top-1 Accuracy: 70.67. Result ID: 69ce2bef47f34ce924b75d77.

[Reporting paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

Recorded evaluation configuration: {"Attention":"QUEST","Epochs":"400 + 20","Training Strategy":"DeiT-3"}. This record does not assert independent reproduction.

### ViT-B/16†

FGSM Top-1 Accuracy: 69.41. Result ID: 69ce2bef47f34ce924b75d76.

[Reporting paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

Recorded evaluation configuration: {"Attention":"Standard","Epochs":"400 + 20","Training Strategy":"DeiT-3"}. This record does not assert independent reproduction.

### ViT-S/16

FGSM Top-1 Accuracy: 67.02. Result ID: 69ce2bef47f34ce924b75d75.

[Reporting paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

Recorded evaluation configuration: {"Attention":"QUEST","Epochs":"200"}. This record does not assert independent reproduction.

### ViT-S/16

FGSM Top-1 Accuracy: 65.9. Result ID: 69ce2bef47f34ce924b75d74.

[Reporting paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

Recorded evaluation configuration: {"Attention":"Standard","Epochs":"200"}. This record does not assert independent reproduction.

### ViT-L/16

FGSM Top-1 Accuracy: 62.85. Result ID: 69ce2bef47f34ce924b75d79.

[Reporting paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

Recorded evaluation configuration: {"Attention":"QUEST","Epochs":"100"}. This record does not assert independent reproduction.

### ViT-L/16

FGSM Top-1 Accuracy: 58.95. Result ID: 69ce2bef47f34ce924b75d78.

[Reporting paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cdf7f20636e5bf4afcc99a.md>)

Recorded evaluation configuration: {"Attention":"QKNorm-LN","Epochs":"100"}. This record does not assert independent reproduction.

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