# Image Classification on ImageNet Linf epsilon=2/255 (test)

Source: [Image Classification on ImageNet Linf epsilon=2/255 (test)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-linf-epsilon-2-255-test>)

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

Data updated: 2026-02-26T05:34:12.179Z

Evaluation of Image Classification robustness on the test split of ImageNet, specifically measuring accuracy under PGD adversarial attacks with an L-infinity constraint of epsilon=2/255.

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 Linf epsilon=2/255 (test) |
| Metric | Accuracy (Clean) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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-imagenet-linf-epsilon-2-255-test.md?metric=Accuracy%20(Clean)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (PGD-10)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-linf-epsilon-2-255-test.md?metric=Robust%20Accuracy%20(PGD-10)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (PGD-50)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-linf-epsilon-2-255-test.md?metric=Robust%20Accuracy%20(PGD-50)>): Higher is better; unit: PERCENT.

- [Robust Accuracy (PGD-100)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-linf-epsilon-2-255-test.md?metric=Robust%20Accuracy%20(PGD-100)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (Clean) | Result date | Report paper |
| --- | --- | --- | --- |
| Free-AT | 64.45 | Not recorded | 68fa6b361e209131cea2cf0b |
| MixACM + RA | 62.05 | Not recorded | 68fa6b361e209131cea2cf0b |
| Fast-AT | 60.9 | Not recorded | 68fa6b361e209131cea2cf0b |
| MixACM | 59.64 | Not recorded | 68fa6b361e209131cea2cf0b |

### Free-AT

Accuracy (Clean): 64.45. Result ID: 698e7150ffc3901de09da2ab.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b361e209131cea2cf0b.md>)

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

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

Recorded evaluation configuration: {"Backbone":"ResNet-50","Train ε":"2/255"}. This record does not assert independent reproduction.

### MixACM + RA

Accuracy (Clean): 62.05. Result ID: 698e7150ffc3901de09da2ae.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b361e209131cea2cf0b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b361e209131cea2cf0b.md>)

[Implementation](<https://www.sota2.com/awaisrauf.github.io/MixACM>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Train ε":"0"}. This record does not assert independent reproduction.

### Fast-AT

Accuracy (Clean): 60.9. Result ID: 698e7150ffc3901de09da2ac.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b361e209131cea2cf0b.md>)

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

Recorded evaluation configuration: {"Backbone":"ResNet-50","Train ε":"2/255"}. This record does not assert independent reproduction.

### MixACM

Accuracy (Clean): 59.64. Result ID: 698e7150ffc3901de09da2ad.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b361e209131cea2cf0b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b361e209131cea2cf0b.md>)

[Implementation](<https://www.sota2.com/awaisrauf.github.io/MixACM>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Train ε":"0"}. This record does not assert independent reproduction.

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