# Image Classification on Pets PGD-l∞ attack, ε=4/255 (test)

Source: [Image Classification on Pets PGD-l∞ attack, ε=4/255 (test)](<https://www.sota2.com/research/sota/image-classification-on-pets-pgd-l-attack-4-255-test>)

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

Data updated: 2026-02-26T06:02:22.067Z

Evaluation of Image Classification robustness on the Pets test set against a Projected Gradient Descent (PGD) L-infinity adversarial attack with epsilon 4/255, reporting Robust Accuracy and Clean Accuracy.

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) | Pets PGD-l∞ attack, ε=4/255 (test) |
| Metric | Robust Accuracy |
| 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

- [Robust Accuracy](<https://www.sota2.com/research/sota/image-classification-on-pets-pgd-l-attack-4-255-test.md?metric=Robust%20Accuracy>): Higher is better; unit: PERCENT.

- [Clean Accuracy](<https://www.sota2.com/research/sota/image-classification-on-pets-pgd-l-attack-4-255-test.md?metric=Clean%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Robust Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Core-tuning | 67.98 | Not recorded | 68fa6b131e209131cea2bffd |
| AT-CE-Con | 65.6 | Not recorded | 68fa6b131e209131cea2bffd |
| AT-CE | 63.59 | Not recorded | 68fa6b131e209131cea2bffd |
| CE | 4.63 | Not recorded | 68fa6b131e209131cea2bffd |

### Core-tuning

Robust Accuracy: 67.98. Result ID: 698e5389ffc3901de09ca88a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b131e209131cea2bffd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b131e209131cea2bffd.md>)

[Implementation](<https://github.com/Vanint/Core-tuning>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Pre-training":"MoCo-v2","Attack Type":"PGD-10","Perturbation Norm":"l∞","Epsilon":"4/255"}. This record does not assert independent reproduction.

### AT-CE-Con

Robust Accuracy: 65.6. Result ID: 698e5389ffc3901de09ca889.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b131e209131cea2bffd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa665c1e209131cea0fcea.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Pre-training":"MoCo-v2","Attack Type":"PGD-10","Perturbation Norm":"l∞","Epsilon":"4/255"}. This record does not assert independent reproduction.

### AT-CE

Robust Accuracy: 63.59. Result ID: 698e5389ffc3901de09ca888.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b131e209131cea2bffd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa665c1e209131cea0fcea.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Pre-training":"MoCo-v2","Attack Type":"PGD-10","Perturbation Norm":"l∞","Epsilon":"4/255"}. This record does not assert independent reproduction.

### CE

Robust Accuracy: 4.63. Result ID: 698e5389ffc3901de09ca887.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b131e209131cea2bffd.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa699f1e209131cea2324f.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Pre-training":"MoCo-v2","Attack Type":"PGD-10","Perturbation Norm":"l∞","Epsilon":"4/255"}. This record does not assert independent reproduction.

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