# Image Classification on ImageNet (val) (L\_inf Epsilon=4/255)

Source: [Image Classification on ImageNet (val) (L\_inf Epsilon=4/255)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-l_inf-epsilon-4-255>)

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

Data updated: 2026-06-02T11:59:22.979Z

Evaluation of Image Classification performance on the ImageNet validation set under adversarial attack constraints, specifically using the L\_infinity norm with an epsilon budget of 4/255. Metrics include clean accuracy and robust 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) | ImageNet l\_inf = 4/255 (val) |
| Metric | Clean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| 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

- [Clean Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-l_inf-epsilon-4-255.md?metric=Clean%20Accuracy>): Higher is better; unit: PERCENT.

- [Robust Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-l_inf-epsilon-4-255.md?metric=Robust%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Clean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Xu | 78.5 | Not recorded | 6a1e6433dbd7150506f8581d |
| Xu | 78 | Not recorded | 6a1e6433dbd7150506f8581d |
| Amini | 77.8 | Not recorded | 6a1e6433dbd7150506f8581d |
| Amini | 76.1 | Not recorded | 6a1e6433dbd7150506f8581d |
| Singh | 72.8 | Not recorded | 6a1e6433dbd7150506f8581d |
| Singh | 68.4 | Not recorded | 6a1e6433dbd7150506f8581d |
| Mo | 68 | Not recorded | 6a1e6433dbd7150506f8581d |
| Mo | 67.7 | Not recorded | 6a1e6433dbd7150506f8581d |

### Xu

Clean Accuracy: 78.5. Result ID: 6a1ec58cfbc373592c1a41e8.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"fully-connected classifier"}. This record does not assert independent reproduction.

### Xu

Clean Accuracy: 78. Result ID: 6a1ec58cfbc373592c1a41ec.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"distance-based classifier","Method framework":"HPM"}. This record does not assert independent reproduction.

### Amini

Clean Accuracy: 77.8. Result ID: 6a1ec58cfbc373592c1a41e9.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"fully-connected classifier"}. This record does not assert independent reproduction.

### Amini

Clean Accuracy: 76.1. Result ID: 6a1ec58cfbc373592c1a41ed.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"distance-based classifier","Method framework":"HPM"}. This record does not assert independent reproduction.

### Singh

Clean Accuracy: 72.8. Result ID: 6a1ec58cfbc373592c1a41ea.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"fully-connected classifier"}. This record does not assert independent reproduction.

### Singh

Clean Accuracy: 68.4. Result ID: 6a1ec58cfbc373592c1a41ee.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"distance-based classifier","Method framework":"HPM"}. This record does not assert independent reproduction.

### Mo

Clean Accuracy: 68. Result ID: 6a1ec58cfbc373592c1a41eb.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"fully-connected classifier"}. This record does not assert independent reproduction.

### Mo

Clean Accuracy: 67.7. Result ID: 6a1ec58cfbc373592c1a41ef.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e6433dbd7150506f8581d.md>)

Recorded evaluation configuration: {"Classifier type":"distance-based classifier","Method framework":"HPM"}. This record does not assert independent reproduction.

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