# Image Classification on CIFAR-10 (Adversarial Attack Robustness)

Source: [Image Classification on CIFAR-10 (Adversarial Attack Robustness)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-adversarial-attack-robustness>)

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

Data updated: 2026-03-10T10:04:21.122Z

Evaluation of Image Classification performance on CIFAR-10, reporting clean accuracy alongside robustness metrics derived from specific adversarial attacks (FGSM, PGD, AGSM, PAGD).

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 |
| Metric | Robustness (FGSM) |
| 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

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

- [Robustness (AGSM)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-adversarial-attack-robustness.md?metric=Robustness%20(AGSM)>): Higher is better; unit: PERCENT.

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

- [Robustness (PAGD)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-adversarial-attack-robustness.md?metric=Robustness%20(PAGD)>): Higher is better; unit: PERCENT.

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

## Recorded results

| Method | Robustness (FGSM) | Result date | Report paper |
| --- | --- | --- | --- |
| Poincare ResNet-32 | 60.68 | Not recorded | 691bfe1a6750be83fe5cecb2 |
| Poincare ResNet-32 | 59.1 | Not recorded | 691bfe1a6750be83fe5cecb2 |
| Poincare ResNet-20 | 56.59 | Not recorded | 691bfe1a6750be83fe5cecb2 |
| Poincare ResNet-20 | 54.43 | Not recorded | 691bfe1a6750be83fe5cecb2 |
| Poincare ResNet-32 | 54.19 | Not recorded | 691bfe1a6750be83fe5cecb2 |
| Poincare ResNet-20 | 49.63 | Not recorded | 691bfe1a6750be83fe5cecb2 |

### Poincare ResNet-32

Robustness (FGSM): 60.68. Result ID: 69afec9a6dbb7075b958d31d.

[Reporting paper](<https://www.sota2.com/research/paper/691bfe1a6750be83fe5cecb2.md>)

Recorded evaluation configuration: {"Model":"PRN 32","ε":"2.4/255 (ε1)"}. This record does not assert independent reproduction.

### Poincare ResNet-32

Robustness (FGSM): 59.1. Result ID: 69afec9a6dbb7075b958d31e.

[Reporting paper](<https://www.sota2.com/research/paper/691bfe1a6750be83fe5cecb2.md>)

Recorded evaluation configuration: {"Model":"PRN 32","ε":"3.2/255 (ε2)"}. This record does not assert independent reproduction.

### Poincare ResNet-20

Robustness (FGSM): 56.59. Result ID: 69afec9a6dbb7075b958d31a.

[Reporting paper](<https://www.sota2.com/research/paper/691bfe1a6750be83fe5cecb2.md>)

Recorded evaluation configuration: {"Model":"PRN 20","ε":"2.4/255 (ε1)"}. This record does not assert independent reproduction.

### Poincare ResNet-20

Robustness (FGSM): 54.43. Result ID: 69afec9a6dbb7075b958d31b.

[Reporting paper](<https://www.sota2.com/research/paper/691bfe1a6750be83fe5cecb2.md>)

Recorded evaluation configuration: {"Model":"PRN 20","ε":"3.2/255 (ε2)"}. This record does not assert independent reproduction.

### Poincare ResNet-32

Robustness (FGSM): 54.19. Result ID: 69afec9a6dbb7075b958d31f.

[Reporting paper](<https://www.sota2.com/research/paper/691bfe1a6750be83fe5cecb2.md>)

Recorded evaluation configuration: {"Model":"PRN 32","ε":"8.0/255 (ε3)"}. This record does not assert independent reproduction.

### Poincare ResNet-20

Robustness (FGSM): 49.63. Result ID: 69afec9a6dbb7075b958d31c.

[Reporting paper](<https://www.sota2.com/research/paper/691bfe1a6750be83fe5cecb2.md>)

Recorded evaluation configuration: {"Model":"PRN 20","ε":"8.0/255 (ε3)"}. This record does not assert independent reproduction.

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