# Image Classification on STL-10 (Adversarial Robustness PGD/FGSM/SPSA)

Source: [Image Classification on STL-10 (Adversarial Robustness PGD/FGSM/SPSA)](<https://www.sota2.com/research/sota/image-classification-on-stl-10-adversarial-robustness-pgd-fgsm-spsa>)

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

Data updated: 2026-06-01T07:46:24.897Z

Evaluation of Image Classification performance on the STL-10 dataset, specifically reporting robustness metrics derived from PGD, FGSM, and SPSA adversarial attacks.

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) | STL-10 |
| Metric | Robustness (PGD) |
| 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

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

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

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

## Recorded results

| Method | Robustness (PGD) | Result date | Report paper |
| --- | --- | --- | --- |
| USMoE | 41 | Not recorded | 68fa77be1e209131cea7510f |
| SoftMoE | 39.8 | Not recorded | 68fa77be1e209131cea7510f |
| EC | 39.5 | Not recorded | 68fa77be1e209131cea7510f |
| TC | 37.7 | Not recorded | 68fa77be1e209131cea7510f |

### USMoE

Robustness (PGD): 41. Result ID: 6a1d38bffbc373592c161d98.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77be1e209131cea7510f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77be1e209131cea7510f.md>)

Recorded evaluation configuration: {"Backbone":"ViT-MoE (10M)"}. This record does not assert independent reproduction.

### SoftMoE

Robustness (PGD): 39.8. Result ID: 6a1d38bffbc373592c161d9b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77be1e209131cea7510f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73521e209131cea5bd78.md>)

Recorded evaluation configuration: {"Backbone":"ViT-MoE (10M)"}. This record does not assert independent reproduction.

### EC

Robustness (PGD): 39.5. Result ID: 6a1d38bffbc373592c161d9a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77be1e209131cea7510f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6da71e209131cea3b2d8.md>)

Recorded evaluation configuration: {"Backbone":"ViT-MoE (10M)"}. This record does not assert independent reproduction.

### TC

Robustness (PGD): 37.7. Result ID: 6a1d38bffbc373592c161d99.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77be1e209131cea7510f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63171e209131ce9fc941.md>)

Recorded evaluation configuration: {"Backbone":"ViT-MoE (10M)"}. This record does not assert independent reproduction.

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