# Image Classification on MNIST (Adversarial Robustness Metrics)

Source: [Image Classification on MNIST (Adversarial Robustness Metrics)](<https://www.sota2.com/research/sota/image-classification-on-mnist-adversarial-robustness-metrics>)

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

Data updated: 2026-05-11T07:43:32.953Z

Evaluation of Image Classification performance on the MNIST dataset, reporting clean accuracy alongside adversarial robustness metrics (Accadv, AURAC) and class count information.

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) | MNIST |
| Metric | Clean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| 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-mnist-adversarial-robustness-metrics.md?metric=Clean%20Accuracy>): Higher is better; unit: PERCENT.

- [Accadv (0.30)](<https://www.sota2.com/research/sota/image-classification-on-mnist-adversarial-robustness-metrics.md?metric=Accadv%20(0.30)>): Higher is better; unit: PERCENT.

- [AURAC](<https://www.sota2.com/research/sota/image-classification-on-mnist-adversarial-robustness-metrics.md?metric=AURAC>): Higher is better; unit: PERCENT.

- [Category Count](<https://www.sota2.com/research/sota/image-classification-on-mnist-adversarial-robustness-metrics.md?metric=Category%20Count>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Clean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Vanilla | 94.5 | Not recorded | 6a01628edbc836b87d8230b7 |
| AdvTrain (off) | 94.5 | Not recorded | 6a01628edbc836b87d8230b7 |
| Selective (off) | 94.5 | Not recorded | 6a01628edbc836b87d8230b7 |
| Selective (on) | 94.5 | Not recorded | 6a01628edbc836b87d8230b7 |
| Two-Stage Sel. | 94.5 | Not recorded | 6a01628edbc836b87d8230b7 |
| Sep-Aware | 94.2 | Not recorded | 6a01628edbc836b87d8230b7 |
| AdvTrain (on) | 94.1 | Not recorded | 6a01628edbc836b87d8230b7 |

### Vanilla

Clean Accuracy: 94.5. Result ID: 6a01888f47f34ce924e30fab.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

### AdvTrain (off)

Clean Accuracy: 94.5. Result ID: 6a01888f47f34ce924e30fac.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

### Selective (off)

Clean Accuracy: 94.5. Result ID: 6a01888f47f34ce924e30fae.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

### Selective (on)

Clean Accuracy: 94.5. Result ID: 6a01888f47f34ce924e30faf.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

### Two-Stage Sel.

Clean Accuracy: 94.5. Result ID: 6a01888f47f34ce924e30fb1.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

### Sep-Aware

Clean Accuracy: 94.2. Result ID: 6a01888f47f34ce924e30fb0.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

### AdvTrain (on)

Clean Accuracy: 94.1. Result ID: 6a01888f47f34ce924e30fad.

[Reporting paper](<https://www.sota2.com/research/paper/6a01628edbc836b87d8230b7.md>)

Recorded evaluation configuration: {"Attack":"Black-Box Transfer PGD-20","ρ":"0.9","Surrogate Model":"SimpleCNN"}. This record does not assert independent reproduction.

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