# Image Classification on CIFAR10 (Adversarial Robustness Metrics)

Source: [Image Classification on CIFAR10 (Adversarial Robustness Metrics)](<https://www.sota2.com/research/sota/image-classification-on-cifar10-adversarial-robustness-metrics>)

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

Data updated: 2026-02-26T05:39:15.675Z

Evaluation of Image Classification performance on the CIFAR-10 dataset, reporting metrics related to adversarial robustness, such as Clean accuracy, Performance drop, and Attack success rate.

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) | CIFAR10 |
| Metric | Clean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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-cifar10-adversarial-robustness-metrics.md?metric=Clean%20Accuracy>): Higher is better; unit: PERCENT.

- [Performance Drop](<https://www.sota2.com/research/sota/image-classification-on-cifar10-adversarial-robustness-metrics.md?metric=Performance%20Drop>): Lower is better; unit: PERCENT.

- [ASR](<https://www.sota2.com/research/sota/image-classification-on-cifar10-adversarial-robustness-metrics.md?metric=ASR>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Clean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Marksman | 94.3 | Not recorded | 68fa6dae1e209131cea3b77d |
| Marksman | 94.1 | Not recorded | 68fa6dae1e209131cea3b77d |
| Marksman | 93.8 | Not recorded | 68fa6dae1e209131cea3b77d |
| RefoolMT | 93.4 | Not recorded | 68fa6dae1e209131cea3b77d |
| PatchMT | 93.3 | Not recorded | 68fa6dae1e209131cea3b77d |
| WaNetMT | 92 | Not recorded | 68fa6dae1e209131cea3b77d |
| RefoolMT | 91 | Not recorded | 68fa6dae1e209131cea3b77d |
| WaNetMT | 89.4 | Not recorded | 68fa6dae1e209131cea3b77d |
| PatchMT | 88.2 | Not recorded | 68fa6dae1e209131cea3b77d |
| Marksman | 36.5 | Not recorded | 68fa6dae1e209131cea3b77d |

### Marksman

Clean Accuracy: 94.3. Result ID: 698fbeae3a6c314849fb9288.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

Recorded evaluation configuration: {"Poisoning Rate":"10%"}. This record does not assert independent reproduction.

### Marksman

Clean Accuracy: 94.1. Result ID: 698fbeae3a6c314849fb9278.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

Recorded evaluation configuration: {"Poisoning rate":"50%"}. This record does not assert independent reproduction.

### Marksman

Clean Accuracy: 93.8. Result ID: 698fbeae3a6c314849fb9292.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

Recorded evaluation configuration: {"Post-training Defense Method":"Adversarial Neuron Pruning (ANP)"}. This record does not assert independent reproduction.

### RefoolMT

Clean Accuracy: 93.4. Result ID: 698fbeae3a6c314849fb9286.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67e61e209131cea18fda.md>)

Recorded evaluation configuration: {"Poisoning Rate":"10%"}. This record does not assert independent reproduction.

### PatchMT

Clean Accuracy: 93.3. Result ID: 698fbeae3a6c314849fb9285.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

Recorded evaluation configuration: {"Poisoning Rate":"10%"}. This record does not assert independent reproduction.

### WaNetMT

Clean Accuracy: 92. Result ID: 698fbeae3a6c314849fb9277.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa696b1e209131cea21e53.md>)

Recorded evaluation configuration: {"Poisoning rate":"50%"}. This record does not assert independent reproduction.

### RefoolMT

Clean Accuracy: 91. Result ID: 698fbeae3a6c314849fb9276.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67e61e209131cea18fda.md>)

Recorded evaluation configuration: {"Poisoning rate":"50%"}. This record does not assert independent reproduction.

### WaNetMT

Clean Accuracy: 89.4. Result ID: 698fbeae3a6c314849fb9287.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa696b1e209131cea21e53.md>)

Recorded evaluation configuration: {"Poisoning Rate":"10%"}. This record does not assert independent reproduction.

### PatchMT

Clean Accuracy: 88.2. Result ID: 698fbeae3a6c314849fb9275.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

Recorded evaluation configuration: {"Poisoning rate":"50%"}. This record does not assert independent reproduction.

### Marksman

Clean Accuracy: 36.5. Result ID: 698fbeae3a6c314849fb9291.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dae1e209131cea3b77d.md>)

Recorded evaluation configuration: {"Post-training Defense Method":"Neural Attention Distillation (NAD)"}. This record does not assert independent reproduction.

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