# Unlearning Attack Success on MNIST, Fashion-MNIST, SVHN, and CIFAR-10 Average (test)

Source: [Unlearning Attack Success on MNIST, Fashion-MNIST, SVHN, and CIFAR-10 Average (test)](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test>)

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

Data updated: 2026-04-10T06:38:29.824Z

Evaluation measuring the success rate of unlearning attacks across the test splits of MNIST, Fashion-MNIST, SVHN, and CIFAR-10 datasets, reporting metrics related to attack efficacy.

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 | Unlearning Attack Success |
| Dataset (including any stated split) | MNIST, Fashion-MNIST, SVHN, and CIFAR-10 Average (test) |
| Metric | NG Success Rate |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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

- [NG Success Rate](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test.md?metric=NG%20Success%20Rate>): Higher is better; unit: PERCENT.

- [FT Success Rate](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test.md?metric=FT%20Success%20Rate>): Higher is better; unit: PERCENT.

- [RL Success Rate](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test.md?metric=RL%20Success%20Rate>): Higher is better; unit: PERCENT.

- [AU Success Rate](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test.md?metric=AU%20Success%20Rate>): Higher is better; unit: PERCENT.

- [RT Success Rate](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test.md?metric=RT%20Success%20Rate>): Higher is better; unit: PERCENT.

- [Overall Average Success Rate](<https://www.sota2.com/research/sota/unlearning-attack-success-on-mnist-fashion-mnist-svhn-and-cifar-10-average-test.md?metric=Overall%20Average%20Success%20Rate>): Higher is better; unit: PERCENT.

## Recorded results

| Method | NG Success Rate | Result date | Report paper |
| --- | --- | --- | --- |
| Model Parameter Difference | 90.31 | Not recorded | 69d884117d0172140ec779fc |
| Model Parameter Dot Product | 81.5 | Not recorded | 69d884117d0172140ec779fc |
| Model Parameter Dot Product | 80.12 | Not recorded | 69d884117d0172140ec779fc |

### Model Parameter Difference

NG Success Rate: 90.31. Result ID: 69d8996447f34ce924c0d67b.

[Reporting paper](<https://www.sota2.com/research/paper/69d884117d0172140ec779fc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d884117d0172140ec779fc.md>)

Recorded evaluation configuration: {"Attack Method":"Difference","Classification Method":"Decision Tree","Unlearning Scenario":"Single-Class"}. This record does not assert independent reproduction.

### Model Parameter Dot Product

NG Success Rate: 81.5. Result ID: 69d8996447f34ce924c0d679.

[Reporting paper](<https://www.sota2.com/research/paper/69d884117d0172140ec779fc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d884117d0172140ec779fc.md>)

Recorded evaluation configuration: {"Attack Method":"Dot Product","Classification Method":"Youden’s Threshold","Unlearning Scenario":"Single-Class"}. This record does not assert independent reproduction.

### Model Parameter Dot Product

NG Success Rate: 80.12. Result ID: 69d8996447f34ce924c0d67a.

[Reporting paper](<https://www.sota2.com/research/paper/69d884117d0172140ec779fc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d884117d0172140ec779fc.md>)

Recorded evaluation configuration: {"Attack Method":"Dot Product","Classification Method":"K-Means","Unlearning Scenario":"Single-Class"}. This record does not assert independent reproduction.

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