# Machine Unlearning on CIFAR-10 10% random data forgetting

Source: [Machine Unlearning on CIFAR-10 10% random data forgetting](<https://www.sota2.com/research/sota/machine-unlearning-on-cifar-10-10-percent-random-data-forgetting>)

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

Data updated: 2026-06-02T08:06:58.719Z

Evaluation of Machine Unlearning effectiveness on CIFAR-10, specifically focusing on the scenario where 10% of the training data was randomly selected for forgetting. Key metrics include utility retention (UA, RA, TA) and privacy leakage (MIA, Avg. Gap).

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 | Machine Unlearning |
| Dataset (including any stated split) | CIFAR-10 10% random data forgetting |
| Metric | Utility Accuracy (UA) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 11 of 11 |
| 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

- [Utility Accuracy (UA)](<https://www.sota2.com/research/sota/machine-unlearning-on-cifar-10-10-percent-random-data-forgetting.md?metric=Utility%20Accuracy%20(UA)>): Higher is better; unit: PERCENT.

- [Retention Accuracy (RA)](<https://www.sota2.com/research/sota/machine-unlearning-on-cifar-10-10-percent-random-data-forgetting.md?metric=Retention%20Accuracy%20(RA)>): Higher is better; unit: PERCENT.

- [Training Accuracy (TA)](<https://www.sota2.com/research/sota/machine-unlearning-on-cifar-10-10-percent-random-data-forgetting.md?metric=Training%20Accuracy%20(TA)>): Higher is better; unit: PERCENT.

- [Membership Inference Attack Success Rate (MIA)](<https://www.sota2.com/research/sota/machine-unlearning-on-cifar-10-10-percent-random-data-forgetting.md?metric=Membership%20Inference%20Attack%20Success%20Rate%20(MIA)>): Lower is better; unit: PERCENT.

- [Average Gap (Privacy Leakage)](<https://www.sota2.com/research/sota/machine-unlearning-on-cifar-10-10-percent-random-data-forgetting.md?metric=Average%20Gap%20(Privacy%20Leakage)>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Utility Accuracy (UA) | Result date | Report paper |
| --- | --- | --- | --- |
| l1-sparse | 56.12 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| RL | 18.75 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| GA | 11.29 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| SalUn | 7.68 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| HeldDist-RAUL-LS | 7.3 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| HeldDist-RAUL-UPGrad | 6.47 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| Retrain | 5.68 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| FT | 3.93 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| UniDist-RAUL-UPGrad | 1.43 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| UniDist-RAUL-LS | 0.97 | Not recorded | 6a1e63b8dbd7150506f852d9 |
| IU | 0.74 | Not recorded | 6a1e63b8dbd7150506f852d9 |

### l1-sparse

Utility Accuracy (UA): 56.12. Result ID: 6a1e8f17fbc373592c188f5a.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72001e209131cea540dc.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

### RL

Utility Accuracy (UA): 18.75. Result ID: 6a1e8f17fbc373592c188f5b.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67551e209131cea158d5.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

### GA

Utility Accuracy (UA): 11.29. Result ID: 6a1e8f17fbc373592c188f58.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf01e209131cea31038.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

### SalUn

Utility Accuracy (UA): 7.68. Result ID: 6a1e8f17fbc373592c188f5c.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72cd1e209131cea5888f.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

### HeldDist-RAUL-LS

Utility Accuracy (UA): 7.3. Result ID: 6a1e8f17fbc373592c188f5f.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Gradient Aggregator":"LS"}. This record does not assert independent reproduction.

### HeldDist-RAUL-UPGrad

Utility Accuracy (UA): 6.47. Result ID: 6a1e8f17fbc373592c188f60.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Gradient Aggregator":"UPGrad"}. This record does not assert independent reproduction.

### Retrain

Utility Accuracy (UA): 5.68. Result ID: 6a1e8f17fbc373592c188f56.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

### FT

Utility Accuracy (UA): 3.93. Result ID: 6a1e8f17fbc373592c188f57.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70381e209131cea49a4c.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

### UniDist-RAUL-UPGrad

Utility Accuracy (UA): 1.43. Result ID: 6a1e8f17fbc373592c188f5e.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Gradient Aggregator":"UPGrad"}. This record does not assert independent reproduction.

### UniDist-RAUL-LS

Utility Accuracy (UA): 0.97. Result ID: 6a1e8f17fbc373592c188f5d.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Gradient Aggregator":"LS"}. This record does not assert independent reproduction.

### IU

Utility Accuracy (UA): 0.74. Result ID: 6a1e8f17fbc373592c188f59.

[Reporting paper](<https://www.sota2.com/research/paper/6a1e63b8dbd7150506f852d9.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa695c1e209131cea21c70.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction.

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