# Machine Unlearning on Tiny-ImageNet (10% Random Forgetting) Source: [Machine Unlearning on Tiny-ImageNet (10% Random Forgetting)]() Collection index: [llms.txt]() Data updated: 2026-05-19T10:22:59.524Z Evaluation of Machine Unlearning performance on the Tiny-ImageNet dataset, specifically configured to forget 10% of the training data via random selection. Metrics include UA, RA, TA, MIA, Avg. Gap, and RTE. 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) | Tiny-ImageNet (10% random data forgetting) | | Metric | Unlearning Accuracy (UA) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [Unlearning Accuracy (UA)](): Higher is better; unit: PERCENT. - [Retraining Accuracy (RA)](): Higher is better; unit: PERCENT. - [Target Accuracy (TA)](): Higher is better; unit: PERCENT. - [Membership Inference Attack Success Rate (MIA)](): Lower is better; unit: PERCENT. - [Average Performance Gap](): Lower is better; unit: PERCENT. - [Retraining Time Error (RTE)](): Lower is better; unit: SCALAR. - [Computation Cost (PFLOPs)](): Lower is better; unit: SIZE. ## Recorded results | Method | Unlearning Accuracy (UA) | Result date | Report paper | | --- | --- | --- | --- | | NegGrad+ | 51.06 | Not recorded | 68fa789e1e209131cea795fa | | MCU-β | 49.92 | Not recorded | 68fa789e1e209131cea795fa | | RT | 45.45 | Not recorded | 68fa789e1e209131cea795fa | | MCU | 42.42 | Not recorded | 68fa789e1e209131cea795fa | | RL | 38.59 | Not recorded | 68fa789e1e209131cea795fa | | NegGrad+ | 38.01 | Not recorded | 68fa7c441e209131cea8d47a | | l1-sparse | 36.96 | Not recorded | 68fa7c441e209131cea8d47a | | SalUn | 36.61 | Not recorded | 68fa789e1e209131cea795fa | | Retrain | 36.16 | Not recorded | 68fa7c441e209131cea8d47a | | NoT | 35.64 | Not recorded | 68fa7c441e209131cea8d47a | | CoUn | 35.1 | Not recorded | 68fa7c441e209131cea8d47a | | SalUn | 34.03 | Not recorded | 68fa7c441e209131cea8d47a | | FT | 32.76 | Not recorded | 68fa7c441e209131cea8d47a | | SFRon | 28.87 | Not recorded | 68fa789e1e209131cea795fa | | FT | 5.76 | Not recorded | 68fa789e1e209131cea795fa | | GA | 5.17 | Not recorded | 68fa789e1e209131cea795fa | | NegTV | 0.81 | Not recorded | 68fa789e1e209131cea795fa | ### NegGrad+ Unlearning Accuracy (UA): 51.06. Result ID: 6a0450f647f34ce924eaca27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### MCU-β Unlearning Accuracy (UA): 49.92. Result ID: 6a0450f647f34ce924eaca2c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### RT Unlearning Accuracy (UA): 45.45. Result ID: 6a0450f647f34ce924eaca23. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### MCU Unlearning Accuracy (UA): 42.42. Result ID: 6a0450f647f34ce924eaca2b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### RL Unlearning Accuracy (UA): 38.59. Result ID: 6a0450f647f34ce924eaca25. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### NegGrad+ Unlearning Accuracy (UA): 38.01. Result ID: 6a0c39ebfbc373592c02f03e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### l1-sparse Unlearning Accuracy (UA): 36.96. Result ID: 6a0c39ebfbc373592c02f03f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### SalUn Unlearning Accuracy (UA): 36.61. Result ID: 6a0450f647f34ce924eaca29. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### Retrain Unlearning Accuracy (UA): 36.16. Result ID: 6a0c39ebfbc373592c02f03c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### NoT Unlearning Accuracy (UA): 35.64. Result ID: 6a0c39ebfbc373592c02f041. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### CoUn Unlearning Accuracy (UA): 35.1. Result ID: 6a0c39ebfbc373592c02f042. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### SalUn Unlearning Accuracy (UA): 34.03. Result ID: 6a0c39ebfbc373592c02f040. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### FT Unlearning Accuracy (UA): 32.76. Result ID: 6a0c39ebfbc373592c02f03d. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ResNet-18"}. This record does not assert independent reproduction. ### SFRon Unlearning Accuracy (UA): 28.87. Result ID: 6a0450f647f34ce924eaca28. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### FT Unlearning Accuracy (UA): 5.76. Result ID: 6a0450f647f34ce924eaca24. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### GA Unlearning Accuracy (UA): 5.17. Result ID: 6a0450f647f34ce924eaca26. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. ### NegTV Unlearning Accuracy (UA): 0.81. Result ID: 6a0450f647f34ce924eaca2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"VGG-16-BN","Forgetting Ratio":"10%"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.