# Adversarial Defense on CIFAR-10 (test) Source: [Adversarial Defense on CIFAR-10 (test)]() Collection index: [llms.txt]() Data updated: 2026-05-27T09:44:44.073Z Evaluation of model robustness against adversarial examples on the CIFAR-10 test set, reporting classification accuracy metrics under adversarial conditions. 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 | Adversarial Defense | | Dataset (including any stated split) | CIFAR-10 (test) | | Metric | Standard Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 30 of 30 | | 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 - [Standard Accuracy](): Higher is better; unit: PERCENT. - [Robust Accuracy](): Higher is better; unit: PERCENT. - [Robust Accuracy (ℓ∞)](): Higher is better; unit: PERCENT. - [Robust Accuracy (ℓ2)](): Higher is better; unit: PERCENT. - [Robust Accuracy (ℓ1)](): Higher is better; unit: PERCENT. - [Robust Accuracy (Union)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Standard Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Wang et al. (2022) + SSNI-L | 93.62 | Not recorded | 68fa792d1e209131cea7c6e4 | | Lee & Kim (2023) + SSNI-L | 93.49 | Not recorded | 68fa792d1e209131cea7c6e4 | | Nie et al. (2022) + SSNI-L | 92.97 | Not recorded | 68fa792d1e209131cea7c6e4 | | Wang et al. (2022) | 92.45 | Not recorded | 68fa792d1e209131cea7c6e4 | | Lee & Kim (2023) | 90.1 | Not recorded | 68fa792d1e209131cea7c6e4 | | Nie et al. (2022) | 89.71 | Not recorded | 68fa792d1e209131cea7c6e4 | | Muon | 89.66 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 87.74 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 87.58 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 86.68 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 85.74 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 85.1 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 84.81 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 84.08 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 84.08 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 82.48 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 82.37 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 81.22 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 80.74 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 79.36 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 79.21 | Not recorded | 6a167ac5dbd7150506f80181 | | SGD | 78.6 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 78.04 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 77.48 | Not recorded | 6a167ac5dbd7150506f80181 | | Muon | 76.3 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 71.5 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 69.98 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 69.22 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 66.26 | Not recorded | 6a167ac5dbd7150506f80181 | | AdamW | 65.18 | Not recorded | 6a167ac5dbd7150506f80181 | ### Wang et al. (2022) + SSNI-L Standard Accuracy: 93.62. Result ID: 699a48025698665ec4cf58e3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Classifier Backbone":"WRN-28-10","Attack Type":"BPDA+EOT ℓ∞","Attack Constraint (ϵ)":"8/255","Defense Method":"SSNI-L"}. This record does not assert independent reproduction. ### Lee & Kim (2023) + SSNI-L Standard Accuracy: 93.49. Result ID: 699a48025698665ec4cf58e5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Classifier Backbone":"WRN-28-10","Attack Type":"BPDA+EOT ℓ∞","Attack Constraint (ϵ)":"8/255","Defense Method":"SSNI-L"}. This record does not assert independent reproduction. ### Nie et al. (2022) + SSNI-L Standard Accuracy: 92.97. Result ID: 699a48025698665ec4cf58e1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Classifier Backbone":"WRN-28-10","Attack Type":"BPDA+EOT ℓ∞","Attack Constraint (ϵ)":"8/255","Defense Method":"SSNI-L"}. This record does not assert independent reproduction. ### Wang et al. (2022) Standard Accuracy: 92.45. Result ID: 699a48025698665ec4cf58e2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Classifier Backbone":"WRN-28-10","Attack Type":"BPDA+EOT ℓ∞","Attack Constraint (ϵ)":"8/255"}. This record does not assert independent reproduction. ### Lee & Kim (2023) Standard Accuracy: 90.1. Result ID: 699a48025698665ec4cf58e4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Classifier Backbone":"WRN-28-10","Attack Type":"BPDA+EOT ℓ∞","Attack Constraint (ϵ)":"8/255"}. This record does not assert independent reproduction. ### Nie et al. (2022) Standard Accuracy: 89.71. Result ID: 699a48025698665ec4cf58e0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Classifier Backbone":"WRN-28-10","Attack Type":"BPDA+EOT ℓ∞","Attack Constraint (ϵ)":"8/255"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 89.66. Result ID: 6a16bcf8fbc373592c104d55. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ2","Optimizer":"Muon","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 87.74. Result ID: 6a16bcf8fbc373592c104d51. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ2","Optimizer":"SGD","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 87.58. Result ID: 6a16bcf8fbc373592c104d50. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ2","Optimizer":"SGD","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 86.68. Result ID: 6a16bcf8fbc373592c104d4f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ1","Optimizer":"Muon","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 85.74. Result ID: 6a16bcf8fbc373592c104d49. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞","Optimizer":"Muon","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 85.1. Result ID: 6a16bcf8fbc373592c104d54. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ2","Optimizer":"Muon","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 84.81. Result ID: 6a16bcf8fbc373592c104d43. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞+ℓ1","Optimizer":"Muon","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 84.08. Result ID: 6a16bcf8fbc373592c104d4a. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ1","Optimizer":"SGD","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 84.08. Result ID: 6a16bcf8fbc373592c104d4b. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ1","Optimizer":"SGD","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 82.48. Result ID: 6a16bcf8fbc373592c104d53. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ2","Optimizer":"AdamW","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 82.37. Result ID: 6a16bcf8fbc373592c104d45. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ∞","Optimizer":"SGD","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 81.22. Result ID: 6a16bcf8fbc373592c104d3f. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ∞+ℓ1","Optimizer":"SGD","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 80.74. Result ID: 6a16bcf8fbc373592c104d52. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ2","Optimizer":"AdamW","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 79.36. Result ID: 6a16bcf8fbc373592c104d4d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ1","Optimizer":"AdamW","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 79.21. Result ID: 6a16bcf8fbc373592c104d3e. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ∞+ℓ1","Optimizer":"SGD","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### SGD Standard Accuracy: 78.6. Result ID: 6a16bcf8fbc373592c104d44. [Reporting paper]() Recorded evaluation configuration: {"Norm":"ℓ∞","Optimizer":"SGD","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 78.04. Result ID: 6a16bcf8fbc373592c104d48. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞","Optimizer":"Muon","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 77.48. Result ID: 6a16bcf8fbc373592c104d42. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞+ℓ1","Optimizer":"Muon","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### Muon Standard Accuracy: 76.3. Result ID: 6a16bcf8fbc373592c104d4e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ1","Optimizer":"Muon","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 71.5. Result ID: 6a16bcf8fbc373592c104d4c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ1","Optimizer":"AdamW","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 69.98. Result ID: 6a16bcf8fbc373592c104d41. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞+ℓ1","Optimizer":"AdamW","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 69.22. Result ID: 6a16bcf8fbc373592c104d47. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞","Optimizer":"AdamW","Checkpoint":"Last","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 66.26. Result ID: 6a16bcf8fbc373592c104d40. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞+ℓ1","Optimizer":"AdamW","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. ### AdamW Standard Accuracy: 65.18. Result ID: 6a16bcf8fbc373592c104d46. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Norm":"ℓ∞","Optimizer":"AdamW","Checkpoint":"Best","Backbone":"PreActResNet-18"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.