# Image Classification on CIFAR-10 L2 norm (epsilon=0.5, test) Source: [Image Classification on CIFAR-10 L2 norm (epsilon=0.5, test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:34:27.226Z Evaluation of Image Classification robustness on the CIFAR-10 test set under an L2 adversarial threat model with a perturbation constraint of epsilon=0.5, reporting standard, PGD, and AutoAttack accuracies. 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) | CIFAR-10 L-2 norm, epsilon=0.5 (test) | | Metric | Accuracy (Standard) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 6 of 6 | | 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 - [Accuracy (Standard)](): Higher is better; unit: PERCENT. - [Accuracy (PGD)](): Higher is better; unit: PERCENT. - [Accuracy (AutoAttack)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy (Standard) | Result date | Report paper | | --- | --- | --- | --- | | Rebuffi et al. | 95.74 | Not recorded | 695c96f953fdd5c361743643 | | Gowal et al. | 94.74 | Not recorded | 695c96f953fdd5c361743643 | | Augustin et al. | 93.96 | Not recorded | 695c96f953fdd5c361743643 | | Rebuffi et al. | 92.41 | Not recorded | 695c96f953fdd5c361743643 | | Rebuffi et al. | 91.79 | Not recorded | 695c96f953fdd5c361743643 | | Sehwag et al. | 90.93 | Not recorded | 695c96f953fdd5c361743643 | ### Rebuffi et al. Accuracy (Standard): 95.74. Result ID: 6996f6e75698665ec4439e38. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"WRN-70-16","Defense Type":"Adversarial Training (AT)","Extra data used":"true"}. This record does not assert independent reproduction. ### Gowal et al. Accuracy (Standard): 94.74. Result ID: 6996f6e75698665ec4439e37. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"WRN-70-16","Defense Type":"Adversarial Training (AT)","Extra data used":"true"}. This record does not assert independent reproduction. ### Augustin et al. Accuracy (Standard): 93.96. Result ID: 6996f6e75698665ec4439e34. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"WRN-28-10","Defense Type":"Adversarial Training (AT)"}. This record does not assert independent reproduction. ### Rebuffi et al. Accuracy (Standard): 92.41. Result ID: 6996f6e75698665ec4439e36. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"WRN-70-16","Defense Type":"Adversarial Training (AT)"}. This record does not assert independent reproduction. ### Rebuffi et al. Accuracy (Standard): 91.79. Result ID: 6996f6e75698665ec4439e35. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"WRN-28-10","Defense Type":"Adversarial Training (AT)","Extra data used":"true"}. This record does not assert independent reproduction. ### Sehwag et al. Accuracy (Standard): 90.93. Result ID: 6996f6e75698665ec4439e33. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"WRN-28-10","Defense Type":"Adversarial Training (AT)"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.