# Adversarial Robustness on CIFAR-10 (L-inf, eps=4/255) Source: [Adversarial Robustness on CIFAR-10 (L-inf, eps=4/255)]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:02:22.956Z Evaluation of model adversarial robustness on the CIFAR-10 dataset under an L-infinity perturbation constraint with epsilon set to 4/255. Metrics focus on specific robust accuracy scores (AA, A3) and associated runtimes. 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 Robustness | | Dataset (including any stated split) | CIFAR-10 L-infinity, epsilon=4/255 | | Metric | Robust Accuracy (AA) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 10 of 10 | | 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 - [Robust Accuracy (AA)](): Higher is better; unit: PERCENT. - [Robust Accuracy (A3)](): Higher is better; unit: PERCENT. - [Robust Accuracy Delta](): Lower is better; unit: SCALAR. - [Runtime (AA)](): Lower is better; unit: TIME. - [Runtime (A3)](): Lower is better; unit: TIME. - [Speed-up](): Higher is better; unit: SCALAR. - [Search Iterations (A3)](): Lower is better; unit: SCALAR. ## Recorded results | Method | Robust Accuracy (AA) | Result date | Report paper | | --- | --- | --- | --- | | Xiao et al. (2020) ADV | 64.91 | Not recorded | 68fa6b161e209131cea2c0c7 | | Xiao et al. (2020) ADV | 64.91 | Not recorded | 68fa6b161e209131cea2c0c7 | | Madry et al. (2018) | 44.78 | Not recorded | 68fa6b161e209131cea2c0c7 | | Guo et al. (2018) | 22.3 | Not recorded | 68fa6b161e209131cea2c0c7 | | Xiao et al. (2020) | 19.82 | Not recorded | 68fa6b161e209131cea2c0c7 | | Metzen et al. (2017) | 6.17 | Not recorded | 68fa6b161e209131cea2c0c7 | | Pang et al. (2019) | 4.14 | Not recorded | 68fa6b161e209131cea2c0c7 | | Papernot et al. (2015) | 2.85 | Not recorded | 68fa6b161e209131cea2c0c7 | | Buckman et al. (2018) | 2.29 | Not recorded | 68fa6b161e209131cea2c0c7 | | Das et al. (2017) + Lee et al. (2018) | 0.59 | Not recorded | 68fa6b161e209131cea2c0c7 | ### Xiao et al. (2020) ADV Robust Accuracy (AA): 64.91. Result ID: 698e5513ffc3901de09cb201. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"training":"adversarial"}. This record does not assert independent reproduction. ### Xiao et al. (2020) ADV Robust Accuracy (AA): 64.91. Result ID: 698e5513ffc3901de09cb202. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"training":"adversarial","attack\_time\_budget\_A3":"T=30"}. This record does not assert independent reproduction. ### Madry et al. (2018) Robust Accuracy (AA): 44.78. Result ID: 698e5513ffc3901de09cb1f9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Guo et al. (2018) Robust Accuracy (AA): 22.3. Result ID: 698e5513ffc3901de09cb1fd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Xiao et al. (2020) Robust Accuracy (AA): 19.82. Result ID: 698e5513ffc3901de09cb200. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Metzen et al. (2017) Robust Accuracy (AA): 6.17. Result ID: 698e5513ffc3901de09cb1fc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Pang et al. (2019) Robust Accuracy (AA): 4.14. Result ID: 698e5513ffc3901de09cb1fe. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"non-differentiable":true}. This record does not assert independent reproduction. ### Papernot et al. (2015) Robust Accuracy (AA): 2.85. Result ID: 698e5513ffc3901de09cb1ff. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Buckman et al. (2018) Robust Accuracy (AA): 2.29. Result ID: 698e5513ffc3901de09cb1fa. [Reporting paper]() Recorded evaluation configuration: {"non-differentiable":true}. This record does not assert independent reproduction. ### Das et al. (2017) + Lee et al. (2018) Robust Accuracy (AA): 0.59. Result ID: 698e5513ffc3901de09cb1fb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.