# Image Classification on ImageNet-1K (val) (Adversarial Robustness Profile) Source: [Image Classification on ImageNet-1K (val) (Adversarial Robustness Profile)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:39:17.259Z Evaluation of Image Classification performance on the validation split of ImageNet-1K, reporting clean accuracy alongside robustness metrics against various adversarial attacks including PGD, CW, and APGD variants. 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) | ImageNet-1K (val) | | Metric | Clean Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 - [Clean Accuracy](): Higher is better; unit: PERCENT. - [PGD Robust Accuracy](): Higher is better; unit: PERCENT. - [CW Robust Accuracy](): Higher is better; unit: PERCENT. - [APGD-DLR Robust Accuracy](): Higher is better; unit: PERCENT. - [APGD-CE Robust Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Clean Accuracy | Result date | Report paper | | --- | --- | --- | --- | | TAPADL-FAN | 84.3 | Not recorded | 68fa75d71e209131cea6a300 | | FAN-B-AdaNCA | 84.1 | Not recorded | 68fa75d71e209131cea6a300 | | FAN-B | 83.9 | Not recorded | 68fa75d71e209131cea6a300 | | Swin-B-AdaNCA | 83.7 | Not recorded | 68fa75d71e209131cea6a300 | | Swin-B | 83.4 | Not recorded | 68fa75d71e209131cea6a300 | | RVT-B-AdaNCA | 83.3 | Not recorded | 68fa75d71e209131cea6a300 | | Swin-B | 83.3 | Not recorded | 68fa75d71e209131cea6a300 | | ConViT-B-AdaNCA | 83.2 | Not recorded | 68fa75d71e209131cea6a300 | | TAPADL-RVT | 83.1 | Not recorded | 68fa75d71e209131cea6a300 | | RVT-B | 82.7 | Not recorded | 68fa75d71e209131cea6a300 | | ConViT-B | 82.7 | Not recorded | 68fa75d71e209131cea6a300 | | ConViT-B | 82.4 | Not recorded | 68fa75d71e209131cea6a300 | ### TAPADL-FAN Clean Accuracy: 84.3. Result ID: 6993e7b93a6c314849195953. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":50.7,"FLOPS (G)":11.8,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### FAN-B-AdaNCA Clean Accuracy: 84.1. Result ID: 6993e7b93a6c314849195954. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":51.7,"FLOPS (G)":12.4,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### FAN-B Clean Accuracy: 83.9. Result ID: 6993e7b93a6c314849195952. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params (M)":50.4,"FLOPS (G)":11.7,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### Swin-B-AdaNCA Clean Accuracy: 83.7. Result ID: 6993e7b93a6c314849195957. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":90.7,"FLOPS (G)":16.3,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### Swin-B Clean Accuracy: 83.4. Result ID: 6993e7b93a6c314849195955. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params (M)":87.8,"FLOPS (G)":15.4,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### RVT-B-AdaNCA Clean Accuracy: 83.3. Result ID: 6993e7b93a6c314849195951. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":91,"FLOPS (G)":19,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### Swin-B Clean Accuracy: 83.3. Result ID: 6993e7b93a6c314849195956. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params (M)":94.1,"FLOPS (G)":16.7,"Pre-training":"ImageNet-22K"}. This record does not assert independent reproduction. ### ConViT-B-AdaNCA Clean Accuracy: 83.2. Result ID: 6993e7b93a6c31484919595a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":89,"FLOPS (G)":19,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### TAPADL-RVT Clean Accuracy: 83.1. Result ID: 6993e7b93a6c314849195950. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":89.4,"FLOPS (G)":17.9,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### RVT-B Clean Accuracy: 82.7. Result ID: 6993e7b93a6c31484919594f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params (M)":88.5,"FLOPS (G)":17.7,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. ### ConViT-B Clean Accuracy: 82.7. Result ID: 6993e7b93a6c314849195959. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params (M)":93.6,"FLOPS (G)":19.2,"Pre-training":"ImageNet-22K"}. This record does not assert independent reproduction. ### ConViT-B Clean Accuracy: 82.4. Result ID: 6993e7b93a6c314849195958. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params (M)":86.5,"FLOPS (G)":17.7,"Pre-training":"ImageNet-1K"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.