{"about":{"non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","site":"https://codewithpapers.app","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/paper/pyramid-adversarial-training-improves-vit","title":"Pyramid Adversarial Training Improves ViT Performance","arxiv_id":"2111.15121","date":"2021-11-30","proceeding":"CVPR 2022 1","authors":["Charles Herrmann","Kyle Sargent","Lu Jiang","Ramin Zabih","Huiwen Chang","Ce Liu","Dilip Krishnan","Deqing Sun"],"abstract":"Aggressive data augmentation is a key component of the strong generalization capabilities of Vision Transformer (ViT). One such data augmentation technique is adversarial training (AT); however, many prior works have shown that this often results in poor clean accuracy. In this work, we present pyramid adversarial training (PyramidAT), a simple and effective technique to improve ViT's overall performance. We pair it with a \"matched\" Dropout and stochastic depth regularization, which adopts the same Dropout and stochastic depth configuration for the clean and adversarial samples. Similar to the improvements on CNNs by AdvProp (not directly applicable to ViT), our pyramid adversarial training breaks the trade-off between in-distribution accuracy and out-of-distribution robustness for ViT and related architectures. It leads to 1.82% absolute improvement on ImageNet clean accuracy for the ViT-B model when trained only on ImageNet-1K data, while simultaneously boosting performance on 7 ImageNet robustness metrics, by absolute numbers ranging from 1.76% to 15.68%. We set a new state-of-the-art for ImageNet-C (41.42 mCE), ImageNet-R (53.92%), and ImageNet-Sketch (41.04%) without extra data, using only the ViT-B/16 backbone and our pyramid adversarial training. Our code is publicly available at pyramidat.github.io.","url_abs":"https://arxiv.org/abs/2111.15121v2","url_pdf":"https://arxiv.org/pdf/2111.15121v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"pyramid-adversarial-training-improves-vit","repo_url":"https://github.com/google-research/scenic/tree/main/scenic/projects/adversarialtraining","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"jax","reach":null}],"tasks":[{"task_slug":"adversarial-attack","task_name":"Adversarial Attack"},{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"domain-generalization","task_name":"Domain Generalization"},{"task_slug":"image-classification","task_name":"Image Classification"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"advprop","method_name":"AdvProp"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"auxiliary-batch-normalization","method_name":"Auxiliary Batch Normalization"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"stochastic-depth","method_name":"Stochastic Depth"},{"method_slug":"transformer","method_name":"Transformer"},{"method_slug":"vision-transformer","method_name":"Vision Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/domain-generalization-on-imagenet-a","task":"Domain Generalization","dataset":"ImageNet-A","model":"Pyramid Adversarial Training Improves ViT (Im21k)","rank_in_archive_order":13,"of":39,"metrics":{"Top-1 accuracy %":"62.44"},"uses_additional_data":true},{"leaderboard":"/sota/domain-generalization-on-imagenet-a","task":"Domain Generalization","dataset":"ImageNet-A","model":"Pyramid Adversarial Training Improves ViT (384x384)","rank_in_archive_order":24,"of":39,"metrics":{"Top-1 accuracy %":"36.41"},"uses_additional_data":false},{"leaderboard":"/sota/domain-generalization-on-imagenet-c","task":"Domain Generalization","dataset":"ImageNet-C","model":"Pyramid Adversarial Training Improves ViT (Im21k)","rank_in_archive_order":9,"of":47,"metrics":{"Number of params":"87M","mean Corruption Error (mCE)":"36.80"},"uses_additional_data":true},{"leaderboard":"/sota/domain-generalization-on-imagenet-c","task":"Domain Generalization","dataset":"ImageNet-C","model":"Pyramid Adversarial Training Improves ViT","rank_in_archive_order":15,"of":47,"metrics":{"mean Corruption Error (mCE)":"41.42"},"uses_additional_data":false},{"leaderboard":"/sota/domain-generalization-on-imagenet-r","task":"Domain Generalization","dataset":"ImageNet-R","model":"Pyramid Adversarial Training Improves ViT (Im21k)","rank_in_archive_order":17,"of":39,"metrics":{"Top-1 Error Rate":"42.16"},"uses_additional_data":true},{"leaderboard":"/sota/domain-generalization-on-imagenet-r","task":"Domain Generalization","dataset":"ImageNet-R","model":"Pyramid Adversarial Training Improves ViT","rank_in_archive_order":22,"of":39,"metrics":{"Top-1 Error Rate":"46.08"},"uses_additional_data":false},{"leaderboard":"/sota/domain-generalization-on-imagenet-sketch","task":"Domain Generalization","dataset":"ImageNet-Sketch","model":"Pyramid Adversarial Training Improves ViT (Im21k)","rank_in_archive_order":14,"of":20,"metrics":{"Top-1 accuracy":"46.03"},"uses_additional_data":true},{"leaderboard":"/sota/domain-generalization-on-imagenet-sketch","task":"Domain Generalization","dataset":"ImageNet-Sketch","model":"Pyramid Adversarial Training Improves ViT","rank_in_archive_order":18,"of":20,"metrics":{"Top-1 accuracy":"41.04"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"ViT-B/16 (512x512) + Pyramid","rank_in_archive_order":27,"of":106,"metrics":{"Top-1 Accuracy":"49.39"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"ViT-B/16 (512x512) + Pixel","rank_in_archive_order":31,"of":106,"metrics":{"Top-1 Accuracy":"47.53"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"ViT-B/16 (512x512)","rank_in_archive_order":34,"of":106,"metrics":{"Top-1 Accuracy":"46.68"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT on 384x384 + Adv Pyramid","rank_in_archive_order":43,"of":106,"metrics":{"Top-1 Accuracy":"39.79"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT on 384x384 + Adv Pixel","rank_in_archive_order":47,"of":106,"metrics":{"Top-1 Accuracy":"37.41"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT on 384x384","rank_in_archive_order":53,"of":106,"metrics":{"Top-1 Accuracy":"35.59"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT on 384x384 + Random Pyramid","rank_in_archive_order":55,"of":106,"metrics":{"Top-1 Accuracy":"34.83"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT on 384x384 + Random Pixel","rank_in_archive_order":57,"of":106,"metrics":{"Top-1 Accuracy":"34.12"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT (RandAug) + Adv Pyramid","rank_in_archive_order":58,"of":106,"metrics":{"Top-1 Accuracy":"32.92"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"Discrete ViT + Pixel","rank_in_archive_order":63,"of":106,"metrics":{"Top-1 Accuracy":"30.98"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"Discrete ViT + Pyramid","rank_in_archive_order":64,"of":106,"metrics":{"Top-1 Accuracy":"30.28"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT (RandAug) + Adv Pixel","rank_in_archive_order":65,"of":106,"metrics":{"Top-1 Accuracy":"30.11"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"Discrete ViT","rank_in_archive_order":66,"of":106,"metrics":{"Top-1 Accuracy":"29.95"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT (RandAug) + Random Pyramid","rank_in_archive_order":68,"of":106,"metrics":{"Top-1 Accuracy":"29.41"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT (RandAug)","rank_in_archive_order":69,"of":106,"metrics":{"Top-1 Accuracy":"29.3"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"RegViT (RandAug) + Random Pixel","rank_in_archive_order":72,"of":106,"metrics":{"Top-1 Accuracy":"28.72"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"MLP-Mixer + Pyramid","rank_in_archive_order":73,"of":106,"metrics":{"Top-1 Accuracy":"28.6"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"MLP-Mixer","rank_in_archive_order":78,"of":106,"metrics":{"Top-1 Accuracy":"25.9"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"ViT + MixUp","rank_in_archive_order":80,"of":106,"metrics":{"Top-1 Accuracy":"25.65"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"MLP-Mixer + Pixel","rank_in_archive_order":82,"of":106,"metrics":{"Top-1 Accuracy":"24.75"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"ViT + CutMix","rank_in_archive_order":87,"of":106,"metrics":{"Top-1 Accuracy":"21.61"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-objectnet","task":"Image Classification","dataset":"ObjectNet","model":"ViT","rank_in_archive_order":96,"of":106,"metrics":{"Top-1 Accuracy":"17.36"},"uses_additional_data":true}],"syntology":{"syntology_url":"https://syntology.ai/paper/2111.15121","atlas_url":"https://app.syntology.ai/?focus=2111.15121","mcp":null,"developers":"https://syntology.ai/developers","held":true,"reason":"arxiv_id_in_atlas","checked_against":{"atlas_ids":549449,"atlas_ids_sha256":"afd72399dafd9906e9062927f70655952385a3d75488558e1f0484e950ab09b7","read_at":"2026-10-05T01:46:21Z"},"agent_calls":[{"tool":"get_citation_path","arguments":{"paper_1":"2111.15121"},"arguments_in":{"paper_2":"another paper's arXiv id or title"},"call":"get_citation_path(paper_1=\"2111.15121\", paper_2=\"…\")"},{"tool":"get_concepts_for_paper","arguments":{"arxiv_id":"2111.15121"},"call":"get_concepts_for_paper(arxiv_id=\"2111.15121\")"}],"by_repo":[{"repo":"google-research/scenic","kind":"official","samples":0,"ran":0,"instrument":0,"unverified":0,"harvested_for_other_papers":6,"harvested_for":["2106.11297","2202.02200","2212.02400","2301.13195","2304.06708","2402.02887"],"harvested_for_calls":[{"arxiv_id":"2106.11297","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2106.11297"},"call":"get_harvested_code_for_paper(arxiv_id=\"2106.11297\")"},{"arxiv_id":"2202.02200","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2202.02200"},"call":"get_harvested_code_for_paper(arxiv_id=\"2202.02200\")"},{"arxiv_id":"2212.02400","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2212.02400"},"call":"get_harvested_code_for_paper(arxiv_id=\"2212.02400\")"},{"arxiv_id":"2301.13195","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2301.13195"},"call":"get_harvested_code_for_paper(arxiv_id=\"2301.13195\")"},{"arxiv_id":"2304.06708","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2304.06708"},"call":"get_harvested_code_for_paper(arxiv_id=\"2304.06708\")"},{"arxiv_id":"2402.02887","tool":"get_harvested_code_for_paper","arguments":{"arxiv_id":"2402.02887"},"call":"get_harvested_code_for_paper(arxiv_id=\"2402.02887\")"}]}],"by_repo_harvested_for_is":"harvested_for: the arXiv ids (as Syntology's graph holds them) of every paper whose samples Syntology's graph links to this repository, in arXiv-id order, complete; the page names the first 3 and says how many more. A link only: it does not say the papers are the same work, that one copied the other, or that the code implements this paper's method; harvested_for_calls gives the get_harvested_code_for_paper call for each arXiv-shaped id among them","by_repo_is":"one row per repository in the archive's code links, Syntology's graph links or the samples; samples 0 means no sample is linked to this paper, a statement about Syntology's coverage, not about the repository: with harvested_for_other_papers present, Syntology harvested the repository for that many other papers, and without it, Syntology harvested nothing from it; placed_by_identical_code counts samples whose link records no repository, placed here because identical code was harvested from this repository for another paper; ran counts executions on a synthesized input and instrument counts failures of Syntology's instrument, not of the code"},"code_links_note":null,"arxiv_metadata":null,"syntology_extracted_results":null,"record_sha256":"39afb0c05df05681f1b081847ca542ef47a4964d599f6e95eb3fa844039ff759","record_changed_at":"2026-09-28","record_changed_at_basis":"first_hashed"}