{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:ZOL5FKOLYKPZ5RXQ2BYMSQHQEC","short_pith_number":"pith:ZOL5FKOL","schema_version":"1.0","canonical_sha256":"cb97d2a9cbc29f9ec6f0d070c940f02089ffc8ae327ef4e68e1041d07ce6f4dc","source":{"kind":"arxiv","id":"2310.06112","version":2},"attestation_state":"computed","paper":{"title":"Theoretical Analysis of Robust Overfitting for Wide DNNs: An NTK Approach","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Di Wang, Shaopeng Fu","submitted_at":"2023-10-09T19:40:25Z","abstract_excerpt":"Adversarial training (AT) is a canonical method for enhancing the robustness of deep neural networks (DNNs). However, recent studies empirically demonstrated that it suffers from robust overfitting, i.e., a long time AT can be detrimental to the robustness of DNNs. This paper presents a theoretical explanation of robust overfitting for DNNs. Specifically, we non-trivially extend the neural tangent kernel (NTK) theory to AT and prove that an adversarially trained wide DNN can be well approximated by a linearized DNN. Moreover, for squared loss, closed-form AT dynamics for the linearized DNN can"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2310.06112","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-10-09T19:40:25Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"8b6e3c05225d54e8b7eaaefb927ae2963a4433e31f01e4f1ce39efc1cedfe3c8","abstract_canon_sha256":"41cd7339b10df8244fbf6e8c0599b2cd314d9819ab7fea4ff6a4a635bf612c09"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:40:57.832186Z","signature_b64":"7CFNZ/ziG/KFamonzqUPh9ezAqeVBDVfoviI7XEvnDk4N07o6T7s2pYHWUPg5fxiWicVKJTI3No2ABNxFokkAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cb97d2a9cbc29f9ec6f0d070c940f02089ffc8ae327ef4e68e1041d07ce6f4dc","last_reissued_at":"2026-07-05T07:40:57.831741Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:40:57.831741Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Theoretical Analysis of Robust Overfitting for Wide DNNs: An NTK Approach","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Di Wang, Shaopeng Fu","submitted_at":"2023-10-09T19:40:25Z","abstract_excerpt":"Adversarial training (AT) is a canonical method for enhancing the robustness of deep neural networks (DNNs). However, recent studies empirically demonstrated that it suffers from robust overfitting, i.e., a long time AT can be detrimental to the robustness of DNNs. This paper presents a theoretical explanation of robust overfitting for DNNs. Specifically, we non-trivially extend the neural tangent kernel (NTK) theory to AT and prove that an adversarially trained wide DNN can be well approximated by a linearized DNN. Moreover, for squared loss, closed-form AT dynamics for the linearized DNN can"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.06112","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2310.06112/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2310.06112","created_at":"2026-07-05T07:40:57.831804+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.06112v2","created_at":"2026-07-05T07:40:57.831804+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.06112","created_at":"2026-07-05T07:40:57.831804+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZOL5FKOLYKPZ","created_at":"2026-07-05T07:40:57.831804+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZOL5FKOLYKPZ5RXQ","created_at":"2026-07-05T07:40:57.831804+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZOL5FKOL","created_at":"2026-07-05T07:40:57.831804+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.19724","citing_title":"Benign Overfitting in Adversarial Training for Vision Transformers","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC","json":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC.json","graph_json":"https://pith.science/api/pith-number/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/graph.json","events_json":"https://pith.science/api/pith-number/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/events.json","paper":"https://pith.science/paper/ZOL5FKOL"},"agent_actions":{"view_html":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC","download_json":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC.json","view_paper":"https://pith.science/paper/ZOL5FKOL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.06112&json=true","fetch_graph":"https://pith.science/api/pith-number/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/graph.json","fetch_events":"https://pith.science/api/pith-number/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/action/storage_attestation","attest_author":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/action/author_attestation","sign_citation":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/action/citation_signature","submit_replication":"https://pith.science/pith/ZOL5FKOLYKPZ5RXQ2BYMSQHQEC/action/replication_record"}},"created_at":"2026-07-05T07:40:57.831804+00:00","updated_at":"2026-07-05T07:40:57.831804+00:00"}