{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:MFANZV7ATB4LUYJ7RTDIMXF35F","short_pith_number":"pith:MFANZV7A","schema_version":"1.0","canonical_sha256":"6140dcd7e09878ba613f8cc6865cbbe94e078951ba30e75cf4d1352c536ff9be","source":{"kind":"arxiv","id":"2306.04226","version":2},"attestation_state":"computed","paper":{"title":"Normalization Layers Are All That Sharpness-Aware Minimization Needs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"David Rolnick, Matthias Hein, Maximilian Mueller, Tiffany Vlaar","submitted_at":"2023-06-07T08:05:46Z","abstract_excerpt":"Sharpness-aware minimization (SAM) was proposed to reduce sharpness of minima and has been shown to enhance generalization performance in various settings. In this work we show that perturbing only the affine normalization parameters (typically comprising 0.1% of the total parameters) in the adversarial step of SAM can outperform perturbing all of the parameters.This finding generalizes to different SAM variants and both ResNet (Batch Normalization) and Vision Transformer (Layer Normalization) architectures. We consider alternative sparse perturbation approaches and find that these do not achi"},"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":"2306.04226","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-07T08:05:46Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"1ee0bc87285aabb7d74f2e6068f68e50dfb1d65103b5f938606dbf205337fd03","abstract_canon_sha256":"31616c58663e81f2b5bad7b896d7d90c91bddd97b6dbcb7381e22cd94c4c651a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:13:42.491156Z","signature_b64":"HIDTBhBWyu6Rr+y4Dq6EAxjbZsNTBqA9D0BB85S6CvDH8QhlJSEgbcxTax/HlzHw1RrL9/Gea+r5iw2mnrPjCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6140dcd7e09878ba613f8cc6865cbbe94e078951ba30e75cf4d1352c536ff9be","last_reissued_at":"2026-07-05T07:13:42.490723Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:13:42.490723Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Normalization Layers Are All That Sharpness-Aware Minimization Needs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"David Rolnick, Matthias Hein, Maximilian Mueller, Tiffany Vlaar","submitted_at":"2023-06-07T08:05:46Z","abstract_excerpt":"Sharpness-aware minimization (SAM) was proposed to reduce sharpness of minima and has been shown to enhance generalization performance in various settings. In this work we show that perturbing only the affine normalization parameters (typically comprising 0.1% of the total parameters) in the adversarial step of SAM can outperform perturbing all of the parameters.This finding generalizes to different SAM variants and both ResNet (Batch Normalization) and Vision Transformer (Layer Normalization) architectures. We consider alternative sparse perturbation approaches and find that these do not achi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.04226","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/2306.04226/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":"2306.04226","created_at":"2026-07-05T07:13:42.490772+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.04226v2","created_at":"2026-07-05T07:13:42.490772+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.04226","created_at":"2026-07-05T07:13:42.490772+00:00"},{"alias_kind":"pith_short_12","alias_value":"MFANZV7ATB4L","created_at":"2026-07-05T07:13:42.490772+00:00"},{"alias_kind":"pith_short_16","alias_value":"MFANZV7ATB4LUYJ7","created_at":"2026-07-05T07:13:42.490772+00:00"},{"alias_kind":"pith_short_8","alias_value":"MFANZV7A","created_at":"2026-07-05T07:13:42.490772+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.07376","citing_title":"Adapter Naturally Serves as Decoupler for Cross-Domain Few-Shot Semantic Segmentation","ref_index":14,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F","json":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F.json","graph_json":"https://pith.science/api/pith-number/MFANZV7ATB4LUYJ7RTDIMXF35F/graph.json","events_json":"https://pith.science/api/pith-number/MFANZV7ATB4LUYJ7RTDIMXF35F/events.json","paper":"https://pith.science/paper/MFANZV7A"},"agent_actions":{"view_html":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F","download_json":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F.json","view_paper":"https://pith.science/paper/MFANZV7A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.04226&json=true","fetch_graph":"https://pith.science/api/pith-number/MFANZV7ATB4LUYJ7RTDIMXF35F/graph.json","fetch_events":"https://pith.science/api/pith-number/MFANZV7ATB4LUYJ7RTDIMXF35F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F/action/storage_attestation","attest_author":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F/action/author_attestation","sign_citation":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F/action/citation_signature","submit_replication":"https://pith.science/pith/MFANZV7ATB4LUYJ7RTDIMXF35F/action/replication_record"}},"created_at":"2026-07-05T07:13:42.490772+00:00","updated_at":"2026-07-05T07:13:42.490772+00:00"}