{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HDDDDXTOP7FLFGV4M4SFA4O3KK","short_pith_number":"pith:HDDDDXTO","schema_version":"1.0","canonical_sha256":"38c631de6e7fcab29abc67245071db52afbe9e6fd4d873ced4dc7d1bc44b7d52","source":{"kind":"arxiv","id":"2603.08290","version":2},"attestation_state":"computed","paper":{"title":"Minor First, Major Last: A Depth-Induced Implicit Bias of Sharpness-Aware Minimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chaewon Moon, Chulhee Yun, Dongkuk Si","submitted_at":"2026-03-09T12:09:14Z","abstract_excerpt":"We study the implicit bias of Sharpness-Aware Minimization (SAM) when training $L$-layer linear diagonal networks on linearly separable binary classification. For linear models ($L=1$), both $\\ell_\\infty$- and $\\ell_2$-SAM recover the $\\ell_2$ max-margin classifier, matching gradient descent (GD). However, for depth $L = 2$, the behavior changes drastically -- even on a single-example dataset. For $\\ell_\\infty$-SAM, the limit direction depends critically on initialization and can converge to $\\mathbf{0}$ or to any standard basis vector, in stark contrast to GD, whose limit aligns with the basi"},"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":"2603.08290","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-09T12:09:14Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2d7aa5c07925957298f40595591e3b4e7f7442717b74019a3df71867cd1ce6ef","abstract_canon_sha256":"0bc6391c39ca6feea5f08f36d18f8d2951af3612bc3af6935fd868005b5f5d6d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:42.901625Z","signature_b64":"OUUuh1TW7jbVFmT3X05EBlLHXA4dVz9dg8zxGRdz29cZ6dYQRB7saIvj9oGRodlk7EN3gTSrkmHGtoh2ARgUAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"38c631de6e7fcab29abc67245071db52afbe9e6fd4d873ced4dc7d1bc44b7d52","last_reissued_at":"2026-05-20T00:05:42.901076Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:42.901076Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Minor First, Major Last: A Depth-Induced Implicit Bias of Sharpness-Aware Minimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chaewon Moon, Chulhee Yun, Dongkuk Si","submitted_at":"2026-03-09T12:09:14Z","abstract_excerpt":"We study the implicit bias of Sharpness-Aware Minimization (SAM) when training $L$-layer linear diagonal networks on linearly separable binary classification. For linear models ($L=1$), both $\\ell_\\infty$- and $\\ell_2$-SAM recover the $\\ell_2$ max-margin classifier, matching gradient descent (GD). However, for depth $L = 2$, the behavior changes drastically -- even on a single-example dataset. For $\\ell_\\infty$-SAM, the limit direction depends critically on initialization and can converge to $\\mathbf{0}$ or to any standard basis vector, in stark contrast to GD, whose limit aligns with the basi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.08290","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/2603.08290/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":"2603.08290","created_at":"2026-05-20T00:05:42.901149+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.08290v2","created_at":"2026-05-20T00:05:42.901149+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.08290","created_at":"2026-05-20T00:05:42.901149+00:00"},{"alias_kind":"pith_short_12","alias_value":"HDDDDXTOP7FL","created_at":"2026-05-20T00:05:42.901149+00:00"},{"alias_kind":"pith_short_16","alias_value":"HDDDDXTOP7FLFGV4","created_at":"2026-05-20T00:05:42.901149+00:00"},{"alias_kind":"pith_short_8","alias_value":"HDDDDXTO","created_at":"2026-05-20T00:05:42.901149+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK","json":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK.json","graph_json":"https://pith.science/api/pith-number/HDDDDXTOP7FLFGV4M4SFA4O3KK/graph.json","events_json":"https://pith.science/api/pith-number/HDDDDXTOP7FLFGV4M4SFA4O3KK/events.json","paper":"https://pith.science/paper/HDDDDXTO"},"agent_actions":{"view_html":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK","download_json":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK.json","view_paper":"https://pith.science/paper/HDDDDXTO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.08290&json=true","fetch_graph":"https://pith.science/api/pith-number/HDDDDXTOP7FLFGV4M4SFA4O3KK/graph.json","fetch_events":"https://pith.science/api/pith-number/HDDDDXTOP7FLFGV4M4SFA4O3KK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK/action/storage_attestation","attest_author":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK/action/author_attestation","sign_citation":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK/action/citation_signature","submit_replication":"https://pith.science/pith/HDDDDXTOP7FLFGV4M4SFA4O3KK/action/replication_record"}},"created_at":"2026-05-20T00:05:42.901149+00:00","updated_at":"2026-05-20T00:05:42.901149+00:00"}