{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ZW5XEE23LILEQUNTBC3LYIY3ZJ","short_pith_number":"pith:ZW5XEE23","schema_version":"1.0","canonical_sha256":"cdbb72135b5a164851b308b6bc231bca738ef0c76b8d84526d718d21cb1329fd","source":{"kind":"arxiv","id":"2510.10777","version":3},"attestation_state":"computed","paper":{"title":"Preconditioned Norms: A Unified Framework for Steepest Descent, Quasi-Newton and Adaptive Methods","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aleksandr Beznosikov, Aleksandr Bogdanov, Andrey Veprikov, Arman Bolatov, Martin Tak\\'a\\v{c}, Samuel Horv\\'ath, Slavomir Hanzely","submitted_at":"2025-10-12T19:39:41Z","abstract_excerpt":"Optimization lies at the core of modern deep learning, yet existing methods often face a fundamental trade-off between adapting to problem geometry and leveraging curvature utilization. Steepest descent algorithms adapt to different geometries through norm choices but remain strictly first-order, whereas quasi-Newton and adaptive optimizers incorporate curvature information but are restricted to Frobenius geometry, limiting their applicability across diverse architectures. In this work, we propose a unified framework generalizing steepest descent, quasi-Newton methods, and adaptive methods thr"},"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":"2510.10777","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-10-12T19:39:41Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"e8a2e9c2b92013e968543a044cbbad37dd2fecec485ff140d5735a893971085f","abstract_canon_sha256":"2f5f734ce898b52a1d7a8e1fc8878df44b49110137e60b6bd1acb5f5664f685e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:02:57.196079Z","signature_b64":"ddu0yP3UZ/q8sBeNbYCG6/gsVFtYLQZvuSJHZTUlVpbAeQKMzsV6DfcYg1MV3lxOJf/8Q9aS8QyO7BJWjzC4Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cdbb72135b5a164851b308b6bc231bca738ef0c76b8d84526d718d21cb1329fd","last_reissued_at":"2026-05-20T00:02:57.195309Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:02:57.195309Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Preconditioned Norms: A Unified Framework for Steepest Descent, Quasi-Newton and Adaptive Methods","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aleksandr Beznosikov, Aleksandr Bogdanov, Andrey Veprikov, Arman Bolatov, Martin Tak\\'a\\v{c}, Samuel Horv\\'ath, Slavomir Hanzely","submitted_at":"2025-10-12T19:39:41Z","abstract_excerpt":"Optimization lies at the core of modern deep learning, yet existing methods often face a fundamental trade-off between adapting to problem geometry and leveraging curvature utilization. Steepest descent algorithms adapt to different geometries through norm choices but remain strictly first-order, whereas quasi-Newton and adaptive optimizers incorporate curvature information but are restricted to Frobenius geometry, limiting their applicability across diverse architectures. In this work, we propose a unified framework generalizing steepest descent, quasi-Newton methods, and adaptive methods thr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.10777","kind":"arxiv","version":3},"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/2510.10777/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":"2510.10777","created_at":"2026-05-20T00:02:57.195445+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.10777v3","created_at":"2026-05-20T00:02:57.195445+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.10777","created_at":"2026-05-20T00:02:57.195445+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZW5XEE23LILE","created_at":"2026-05-20T00:02:57.195445+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZW5XEE23LILEQUNT","created_at":"2026-05-20T00:02:57.195445+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZW5XEE23","created_at":"2026-05-20T00:02:57.195445+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.18106","citing_title":"Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers","ref_index":150,"is_internal_anchor":true},{"citing_arxiv_id":"2605.19811","citing_title":"LionMuon: Alternating Spectral and Sign Descent for Efficient Training","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ","json":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ.json","graph_json":"https://pith.science/api/pith-number/ZW5XEE23LILEQUNTBC3LYIY3ZJ/graph.json","events_json":"https://pith.science/api/pith-number/ZW5XEE23LILEQUNTBC3LYIY3ZJ/events.json","paper":"https://pith.science/paper/ZW5XEE23"},"agent_actions":{"view_html":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ","download_json":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ.json","view_paper":"https://pith.science/paper/ZW5XEE23","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.10777&json=true","fetch_graph":"https://pith.science/api/pith-number/ZW5XEE23LILEQUNTBC3LYIY3ZJ/graph.json","fetch_events":"https://pith.science/api/pith-number/ZW5XEE23LILEQUNTBC3LYIY3ZJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ/action/storage_attestation","attest_author":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ/action/author_attestation","sign_citation":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ/action/citation_signature","submit_replication":"https://pith.science/pith/ZW5XEE23LILEQUNTBC3LYIY3ZJ/action/replication_record"}},"created_at":"2026-05-20T00:02:57.195445+00:00","updated_at":"2026-05-20T00:02:57.195445+00:00"}