{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PEOM375UR3HZXWNB2GTD7ZYLJV","short_pith_number":"pith:PEOM375U","schema_version":"1.0","canonical_sha256":"791ccdffb48ecf9bd9a1d1a63fe70b4d76d02914d364084284e7332d24ffd516","source":{"kind":"arxiv","id":"2606.02328","version":1},"attestation_state":"computed","paper":{"title":"Riemannian Gradient Descent for Low-Rank Architectures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Nicholas Knight","submitted_at":"2026-06-01T14:40:00Z","abstract_excerpt":"We explore Riemannian optimization techniques for rank-factored matrix parameters, targeting contemporary deep learning applications. We examine ten points in the algorithm design space: two geometries for rank-$r$ matrices, three geometries for rank-$r$ partial isometries, and block-matrix variants of these five, where factors are shared across block-rows and block-columns. We apply our methods to the multihead attention parameters in small language models. After tuning learning rates, our methods do not conclusively outperform an AdamW baseline. Our implementations are available online."},"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":"2606.02328","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-01T14:40:00Z","cross_cats_sorted":[],"title_canon_sha256":"77e813f30704302f40e8bdaee595bd483d32dea6d9fb7ca5fbd4107361284881","abstract_canon_sha256":"33908af3abc56afbc8eabb9107aebdb780ef02414c5bf03b474344ad489a9736"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T03:04:56.251102Z","signature_b64":"651JCmDPG2pp3h89J0hdJf7l7xHqwcL7KQdt4we9fGIJJUuyY8ImkgkUZkGnobaA0kIh7+RkbN5SrBS0BpeFBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"791ccdffb48ecf9bd9a1d1a63fe70b4d76d02914d364084284e7332d24ffd516","last_reissued_at":"2026-06-02T03:04:56.250774Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T03:04:56.250774Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Riemannian Gradient Descent for Low-Rank Architectures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Nicholas Knight","submitted_at":"2026-06-01T14:40:00Z","abstract_excerpt":"We explore Riemannian optimization techniques for rank-factored matrix parameters, targeting contemporary deep learning applications. We examine ten points in the algorithm design space: two geometries for rank-$r$ matrices, three geometries for rank-$r$ partial isometries, and block-matrix variants of these five, where factors are shared across block-rows and block-columns. We apply our methods to the multihead attention parameters in small language models. After tuning learning rates, our methods do not conclusively outperform an AdamW baseline. Our implementations are available online."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.02328","kind":"arxiv","version":1},"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/2606.02328/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":"2606.02328","created_at":"2026-06-02T03:04:56.250828+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.02328v1","created_at":"2026-06-02T03:04:56.250828+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.02328","created_at":"2026-06-02T03:04:56.250828+00:00"},{"alias_kind":"pith_short_12","alias_value":"PEOM375UR3HZ","created_at":"2026-06-02T03:04:56.250828+00:00"},{"alias_kind":"pith_short_16","alias_value":"PEOM375UR3HZXWNB","created_at":"2026-06-02T03:04:56.250828+00:00"},{"alias_kind":"pith_short_8","alias_value":"PEOM375U","created_at":"2026-06-02T03:04:56.250828+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/PEOM375UR3HZXWNB2GTD7ZYLJV","json":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV.json","graph_json":"https://pith.science/api/pith-number/PEOM375UR3HZXWNB2GTD7ZYLJV/graph.json","events_json":"https://pith.science/api/pith-number/PEOM375UR3HZXWNB2GTD7ZYLJV/events.json","paper":"https://pith.science/paper/PEOM375U"},"agent_actions":{"view_html":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV","download_json":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV.json","view_paper":"https://pith.science/paper/PEOM375U","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.02328&json=true","fetch_graph":"https://pith.science/api/pith-number/PEOM375UR3HZXWNB2GTD7ZYLJV/graph.json","fetch_events":"https://pith.science/api/pith-number/PEOM375UR3HZXWNB2GTD7ZYLJV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV/action/storage_attestation","attest_author":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV/action/author_attestation","sign_citation":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV/action/citation_signature","submit_replication":"https://pith.science/pith/PEOM375UR3HZXWNB2GTD7ZYLJV/action/replication_record"}},"created_at":"2026-06-02T03:04:56.250828+00:00","updated_at":"2026-06-02T03:04:56.250828+00:00"}