{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZHU2NH44ZN5PRP54WP7JVKC2M5","short_pith_number":"pith:ZHU2NH44","schema_version":"1.0","canonical_sha256":"c9e9a69f9ccb7af8bfbcb3fe9aa85a677998aec371beb1cb7e4ef9e3ef560dd5","source":{"kind":"arxiv","id":"2602.12429","version":2},"attestation_state":"computed","paper":{"title":"Stabilizing Native Low-Rank LLM Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Edouard Oyallon, Eugene Belilovsky, Paul Janson","submitted_at":"2026-02-12T21:33:14Z","abstract_excerpt":"Foundation models have achieved remarkable success, yet their growing parameter counts pose significant computational and memory challenges. Low-rank factorization offers a promising route to reduce training and inference costs, but the community lacks a stable recipe for training models from scratch using exclusively low-rank weights while matching the performance of the dense model. We demonstrate that Large Language Models (LLMs) can be trained from scratch using exclusively low-rank factorized weights for all non-embedding matrices without auxiliary \"full-rank\" guidance required by prior m"},"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":"2602.12429","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-12T21:33:14Z","cross_cats_sorted":[],"title_canon_sha256":"a6ba30bbe36bc68fb7287ec2fd9fb9eab1e12bddce0d74721f6b45b3a0818b9e","abstract_canon_sha256":"7cb78fe443fe5ac0b3d156f373cae268d449cfea00cd486fa4460a33c0a7ec44"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T00:20:44.742744Z","signature_b64":"cbyLU/elFz2JtASGx+UNDF1Q+Sno9fpzoKGvFrDSuPtAJO3mYlJl3P13xCW5xl6J9EKcX31O5pFbZIyrv/lSBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c9e9a69f9ccb7af8bfbcb3fe9aa85a677998aec371beb1cb7e4ef9e3ef560dd5","last_reissued_at":"2026-07-17T00:20:44.741807Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T00:20:44.741807Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stabilizing Native Low-Rank LLM Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Edouard Oyallon, Eugene Belilovsky, Paul Janson","submitted_at":"2026-02-12T21:33:14Z","abstract_excerpt":"Foundation models have achieved remarkable success, yet their growing parameter counts pose significant computational and memory challenges. Low-rank factorization offers a promising route to reduce training and inference costs, but the community lacks a stable recipe for training models from scratch using exclusively low-rank weights while matching the performance of the dense model. We demonstrate that Large Language Models (LLMs) can be trained from scratch using exclusively low-rank factorized weights for all non-embedding matrices without auxiliary \"full-rank\" guidance required by prior m"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.12429","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/2602.12429/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":"2602.12429","created_at":"2026-07-17T00:20:44.742249+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.12429v2","created_at":"2026-07-17T00:20:44.742249+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.12429","created_at":"2026-07-17T00:20:44.742249+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZHU2NH44ZN5P","created_at":"2026-07-17T00:20:44.742249+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZHU2NH44ZN5PRP54","created_at":"2026-07-17T00:20:44.742249+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZHU2NH44","created_at":"2026-07-17T00:20:44.742249+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.12921","citing_title":"LoRA-Muon: Spectral Steepest Descent on the Low-Rank Manifold","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09238","citing_title":"Intrinsic Muon: Spectral Optimization on Riemannian Matrix Manifolds","ref_index":26,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5","json":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5.json","graph_json":"https://pith.science/api/pith-number/ZHU2NH44ZN5PRP54WP7JVKC2M5/graph.json","events_json":"https://pith.science/api/pith-number/ZHU2NH44ZN5PRP54WP7JVKC2M5/events.json","paper":"https://pith.science/paper/ZHU2NH44"},"agent_actions":{"view_html":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5","download_json":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5.json","view_paper":"https://pith.science/paper/ZHU2NH44","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.12429&json=true","fetch_graph":"https://pith.science/api/pith-number/ZHU2NH44ZN5PRP54WP7JVKC2M5/graph.json","fetch_events":"https://pith.science/api/pith-number/ZHU2NH44ZN5PRP54WP7JVKC2M5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5/action/storage_attestation","attest_author":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5/action/author_attestation","sign_citation":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5/action/citation_signature","submit_replication":"https://pith.science/pith/ZHU2NH44ZN5PRP54WP7JVKC2M5/action/replication_record"}},"created_at":"2026-07-17T00:20:44.742249+00:00","updated_at":"2026-07-17T00:20:44.742249+00:00"}