{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:LGEPWDFEKBCQJKLGGYUI3CGTWM","short_pith_number":"pith:LGEPWDFE","schema_version":"1.0","canonical_sha256":"5988fb0ca4504504a96636288d88d3b3146048919c8c63a38213fa3ff88a2558","source":{"kind":"arxiv","id":"2209.13569","version":1},"attestation_state":"computed","paper":{"title":"Exploring Low Rank Training of Deep Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Acyr Locatelli, Aidan N. Gomez, Bharat Venkitesh, Jimmy Ba, Siddhartha Rao Kamalakara, Yarin Gal","submitted_at":"2022-09-27T17:43:45Z","abstract_excerpt":"Training deep neural networks in low rank, i.e. with factorised layers, is of particular interest to the community: it offers efficiency over unfactorised training in terms of both memory consumption and training time. Prior work has focused on low rank approximations of pre-trained networks and training in low rank space with additional objectives, offering various ad hoc explanations for chosen practice. We analyse techniques that work well in practice, and through extensive ablations on models such as GPT2 we provide evidence falsifying common beliefs in the field, hinting in the process at"},"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":"2209.13569","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-09-27T17:43:45Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"be1ac159102d12f592b7d07981dd1a3e3e9801d7ec267d85878beb84a847220d","abstract_canon_sha256":"64ec6d7acb9d436afe9a7d229ba2c8123c4abf137724976058606aebcb4e66bf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:01:33.862926Z","signature_b64":"6gcaHEZxqWtcFup2jw4g/2iNFEFK+wrYtag7WFbSaf4kTF14556iupE4w9Md6q2vFKYRJx6pE/vgnWdwoWxTCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5988fb0ca4504504a96636288d88d3b3146048919c8c63a38213fa3ff88a2558","last_reissued_at":"2026-07-05T05:01:33.862435Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:01:33.862435Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Exploring Low Rank Training of Deep Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Acyr Locatelli, Aidan N. Gomez, Bharat Venkitesh, Jimmy Ba, Siddhartha Rao Kamalakara, Yarin Gal","submitted_at":"2022-09-27T17:43:45Z","abstract_excerpt":"Training deep neural networks in low rank, i.e. with factorised layers, is of particular interest to the community: it offers efficiency over unfactorised training in terms of both memory consumption and training time. Prior work has focused on low rank approximations of pre-trained networks and training in low rank space with additional objectives, offering various ad hoc explanations for chosen practice. We analyse techniques that work well in practice, and through extensive ablations on models such as GPT2 we provide evidence falsifying common beliefs in the field, hinting in the process at"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2209.13569","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/2209.13569/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":"2209.13569","created_at":"2026-07-05T05:01:33.862492+00:00"},{"alias_kind":"arxiv_version","alias_value":"2209.13569v1","created_at":"2026-07-05T05:01:33.862492+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2209.13569","created_at":"2026-07-05T05:01:33.862492+00:00"},{"alias_kind":"pith_short_12","alias_value":"LGEPWDFEKBCQ","created_at":"2026-07-05T05:01:33.862492+00:00"},{"alias_kind":"pith_short_16","alias_value":"LGEPWDFEKBCQJKLG","created_at":"2026-07-05T05:01:33.862492+00:00"},{"alias_kind":"pith_short_8","alias_value":"LGEPWDFE","created_at":"2026-07-05T05:01:33.862492+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.28932","citing_title":"DLR: Zero-Inference-Cost Latent Residuals for Low-Rank Pre-Training","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18993","citing_title":"CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2403.03507","citing_title":"GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2512.12131","citing_title":"BOOST: BOttleneck-Optimized Scalable Training Framework for Low-Rank Large Language Models","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07063","citing_title":"Dr. Post-Training: A Data Regularization Perspective on LLM Post-Training","ref_index":55,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM","json":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM.json","graph_json":"https://pith.science/api/pith-number/LGEPWDFEKBCQJKLGGYUI3CGTWM/graph.json","events_json":"https://pith.science/api/pith-number/LGEPWDFEKBCQJKLGGYUI3CGTWM/events.json","paper":"https://pith.science/paper/LGEPWDFE"},"agent_actions":{"view_html":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM","download_json":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM.json","view_paper":"https://pith.science/paper/LGEPWDFE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2209.13569&json=true","fetch_graph":"https://pith.science/api/pith-number/LGEPWDFEKBCQJKLGGYUI3CGTWM/graph.json","fetch_events":"https://pith.science/api/pith-number/LGEPWDFEKBCQJKLGGYUI3CGTWM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM/action/storage_attestation","attest_author":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM/action/author_attestation","sign_citation":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM/action/citation_signature","submit_replication":"https://pith.science/pith/LGEPWDFEKBCQJKLGGYUI3CGTWM/action/replication_record"}},"created_at":"2026-07-05T05:01:33.862492+00:00","updated_at":"2026-07-05T05:01:33.862492+00:00"}