{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MTIWQXOCKCSIGVUWNC2JJW4L2J","short_pith_number":"pith:MTIWQXOC","schema_version":"1.0","canonical_sha256":"64d1685dc250a483569668b494db8bd2444c258ce092271ea1bf7942d1601705","source":{"kind":"arxiv","id":"2405.16528","version":4},"attestation_state":"computed","paper":{"title":"LoQT: Low-Rank Adapters for Quantized Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Mads Toftrup, Michael J. Kastoryano, Sebastian Loeschcke, Serge Belongie, V\\'esteinn Sn{\\ae}bjarnarson","submitted_at":"2024-05-26T11:29:57Z","abstract_excerpt":"Despite advances using low-rank adapters and quantization, pretraining of large models on consumer hardware has not been possible without model sharding, offloading during training, or per-layer gradient updates. To address these limitations, we propose Low-Rank Adapters for Quantized Training (LoQT), a method for efficiently training quantized models. LoQT uses gradient-based tensor factorization to initialize low-rank trainable weight matrices that are periodically merged into quantized full-rank weight matrices. Our approach is suitable for both pretraining and fine-tuning models. We demons"},"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":"2405.16528","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-26T11:29:57Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"689c38c7da7c848675e8200859c86d657f70a9ef1226492f99a41cdc6ee1d825","abstract_canon_sha256":"5cc4215dc194cde048c173a29812684a1efff03ab4e67a87a569008e81bcc50b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:30:15.008775Z","signature_b64":"1vyUZxSBUWTV8ho0QPJ/IawUU18WBXReqybaTfuDPOkJ+k4dYwJNSZUjGoTBwmID68qrvbbRcATiJU1SXzdvBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"64d1685dc250a483569668b494db8bd2444c258ce092271ea1bf7942d1601705","last_reissued_at":"2026-07-05T09:30:15.008263Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:30:15.008263Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoQT: Low-Rank Adapters for Quantized Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Mads Toftrup, Michael J. Kastoryano, Sebastian Loeschcke, Serge Belongie, V\\'esteinn Sn{\\ae}bjarnarson","submitted_at":"2024-05-26T11:29:57Z","abstract_excerpt":"Despite advances using low-rank adapters and quantization, pretraining of large models on consumer hardware has not been possible without model sharding, offloading during training, or per-layer gradient updates. To address these limitations, we propose Low-Rank Adapters for Quantized Training (LoQT), a method for efficiently training quantized models. LoQT uses gradient-based tensor factorization to initialize low-rank trainable weight matrices that are periodically merged into quantized full-rank weight matrices. Our approach is suitable for both pretraining and fine-tuning models. We demons"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.16528","kind":"arxiv","version":4},"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/2405.16528/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":"2405.16528","created_at":"2026-07-05T09:30:15.008321+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.16528v4","created_at":"2026-07-05T09:30:15.008321+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.16528","created_at":"2026-07-05T09:30:15.008321+00:00"},{"alias_kind":"pith_short_12","alias_value":"MTIWQXOCKCSI","created_at":"2026-07-05T09:30:15.008321+00:00"},{"alias_kind":"pith_short_16","alias_value":"MTIWQXOCKCSIGVUW","created_at":"2026-07-05T09:30:15.008321+00:00"},{"alias_kind":"pith_short_8","alias_value":"MTIWQXOC","created_at":"2026-07-05T09:30:15.008321+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.13652","citing_title":"Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13652","citing_title":"Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J","json":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J.json","graph_json":"https://pith.science/api/pith-number/MTIWQXOCKCSIGVUWNC2JJW4L2J/graph.json","events_json":"https://pith.science/api/pith-number/MTIWQXOCKCSIGVUWNC2JJW4L2J/events.json","paper":"https://pith.science/paper/MTIWQXOC"},"agent_actions":{"view_html":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J","download_json":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J.json","view_paper":"https://pith.science/paper/MTIWQXOC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.16528&json=true","fetch_graph":"https://pith.science/api/pith-number/MTIWQXOCKCSIGVUWNC2JJW4L2J/graph.json","fetch_events":"https://pith.science/api/pith-number/MTIWQXOCKCSIGVUWNC2JJW4L2J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J/action/storage_attestation","attest_author":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J/action/author_attestation","sign_citation":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J/action/citation_signature","submit_replication":"https://pith.science/pith/MTIWQXOCKCSIGVUWNC2JJW4L2J/action/replication_record"}},"created_at":"2026-07-05T09:30:15.008321+00:00","updated_at":"2026-07-05T09:30:15.008321+00:00"}