{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3CU2INF2GMHF2UDJ4YN7DTU5OV","short_pith_number":"pith:3CU2INF2","schema_version":"1.0","canonical_sha256":"d8a9a434ba330e5d5069e61bf1ce9d7552d2358ee48ac3c931bf58963f4b6b6e","source":{"kind":"arxiv","id":"2412.19820","version":1},"attestation_state":"computed","paper":{"title":"GaLore$+$: Boosting Low-Rank Adaptation for LLMs with Cross-Head Projection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Shaohui Li, Xutao Liao, You He, Yuhui Xu, Yu Liu, Zhi Li","submitted_at":"2024-12-15T12:28:13Z","abstract_excerpt":"Recent low-rank training methods, such as GaLore, have significantly reduced the memory required to optimize large language models (LLMs). However, these methods often suffer from time-consuming low-rank projection estimations. In particular, the singular value decomposition (SVD) in GaLore can consume more than 80\\% of the total training time. To address this issue, we propose GaLore$+$, which uses cross-head low-rank projection to reduce the substantial time consumption in estimating low-rank projections for multi-head attention. In addition, we employ randomized subspace iteration to achiev"},"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":"2412.19820","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-12-15T12:28:13Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"f1d787a196fdf8a90fec6fde95b4dd1696faed4751b220e67fb57010e47ba360","abstract_canon_sha256":"cd4ffb308bae9d460564a324bca55bd08475f4e07ddf64d9077202ccaf615f2c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:55:16.659766Z","signature_b64":"zy+BUfmG/T2W2cg0W72pRtT1CvUQ5P+EBUwn3n2nV2jIkqg97Lll9b+AeLZC2p/Judve8/EM/0OKNNSqAJfRBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d8a9a434ba330e5d5069e61bf1ce9d7552d2358ee48ac3c931bf58963f4b6b6e","last_reissued_at":"2026-07-05T09:55:16.659279Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:55:16.659279Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GaLore$+$: Boosting Low-Rank Adaptation for LLMs with Cross-Head Projection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Shaohui Li, Xutao Liao, You He, Yuhui Xu, Yu Liu, Zhi Li","submitted_at":"2024-12-15T12:28:13Z","abstract_excerpt":"Recent low-rank training methods, such as GaLore, have significantly reduced the memory required to optimize large language models (LLMs). However, these methods often suffer from time-consuming low-rank projection estimations. In particular, the singular value decomposition (SVD) in GaLore can consume more than 80\\% of the total training time. To address this issue, we propose GaLore$+$, which uses cross-head low-rank projection to reduce the substantial time consumption in estimating low-rank projections for multi-head attention. In addition, we employ randomized subspace iteration to achiev"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.19820","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/2412.19820/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":"2412.19820","created_at":"2026-07-05T09:55:16.659343+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.19820v1","created_at":"2026-07-05T09:55:16.659343+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.19820","created_at":"2026-07-05T09:55:16.659343+00:00"},{"alias_kind":"pith_short_12","alias_value":"3CU2INF2GMHF","created_at":"2026-07-05T09:55:16.659343+00:00"},{"alias_kind":"pith_short_16","alias_value":"3CU2INF2GMHF2UDJ","created_at":"2026-07-05T09:55:16.659343+00:00"},{"alias_kind":"pith_short_8","alias_value":"3CU2INF2","created_at":"2026-07-05T09:55:16.659343+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2506.16659","citing_title":"Memory-Efficient LLM Pretraining via Minimalist Optimizer Design","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV","json":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV.json","graph_json":"https://pith.science/api/pith-number/3CU2INF2GMHF2UDJ4YN7DTU5OV/graph.json","events_json":"https://pith.science/api/pith-number/3CU2INF2GMHF2UDJ4YN7DTU5OV/events.json","paper":"https://pith.science/paper/3CU2INF2"},"agent_actions":{"view_html":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV","download_json":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV.json","view_paper":"https://pith.science/paper/3CU2INF2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.19820&json=true","fetch_graph":"https://pith.science/api/pith-number/3CU2INF2GMHF2UDJ4YN7DTU5OV/graph.json","fetch_events":"https://pith.science/api/pith-number/3CU2INF2GMHF2UDJ4YN7DTU5OV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV/action/storage_attestation","attest_author":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV/action/author_attestation","sign_citation":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV/action/citation_signature","submit_replication":"https://pith.science/pith/3CU2INF2GMHF2UDJ4YN7DTU5OV/action/replication_record"}},"created_at":"2026-07-05T09:55:16.659343+00:00","updated_at":"2026-07-05T09:55:16.659343+00:00"}