{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:VPTAFFPPFKB345U32LG54TVE57","short_pith_number":"pith:VPTAFFPP","schema_version":"1.0","canonical_sha256":"abe60295ef2a83be769bd2cdde4ea4efd094b51944c35210db5009830457a1e9","source":{"kind":"arxiv","id":"2412.05270","version":4},"attestation_state":"computed","paper":{"title":"APOLLO: SGD-like Memory, AdamW-level Performance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.PF"],"primary_cat":"cs.LG","authors_text":"Bo Long, David Z. Pan, Hanqing Zhu, Jinwon Lee, Sem Park, Vikas Chandra, Wenyan Cong, Xi Liu, Zhangyang Wang, Zhenyu Zhang","submitted_at":"2024-12-06T18:55:34Z","abstract_excerpt":"Large language models (LLMs) are notoriously memory-intensive during training, particularly with the popular AdamW optimizer. This memory burden necessitates using more or higher-end GPUs or reducing batch sizes, limiting training scalability and throughput. To address this, various memory-efficient optimizers have been proposed to reduce optimizer memory usage. However, they face critical challenges: (i) reliance on costly SVD operations; (ii) significant performance trade-offs compared to AdamW; and (iii) still substantial optimizer memory overhead to maintain competitive performance.\n  In t"},"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":"2412.05270","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-12-06T18:55:34Z","cross_cats_sorted":["cs.AI","cs.PF"],"title_canon_sha256":"e74e6fd5f53e4df392d76d2df350ba42cbd49ba40d1688b5412a49332cfe0763","abstract_canon_sha256":"4e11076902e916fd0df577c4e9dc004ba96468b01ffde9100362b13bbd9b16fa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:23.634688Z","signature_b64":"wBnvqKg4zT+UegTjg49DUsSQIVaeVVB/sQn+MtLYWirGGIVgZJ2VGfUtWMDlJvQqthe1fsC7e2XauazftAvBBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"abe60295ef2a83be769bd2cdde4ea4efd094b51944c35210db5009830457a1e9","last_reissued_at":"2026-07-05T10:15:23.634208Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:23.634208Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"APOLLO: SGD-like Memory, AdamW-level Performance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.PF"],"primary_cat":"cs.LG","authors_text":"Bo Long, David Z. Pan, Hanqing Zhu, Jinwon Lee, Sem Park, Vikas Chandra, Wenyan Cong, Xi Liu, Zhangyang Wang, Zhenyu Zhang","submitted_at":"2024-12-06T18:55:34Z","abstract_excerpt":"Large language models (LLMs) are notoriously memory-intensive during training, particularly with the popular AdamW optimizer. This memory burden necessitates using more or higher-end GPUs or reducing batch sizes, limiting training scalability and throughput. To address this, various memory-efficient optimizers have been proposed to reduce optimizer memory usage. However, they face critical challenges: (i) reliance on costly SVD operations; (ii) significant performance trade-offs compared to AdamW; and (iii) still substantial optimizer memory overhead to maintain competitive performance.\n  In t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.05270","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/2412.05270/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":"2412.05270","created_at":"2026-07-05T10:15:23.634258+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.05270v4","created_at":"2026-07-05T10:15:23.634258+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.05270","created_at":"2026-07-05T10:15:23.634258+00:00"},{"alias_kind":"pith_short_12","alias_value":"VPTAFFPPFKB3","created_at":"2026-07-05T10:15:23.634258+00:00"},{"alias_kind":"pith_short_16","alias_value":"VPTAFFPPFKB345U3","created_at":"2026-07-05T10:15:23.634258+00:00"},{"alias_kind":"pith_short_8","alias_value":"VPTAFFPP","created_at":"2026-07-05T10:15:23.634258+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05872","citing_title":"No Subspace to Track: Non-Identifiability and Optimizer State in Low-Rank Training","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2509.18993","citing_title":"CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure","ref_index":66,"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":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09176","citing_title":"Navigating LLM Valley: From AdamW to Memory-Efficient and Matrix-Based Optimizers","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11867","citing_title":"Disposition Distillation at Small Scale: A Three-Arc Negative Result","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07663","citing_title":"SAGE: Sign-Adaptive Gradient for Memory-Efficient LLM Optimization","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57","json":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57.json","graph_json":"https://pith.science/api/pith-number/VPTAFFPPFKB345U32LG54TVE57/graph.json","events_json":"https://pith.science/api/pith-number/VPTAFFPPFKB345U32LG54TVE57/events.json","paper":"https://pith.science/paper/VPTAFFPP"},"agent_actions":{"view_html":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57","download_json":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57.json","view_paper":"https://pith.science/paper/VPTAFFPP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.05270&json=true","fetch_graph":"https://pith.science/api/pith-number/VPTAFFPPFKB345U32LG54TVE57/graph.json","fetch_events":"https://pith.science/api/pith-number/VPTAFFPPFKB345U32LG54TVE57/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57/action/storage_attestation","attest_author":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57/action/author_attestation","sign_citation":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57/action/citation_signature","submit_replication":"https://pith.science/pith/VPTAFFPPFKB345U32LG54TVE57/action/replication_record"}},"created_at":"2026-07-05T10:15:23.634258+00:00","updated_at":"2026-07-05T10:15:23.634258+00:00"}