{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:U6EERTYPBFZB3FSJLY4SFXOH7Z","short_pith_number":"pith:U6EERTYP","schema_version":"1.0","canonical_sha256":"a78848cf0f09721d96495e3922ddc7fe532985cbb4eb58f4248c5026b7935cc4","source":{"kind":"arxiv","id":"2410.09823","version":1},"attestation_state":"computed","paper":{"title":"Simultaneous Computation and Memory Efficient Zeroth-Order Optimizer for Fine-Tuning Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Changxing Ding, Chao Xue, Fei Wang, Liang Ding, Li Shen, Ye Liu","submitted_at":"2024-10-13T12:47:37Z","abstract_excerpt":"Fine-tuning is powerful for adapting large language models to downstream tasks, but it often results in huge memory usages. A promising approach to mitigate this is using Zeroth-Order (ZO) optimization, which estimates gradients to replace First-Order (FO) gradient calculations, albeit with longer training time due to its stochastic nature. By revisiting the Memory-efficient ZO (MeZO) optimizer, we discover that the full-parameter perturbation and updating processes consume over 50% of its overall fine-tuning time cost. Based on these observations, we introduce a novel layer-wise sparse comput"},"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":"2410.09823","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-13T12:47:37Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"516b69c00e2c05ad79c972ddcf418df725800a294db47cb1efdb04cfc8e52874","abstract_canon_sha256":"f85482f07f1319a8421df1e1a40a294245f9e3f20a2b00cdc604f8273238604b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:20:06.663365Z","signature_b64":"WTrrVRpk26QetU5FW+wlVe98n/NI/H/CHd1ZjRF9kn6fPWXLiZtTMUJIG+PIGZoRHrjhbjX9yRmcw6zHrzNGCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a78848cf0f09721d96495e3922ddc7fe532985cbb4eb58f4248c5026b7935cc4","last_reissued_at":"2026-07-05T09:20:06.662910Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:20:06.662910Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Simultaneous Computation and Memory Efficient Zeroth-Order Optimizer for Fine-Tuning Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Changxing Ding, Chao Xue, Fei Wang, Liang Ding, Li Shen, Ye Liu","submitted_at":"2024-10-13T12:47:37Z","abstract_excerpt":"Fine-tuning is powerful for adapting large language models to downstream tasks, but it often results in huge memory usages. A promising approach to mitigate this is using Zeroth-Order (ZO) optimization, which estimates gradients to replace First-Order (FO) gradient calculations, albeit with longer training time due to its stochastic nature. By revisiting the Memory-efficient ZO (MeZO) optimizer, we discover that the full-parameter perturbation and updating processes consume over 50% of its overall fine-tuning time cost. Based on these observations, we introduce a novel layer-wise sparse comput"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.09823","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/2410.09823/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":"2410.09823","created_at":"2026-07-05T09:20:06.662965+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.09823v1","created_at":"2026-07-05T09:20:06.662965+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.09823","created_at":"2026-07-05T09:20:06.662965+00:00"},{"alias_kind":"pith_short_12","alias_value":"U6EERTYPBFZB","created_at":"2026-07-05T09:20:06.662965+00:00"},{"alias_kind":"pith_short_16","alias_value":"U6EERTYPBFZB3FSJ","created_at":"2026-07-05T09:20:06.662965+00:00"},{"alias_kind":"pith_short_8","alias_value":"U6EERTYP","created_at":"2026-07-05T09:20:06.662965+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.18264","citing_title":"Universally Empowering Zeroth-Order Optimization via Adaptive Layer-wise Sampling","ref_index":41,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z","json":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z.json","graph_json":"https://pith.science/api/pith-number/U6EERTYPBFZB3FSJLY4SFXOH7Z/graph.json","events_json":"https://pith.science/api/pith-number/U6EERTYPBFZB3FSJLY4SFXOH7Z/events.json","paper":"https://pith.science/paper/U6EERTYP"},"agent_actions":{"view_html":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z","download_json":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z.json","view_paper":"https://pith.science/paper/U6EERTYP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.09823&json=true","fetch_graph":"https://pith.science/api/pith-number/U6EERTYPBFZB3FSJLY4SFXOH7Z/graph.json","fetch_events":"https://pith.science/api/pith-number/U6EERTYPBFZB3FSJLY4SFXOH7Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z/action/storage_attestation","attest_author":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z/action/author_attestation","sign_citation":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z/action/citation_signature","submit_replication":"https://pith.science/pith/U6EERTYPBFZB3FSJLY4SFXOH7Z/action/replication_record"}},"created_at":"2026-07-05T09:20:06.662965+00:00","updated_at":"2026-07-05T09:20:06.662965+00:00"}