{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TL6IPOFXSO3M6CV2UWQFZTR3QQ","short_pith_number":"pith:TL6IPOFX","schema_version":"1.0","canonical_sha256":"9afc87b8b793b6cf0abaa5a05cce3b840f86d730f79c3d8a9bbcea1135c7aca1","source":{"kind":"arxiv","id":"2402.11592","version":3},"attestation_state":"computed","paper":{"title":"Revisiting Zeroth-Order Optimization for Memory-Efficient LLM Fine-Tuning: A Benchmark","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jason D. Lee, Jiaxiang Li, Junyuan Hong, Mingyi Hong, Pingzhi Li, Pin-Yu Chen, Sijia Liu, Tianlong Chen, Wenqing Zheng, Wotao Yin, Yihua Zhang, Yimeng Zhang, Zhangyang Wang","submitted_at":"2024-02-18T14:08:48Z","abstract_excerpt":"In the evolving landscape of natural language processing (NLP), fine-tuning pre-trained Large Language Models (LLMs) with first-order (FO) optimizers like SGD and Adam has become standard. Yet, as LLMs grow {in size}, the substantial memory overhead from back-propagation (BP) for FO gradient computation presents a significant challenge. Addressing this issue is crucial, especially for applications like on-device training where memory efficiency is paramount. This paper proposes a shift towards BP-free, zeroth-order (ZO) optimization as a solution for reducing memory costs during LLM fine-tunin"},"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":"2402.11592","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-18T14:08:48Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"b9721e42e079bf1e682ce83974889b1b015edba25999d9b23a682dd2343800b1","abstract_canon_sha256":"20cd7237438600ad488b4bb07c4c15ef25db14c99f9802dc202bf0c5e5b31504"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:24:02.149405Z","signature_b64":"iwp0ONtf4g3P6suaW5fYzMgiz+yGP96YukjGBEGI9GFrIZf4NyK9KpQgwpD0rJo8ELRFh7z+VODHq3DwDtuFBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9afc87b8b793b6cf0abaa5a05cce3b840f86d730f79c3d8a9bbcea1135c7aca1","last_reissued_at":"2026-07-05T08:24:02.148919Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:24:02.148919Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Revisiting Zeroth-Order Optimization for Memory-Efficient LLM Fine-Tuning: A Benchmark","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jason D. Lee, Jiaxiang Li, Junyuan Hong, Mingyi Hong, Pingzhi Li, Pin-Yu Chen, Sijia Liu, Tianlong Chen, Wenqing Zheng, Wotao Yin, Yihua Zhang, Yimeng Zhang, Zhangyang Wang","submitted_at":"2024-02-18T14:08:48Z","abstract_excerpt":"In the evolving landscape of natural language processing (NLP), fine-tuning pre-trained Large Language Models (LLMs) with first-order (FO) optimizers like SGD and Adam has become standard. Yet, as LLMs grow {in size}, the substantial memory overhead from back-propagation (BP) for FO gradient computation presents a significant challenge. Addressing this issue is crucial, especially for applications like on-device training where memory efficiency is paramount. This paper proposes a shift towards BP-free, zeroth-order (ZO) optimization as a solution for reducing memory costs during LLM fine-tunin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.11592","kind":"arxiv","version":3},"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/2402.11592/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":"2402.11592","created_at":"2026-07-05T08:24:02.148977+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.11592v3","created_at":"2026-07-05T08:24:02.148977+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.11592","created_at":"2026-07-05T08:24:02.148977+00:00"},{"alias_kind":"pith_short_12","alias_value":"TL6IPOFXSO3M","created_at":"2026-07-05T08:24:02.148977+00:00"},{"alias_kind":"pith_short_16","alias_value":"TL6IPOFXSO3M6CV2","created_at":"2026-07-05T08:24:02.148977+00:00"},{"alias_kind":"pith_short_8","alias_value":"TL6IPOFX","created_at":"2026-07-05T08:24:02.148977+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.14970","citing_title":"Zero-order Parameter-free Optimization for LMO-based Methods: Novel Approach for Efficient Fine-tuning","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12908","citing_title":"SENTINEL: Failure-Driven Reinforcement Learning for Training Tool-Using Language Model Agents","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12124","citing_title":"A Unified Zeroth-Order Approach for Decentralized Minimax Optimization","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02869","citing_title":"ZOAF: Towards Efficient Zeroth-Order Optimization for Analog/RF Circuit Design","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31208","citing_title":"Probing Memorization of Tabular In-Context Learning","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29396","citing_title":"Aligned but Fragile: Enhancing LLM Safety Robustness via Zeroth-Order Optimization","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09034","citing_title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2510.00761","citing_title":"Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12813","citing_title":"REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations","ref_index":165,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09034","citing_title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00650","citing_title":"AdaMeZO: Adam-style Zeroth-Order Optimizer for LLM Fine-tuning Without Maintaining the Moments","ref_index":67,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ","json":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ.json","graph_json":"https://pith.science/api/pith-number/TL6IPOFXSO3M6CV2UWQFZTR3QQ/graph.json","events_json":"https://pith.science/api/pith-number/TL6IPOFXSO3M6CV2UWQFZTR3QQ/events.json","paper":"https://pith.science/paper/TL6IPOFX"},"agent_actions":{"view_html":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ","download_json":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ.json","view_paper":"https://pith.science/paper/TL6IPOFX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.11592&json=true","fetch_graph":"https://pith.science/api/pith-number/TL6IPOFXSO3M6CV2UWQFZTR3QQ/graph.json","fetch_events":"https://pith.science/api/pith-number/TL6IPOFXSO3M6CV2UWQFZTR3QQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ/action/storage_attestation","attest_author":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ/action/author_attestation","sign_citation":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ/action/citation_signature","submit_replication":"https://pith.science/pith/TL6IPOFXSO3M6CV2UWQFZTR3QQ/action/replication_record"}},"created_at":"2026-07-05T08:24:02.148977+00:00","updated_at":"2026-07-05T08:24:02.148977+00:00"}