{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NKVAFVXZTIMRM7CABDQWN2SRPI","short_pith_number":"pith:NKVAFVXZ","schema_version":"1.0","canonical_sha256":"6aaa02d6f99a19167c4008e166ea517a2bd01aeb72a077663e6bc17ab533525f","source":{"kind":"arxiv","id":"2402.15173","version":4},"attestation_state":"computed","paper":{"title":"Second-Order Fine-Tuning without Pain for LLMs:A Hessian Informed Zeroth-Order Optimizer","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guang Dai, Haishan Ye, Ivor W.Tsang, Sizhe Dang, Yanjun Zhao, Yi Qian","submitted_at":"2024-02-23T08:11:55Z","abstract_excerpt":"Fine-tuning large language models (LLMs) with classic first-order optimizers entails prohibitive GPU memory due to the backpropagation process. Recent works have turned to zeroth-order optimizers for fine-tuning, which save substantial memory by using two forward passes. However, these optimizers are plagued by the heterogeneity of parameter curvatures across different dimensions. In this work, we propose HiZOO, a diagonal Hessian informed zeroth-order optimizer which is the first work to leverage the diagonal Hessian to enhance zeroth-order optimizer for fine-tuning LLMs. What's more, HiZOO a"},"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":"2402.15173","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-23T08:11:55Z","cross_cats_sorted":[],"title_canon_sha256":"a75bbb6f5f8c4aa300b27bdd8334bc90c15af8783a0f4024e320e5338573f17c","abstract_canon_sha256":"c0c7310ae3fc71f16baeb8e07ff69bc1f3a4326d400a8eec84b53bba1667ee82"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:16:07.514377Z","signature_b64":"5vRJHTB5Q+uGSGv7QAGaxHAzAtbE0lxYcAG9IarTWoyUGKrCtgAmslsvth/hpWQByCbXMXtI+hTZ24Twy2eiAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6aaa02d6f99a19167c4008e166ea517a2bd01aeb72a077663e6bc17ab533525f","last_reissued_at":"2026-07-05T10:16:07.513886Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:16:07.513886Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Second-Order Fine-Tuning without Pain for LLMs:A Hessian Informed Zeroth-Order Optimizer","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guang Dai, Haishan Ye, Ivor W.Tsang, Sizhe Dang, Yanjun Zhao, Yi Qian","submitted_at":"2024-02-23T08:11:55Z","abstract_excerpt":"Fine-tuning large language models (LLMs) with classic first-order optimizers entails prohibitive GPU memory due to the backpropagation process. Recent works have turned to zeroth-order optimizers for fine-tuning, which save substantial memory by using two forward passes. However, these optimizers are plagued by the heterogeneity of parameter curvatures across different dimensions. In this work, we propose HiZOO, a diagonal Hessian informed zeroth-order optimizer which is the first work to leverage the diagonal Hessian to enhance zeroth-order optimizer for fine-tuning LLMs. What's more, HiZOO a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.15173","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/2402.15173/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":"2402.15173","created_at":"2026-07-05T10:16:07.513955+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.15173v4","created_at":"2026-07-05T10:16:07.513955+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.15173","created_at":"2026-07-05T10:16:07.513955+00:00"},{"alias_kind":"pith_short_12","alias_value":"NKVAFVXZTIMR","created_at":"2026-07-05T10:16:07.513955+00:00"},{"alias_kind":"pith_short_16","alias_value":"NKVAFVXZTIMRM7CA","created_at":"2026-07-05T10:16:07.513955+00:00"},{"alias_kind":"pith_short_8","alias_value":"NKVAFVXZ","created_at":"2026-07-05T10:16:07.513955+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"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":36,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05516","citing_title":"Dominant-Layer ZO: A Single Layer Dominates Zeroth-Order Fine-Tuning of LLMs","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09034","citing_title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12813","citing_title":"REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations","ref_index":170,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09034","citing_title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration","ref_index":20,"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":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21368","citing_title":"Distributed Zeroth-Order Optimization with Rademacher Perturbations and Momentum Gradient Tracking","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI","json":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI.json","graph_json":"https://pith.science/api/pith-number/NKVAFVXZTIMRM7CABDQWN2SRPI/graph.json","events_json":"https://pith.science/api/pith-number/NKVAFVXZTIMRM7CABDQWN2SRPI/events.json","paper":"https://pith.science/paper/NKVAFVXZ"},"agent_actions":{"view_html":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI","download_json":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI.json","view_paper":"https://pith.science/paper/NKVAFVXZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.15173&json=true","fetch_graph":"https://pith.science/api/pith-number/NKVAFVXZTIMRM7CABDQWN2SRPI/graph.json","fetch_events":"https://pith.science/api/pith-number/NKVAFVXZTIMRM7CABDQWN2SRPI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI/action/storage_attestation","attest_author":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI/action/author_attestation","sign_citation":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI/action/citation_signature","submit_replication":"https://pith.science/pith/NKVAFVXZTIMRM7CABDQWN2SRPI/action/replication_record"}},"created_at":"2026-07-05T10:16:07.513955+00:00","updated_at":"2026-07-05T10:16:07.513955+00:00"}