{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TGGTWQNSM3MXPWHVAYZ5UYQBU3","short_pith_number":"pith:TGGTWQNS","schema_version":"1.0","canonical_sha256":"998d3b41b266d977d8f50633da6201a6f89f1b0c5ce7aa77dc3a0871551a490a","source":{"kind":"arxiv","id":"2404.08080","version":1},"attestation_state":"computed","paper":{"title":"Variance-reduced Zeroth-Order Methods for Fine-Tuning Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","math.OC"],"primary_cat":"cs.LG","authors_text":"Hao Zhou, Parameswaran Raman, Tanmay Gautam, Wooseok Ha, Youngsuk Park","submitted_at":"2024-04-11T18:35:49Z","abstract_excerpt":"Fine-tuning language models (LMs) has demonstrated success in a wide array of downstream tasks. However, as LMs are scaled up, the memory requirements for backpropagation become prohibitively high. Zeroth-order (ZO) optimization methods can leverage memory-efficient forward passes to estimate gradients. More recently, MeZO, an adaptation of ZO-SGD, has been shown to consistently outperform zero-shot and in-context learning when combined with suitable task prompts. In this work, we couple ZO methods with variance reduction techniques to enhance stability and convergence for inference-based LM f"},"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":"2404.08080","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-04-11T18:35:49Z","cross_cats_sorted":["cs.AI","cs.CL","math.OC"],"title_canon_sha256":"9562a2f69554d57b162fcf3128cc02c416dd9c0f21c3aeab520d5d0b0f057f17","abstract_canon_sha256":"14c46ab08506cf8eb6313f900a05d9200c85bac1e79f035fc3ff7e54666fe259"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:07:02.307297Z","signature_b64":"PCDxErNT4xajLL7SudKKZzMc31+RvBIPjcUmtmgEiq5Uxm/wiHlHtmk01W/QThcYMqO965mvJ7LE+wnIGZODBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"998d3b41b266d977d8f50633da6201a6f89f1b0c5ce7aa77dc3a0871551a490a","last_reissued_at":"2026-07-05T08:07:02.306848Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:07:02.306848Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Variance-reduced Zeroth-Order Methods for Fine-Tuning Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","math.OC"],"primary_cat":"cs.LG","authors_text":"Hao Zhou, Parameswaran Raman, Tanmay Gautam, Wooseok Ha, Youngsuk Park","submitted_at":"2024-04-11T18:35:49Z","abstract_excerpt":"Fine-tuning language models (LMs) has demonstrated success in a wide array of downstream tasks. However, as LMs are scaled up, the memory requirements for backpropagation become prohibitively high. Zeroth-order (ZO) optimization methods can leverage memory-efficient forward passes to estimate gradients. More recently, MeZO, an adaptation of ZO-SGD, has been shown to consistently outperform zero-shot and in-context learning when combined with suitable task prompts. In this work, we couple ZO methods with variance reduction techniques to enhance stability and convergence for inference-based LM f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.08080","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/2404.08080/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":"2404.08080","created_at":"2026-07-05T08:07:02.306904+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.08080v1","created_at":"2026-07-05T08:07:02.306904+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.08080","created_at":"2026-07-05T08:07:02.306904+00:00"},{"alias_kind":"pith_short_12","alias_value":"TGGTWQNSM3MX","created_at":"2026-07-05T08:07:02.306904+00:00"},{"alias_kind":"pith_short_16","alias_value":"TGGTWQNSM3MXPWHV","created_at":"2026-07-05T08:07:02.306904+00:00"},{"alias_kind":"pith_short_8","alias_value":"TGGTWQNS","created_at":"2026-07-05T08:07:02.306904+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"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":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15622","citing_title":"Position: Zeroth-Order Optimization in Deep Learning Is Underexplored, Not Underpowered","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09034","citing_title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18993","citing_title":"CR-Net: Scaling Parameter-Efficient Training with Cross-Layer Low-Rank Structure","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09034","citing_title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3","json":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3.json","graph_json":"https://pith.science/api/pith-number/TGGTWQNSM3MXPWHVAYZ5UYQBU3/graph.json","events_json":"https://pith.science/api/pith-number/TGGTWQNSM3MXPWHVAYZ5UYQBU3/events.json","paper":"https://pith.science/paper/TGGTWQNS"},"agent_actions":{"view_html":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3","download_json":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3.json","view_paper":"https://pith.science/paper/TGGTWQNS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.08080&json=true","fetch_graph":"https://pith.science/api/pith-number/TGGTWQNSM3MXPWHVAYZ5UYQBU3/graph.json","fetch_events":"https://pith.science/api/pith-number/TGGTWQNSM3MXPWHVAYZ5UYQBU3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3/action/storage_attestation","attest_author":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3/action/author_attestation","sign_citation":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3/action/citation_signature","submit_replication":"https://pith.science/pith/TGGTWQNSM3MXPWHVAYZ5UYQBU3/action/replication_record"}},"created_at":"2026-07-05T08:07:02.306904+00:00","updated_at":"2026-07-05T08:07:02.306904+00:00"}