{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QKH3YQYAPUSK3X6XDWO4V4DDAZ","short_pith_number":"pith:QKH3YQYA","schema_version":"1.0","canonical_sha256":"828fbc43007d24addfd71d9dcaf0630662c1af204739de00495489eb2287142a","source":{"kind":"arxiv","id":"2410.08989","version":3},"attestation_state":"computed","paper":{"title":"Zeroth-Order Fine-Tuning of LLMs in Random Subspaces","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hua Huang, Jia Li, Mi Tian, Pan Zhou, Sike Wang, Ziming Yu","submitted_at":"2024-10-11T17:01:43Z","abstract_excerpt":"Fine-tuning Large Language Models (LLMs) has proven effective for a variety of downstream tasks. However, as LLMs grow in size, the memory demands for backpropagation become increasingly prohibitive. Zeroth-order (ZO) optimization methods offer a memory-efficient alternative by using forward passes to estimate gradients, but the variance of gradient estimates typically scales linearly with the model's parameter dimension$\\unicode{x2013}$a significant issue for LLMs. In this paper, we propose the random Subspace Zeroth-order (SubZero) optimization to address the challenges posed by LLMs' high d"},"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":"2410.08989","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-11T17:01:43Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c9efa3f765463be2176590d68d89670b2137482e803b8ef9e294f857e6666eaa","abstract_canon_sha256":"28cb97df8dcea103a1233be58e8f47833fdfba02fb2b4ec8bbac8fe9645b9570"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:42:23.338908Z","signature_b64":"BRzX0as8xHJfKSgnllCpYRA3t/CPoZEuHkyWrS1Ewrm7haaN1XgoawM/CoGTHefoaaHV/B+purqdzYZOuyY/Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"828fbc43007d24addfd71d9dcaf0630662c1af204739de00495489eb2287142a","last_reissued_at":"2026-07-05T11:42:23.338415Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:42:23.338415Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Zeroth-Order Fine-Tuning of LLMs in Random Subspaces","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hua Huang, Jia Li, Mi Tian, Pan Zhou, Sike Wang, Ziming Yu","submitted_at":"2024-10-11T17:01:43Z","abstract_excerpt":"Fine-tuning Large Language Models (LLMs) has proven effective for a variety of downstream tasks. However, as LLMs grow in size, the memory demands for backpropagation become increasingly prohibitive. Zeroth-order (ZO) optimization methods offer a memory-efficient alternative by using forward passes to estimate gradients, but the variance of gradient estimates typically scales linearly with the model's parameter dimension$\\unicode{x2013}$a significant issue for LLMs. In this paper, we propose the random Subspace Zeroth-order (SubZero) optimization to address the challenges posed by LLMs' high d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.08989","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/2410.08989/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":"2410.08989","created_at":"2026-07-05T11:42:23.338477+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.08989v3","created_at":"2026-07-05T11:42:23.338477+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.08989","created_at":"2026-07-05T11:42:23.338477+00:00"},{"alias_kind":"pith_short_12","alias_value":"QKH3YQYAPUSK","created_at":"2026-07-05T11:42:23.338477+00:00"},{"alias_kind":"pith_short_16","alias_value":"QKH3YQYAPUSK3X6X","created_at":"2026-07-05T11:42:23.338477+00:00"},{"alias_kind":"pith_short_8","alias_value":"QKH3YQYA","created_at":"2026-07-05T11:42:23.338477+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.17264","citing_title":"OAT-Rephrase: Optimization-Aware Training Data Rephrasing for Zeroth-Order LLM Fine-Tuning","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ","json":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ.json","graph_json":"https://pith.science/api/pith-number/QKH3YQYAPUSK3X6XDWO4V4DDAZ/graph.json","events_json":"https://pith.science/api/pith-number/QKH3YQYAPUSK3X6XDWO4V4DDAZ/events.json","paper":"https://pith.science/paper/QKH3YQYA"},"agent_actions":{"view_html":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ","download_json":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ.json","view_paper":"https://pith.science/paper/QKH3YQYA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.08989&json=true","fetch_graph":"https://pith.science/api/pith-number/QKH3YQYAPUSK3X6XDWO4V4DDAZ/graph.json","fetch_events":"https://pith.science/api/pith-number/QKH3YQYAPUSK3X6XDWO4V4DDAZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ/action/storage_attestation","attest_author":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ/action/author_attestation","sign_citation":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ/action/citation_signature","submit_replication":"https://pith.science/pith/QKH3YQYAPUSK3X6XDWO4V4DDAZ/action/replication_record"}},"created_at":"2026-07-05T11:42:23.338477+00:00","updated_at":"2026-07-05T11:42:23.338477+00:00"}