{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:6H4SEO2C6UGGTCHW5THJ7WEFRL","short_pith_number":"pith:6H4SEO2C","schema_version":"1.0","canonical_sha256":"f1f9223b42f50c6988f6ecce9fd8858af392b74e1d1aa80ce0001993dc23a3a6","source":{"kind":"arxiv","id":"2505.18886","version":1},"attestation_state":"computed","paper":{"title":"KerZOO: Kernel Function Informed Zeroth-Order Optimization for Accurate and Accelerated LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Geng Yuan, Qitao Tan, Shaoyi Huang, Xiaodong Yu, Zhendong Mi, Zining Zhu","submitted_at":"2025-05-24T21:56:03Z","abstract_excerpt":"Large language models (LLMs) have demonstrated impressive capabilities across numerous NLP tasks. Nevertheless, conventional first-order fine-tuning techniques impose heavy memory demands, creating practical obstacles to real-world applications. Zeroth-order (ZO) optimization has recently emerged as a promising memory-efficient alternative, as it circumvents the need for backpropagation by estimating gradients solely through forward passes--making it particularly suitable for resource-limited environments. Despite its efficiency, ZO optimization suffers from gradient estimation bias, which sig"},"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":"2505.18886","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-24T21:56:03Z","cross_cats_sorted":[],"title_canon_sha256":"1203c5fe690ab3ed0bcaa6098f94007dc135ffc46f3c5dcdb0b4fcff165b3609","abstract_canon_sha256":"de261e8ad10c666fefdef36672d42fe734378c2bbc07ef3d5e767335f54654de"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:21.270077Z","signature_b64":"fYpGU0bkfdjUMJeO4+Oc3Gg17Npz8OUgmlm34hZVFNcozKePDfwDwks89h6pbcxp7kvUGpsowvP4DH56OgNVDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f1f9223b42f50c6988f6ecce9fd8858af392b74e1d1aa80ce0001993dc23a3a6","last_reissued_at":"2026-07-05T11:09:21.269606Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:21.269606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KerZOO: Kernel Function Informed Zeroth-Order Optimization for Accurate and Accelerated LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Geng Yuan, Qitao Tan, Shaoyi Huang, Xiaodong Yu, Zhendong Mi, Zining Zhu","submitted_at":"2025-05-24T21:56:03Z","abstract_excerpt":"Large language models (LLMs) have demonstrated impressive capabilities across numerous NLP tasks. Nevertheless, conventional first-order fine-tuning techniques impose heavy memory demands, creating practical obstacles to real-world applications. Zeroth-order (ZO) optimization has recently emerged as a promising memory-efficient alternative, as it circumvents the need for backpropagation by estimating gradients solely through forward passes--making it particularly suitable for resource-limited environments. Despite its efficiency, ZO optimization suffers from gradient estimation bias, which sig"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18886","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/2505.18886/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":"2505.18886","created_at":"2026-07-05T11:09:21.269663+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.18886v1","created_at":"2026-07-05T11:09:21.269663+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18886","created_at":"2026-07-05T11:09:21.269663+00:00"},{"alias_kind":"pith_short_12","alias_value":"6H4SEO2C6UGG","created_at":"2026-07-05T11:09:21.269663+00:00"},{"alias_kind":"pith_short_16","alias_value":"6H4SEO2C6UGGTCHW","created_at":"2026-07-05T11:09:21.269663+00:00"},{"alias_kind":"pith_short_8","alias_value":"6H4SEO2C","created_at":"2026-07-05T11:09:21.269663+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.00761","citing_title":"Downgrade to Upgrade: Optimizer Simplification Enhances Robustness in LLM Unlearning","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03373","citing_title":"Learning Dynamics of Zeroth-Order Optimization: A Kernel Perspective","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL","json":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL.json","graph_json":"https://pith.science/api/pith-number/6H4SEO2C6UGGTCHW5THJ7WEFRL/graph.json","events_json":"https://pith.science/api/pith-number/6H4SEO2C6UGGTCHW5THJ7WEFRL/events.json","paper":"https://pith.science/paper/6H4SEO2C"},"agent_actions":{"view_html":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL","download_json":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL.json","view_paper":"https://pith.science/paper/6H4SEO2C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.18886&json=true","fetch_graph":"https://pith.science/api/pith-number/6H4SEO2C6UGGTCHW5THJ7WEFRL/graph.json","fetch_events":"https://pith.science/api/pith-number/6H4SEO2C6UGGTCHW5THJ7WEFRL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL/action/storage_attestation","attest_author":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL/action/author_attestation","sign_citation":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL/action/citation_signature","submit_replication":"https://pith.science/pith/6H4SEO2C6UGGTCHW5THJ7WEFRL/action/replication_record"}},"created_at":"2026-07-05T11:09:21.269663+00:00","updated_at":"2026-07-05T11:09:21.269663+00:00"}