{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5X47XB6UG424ITSXSMO7ZLMCTY","short_pith_number":"pith:5X47XB6U","schema_version":"1.0","canonical_sha256":"edf9fb87d43735c44e57931dfcad829e221405acfda987f11153ed6c3beabe01","source":{"kind":"arxiv","id":"2607.01125","version":1},"attestation_state":"computed","paper":{"title":"ZO-Act: Efficient Zeroth-Order Fine-Tuning via One-Shot Activation-Informed Low-Rank Subspaces","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Naigang Wang, Penghang Yin, Xin Li, Xun Dong, Yibo Xu, Zi Yang","submitted_at":"2026-07-01T16:12:50Z","abstract_excerpt":"Zeroth-order (ZO) optimization enables fine-tuning large language models when backpropagation is unavailable or memory-prohibitive, but existing methods often perturb full model weights or randomly constructed low-dimensional subspaces, yielding high-variance estimates and limited performance. We propose ZO-Act, an activation-informed ZO fine-tuning method that restricts perturbations to a fixed low-rank subspace derived from input activations. For each linear layer, ZO-Act computes a small activation basis once at initialization and optimizes only lightweight coefficient matrices using forwar"},"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":"2607.01125","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-01T16:12:50Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"53c9c8713a8c322d469419a1b4b6eed24795b457ff1090ce11e64610558b291a","abstract_canon_sha256":"633ac640125f9f806077958118188e661b4b8690970a6eb15abaf9fe7b9d9a3a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-02T01:18:29.807188Z","signature_b64":"nW2zKfR5UzlBBumW8KPfep85M54ZEg7KqqCRAkZcfOJ+KxNaACWlI7IXcrMlypgVKvTASlUBauKFaWSDnRrMAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"edf9fb87d43735c44e57931dfcad829e221405acfda987f11153ed6c3beabe01","last_reissued_at":"2026-07-02T01:18:29.806782Z","signature_status":"signed_v1","first_computed_at":"2026-07-02T01:18:29.806782Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ZO-Act: Efficient Zeroth-Order Fine-Tuning via One-Shot Activation-Informed Low-Rank Subspaces","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Naigang Wang, Penghang Yin, Xin Li, Xun Dong, Yibo Xu, Zi Yang","submitted_at":"2026-07-01T16:12:50Z","abstract_excerpt":"Zeroth-order (ZO) optimization enables fine-tuning large language models when backpropagation is unavailable or memory-prohibitive, but existing methods often perturb full model weights or randomly constructed low-dimensional subspaces, yielding high-variance estimates and limited performance. We propose ZO-Act, an activation-informed ZO fine-tuning method that restricts perturbations to a fixed low-rank subspace derived from input activations. For each linear layer, ZO-Act computes a small activation basis once at initialization and optimizes only lightweight coefficient matrices using forwar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.01125","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/2607.01125/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":"2607.01125","created_at":"2026-07-02T01:18:29.806837+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.01125v1","created_at":"2026-07-02T01:18:29.806837+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.01125","created_at":"2026-07-02T01:18:29.806837+00:00"},{"alias_kind":"pith_short_12","alias_value":"5X47XB6UG424","created_at":"2026-07-02T01:18:29.806837+00:00"},{"alias_kind":"pith_short_16","alias_value":"5X47XB6UG424ITSX","created_at":"2026-07-02T01:18:29.806837+00:00"},{"alias_kind":"pith_short_8","alias_value":"5X47XB6U","created_at":"2026-07-02T01:18:29.806837+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY","json":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY.json","graph_json":"https://pith.science/api/pith-number/5X47XB6UG424ITSXSMO7ZLMCTY/graph.json","events_json":"https://pith.science/api/pith-number/5X47XB6UG424ITSXSMO7ZLMCTY/events.json","paper":"https://pith.science/paper/5X47XB6U"},"agent_actions":{"view_html":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY","download_json":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY.json","view_paper":"https://pith.science/paper/5X47XB6U","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.01125&json=true","fetch_graph":"https://pith.science/api/pith-number/5X47XB6UG424ITSXSMO7ZLMCTY/graph.json","fetch_events":"https://pith.science/api/pith-number/5X47XB6UG424ITSXSMO7ZLMCTY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY/action/storage_attestation","attest_author":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY/action/author_attestation","sign_citation":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY/action/citation_signature","submit_replication":"https://pith.science/pith/5X47XB6UG424ITSXSMO7ZLMCTY/action/replication_record"}},"created_at":"2026-07-02T01:18:29.806837+00:00","updated_at":"2026-07-02T01:18:29.806837+00:00"}