{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:WYOTTIM7ROJ4XIBGSHJUUJL54H","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"bb98591811764327e9dcc45e04ece45f1493c8f7f3df164546662746b696236a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T16:16:45Z","title_canon_sha256":"99cf8b2f472b177bbce07c93deb1fc36c779117d99a8bd71f72d650e982582fe"},"schema_version":"1.0","source":{"id":"2605.09034","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.09034","created_at":"2026-05-20T00:01:43Z"},{"alias_kind":"arxiv_version","alias_value":"2605.09034v2","created_at":"2026-05-20T00:01:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.09034","created_at":"2026-05-20T00:01:43Z"},{"alias_kind":"pith_short_12","alias_value":"WYOTTIM7ROJ4","created_at":"2026-05-20T00:01:43Z"},{"alias_kind":"pith_short_16","alias_value":"WYOTTIM7ROJ4XIBG","created_at":"2026-05-20T00:01:43Z"},{"alias_kind":"pith_short_8","alias_value":"WYOTTIM7","created_at":"2026-05-20T00:01:43Z"}],"graph_snapshots":[{"event_id":"sha256:e86d3fdb746a737e700c895546ee716cbfae98427e0ca9423744e00003491190","target":"graph","created_at":"2026-05-20T00:01:43Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Experiments on LLM fine-tuning show that our method can achieve from 1.5x to 4x the convergence speed of ZO-Muon, the current SOTA algorithm, across SuperGlue datasets in the OPT-13B model."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that projecting the search space onto a momentum-derived subspace sufficiently lowers gradient variance to stabilize the streaming power-iteration procedure and enable effective partial orthogonalization in the zeroth-order regime (abstract, paragraph on streaming variant)."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ZO-MOPI accelerates zeroth-order LLM fine-tuning by applying partial spectral orthogonalization from power iteration inside a momentum-projected subspace to reduce variance and exploit dominant directions."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Partial orthogonalization via power iteration accelerates zeroth-order fine-tuning of large language models."}],"snapshot_sha256":"4d7a9efff25927d6fe9bd319bf8316ea0d31673a41574332e9d1b313f506d991"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"45419483cb771843c2b0efc9430898bd183041cf1e157286c56184090b577c40"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T20:38:49.630487Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T13:31:19.289094Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T10:32:51.236600Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.09034/integrity.json","findings":[],"snapshot_sha256":"914af79f9f85df6ed1c63de74cd5b79ec755b3e4e6dd922eeace2d9edc4b54bf","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Zeroth-order (ZO) optimization has become increasingly popular and important in fine-tuning large language models (LLMs), especially on edge devices due to its ability to adjust the model to local data without the need for memory-intensive back-propagation. Recent works try to reduce ZO variance through low-dimensional subspace search, but subspace restriction alone leaves key optimization geometry under-exploited, motivating additional acceleration. In this work, we focus on the hidden layer training problem in which spectral optimizers like Muon outperform AdamW due to its ability to exploit","authors_text":"Jiahe Chen, Ziye Ma","cross_cats":[],"headline":"Partial orthogonalization via power iteration accelerates zeroth-order fine-tuning of large language models.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T16:16:45Z","title":"Accelerating Zeroth-Order Spectral Optimization with Partial Orthogonalization from Power Iteration"},"references":{"count":26,"internal_anchors":6,"resolved_work":26,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Dion: Distributed orthonormal- ized updates.arXiv preprint: 2504.05295","work_id":"4ecfbe15-0efd-41f4-a129-82f6a2164d2f","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Enhancing zeroth-order fine-tuning for language models with low-rank structures.arXiv preprint arXiv:2410.07698","work_id":"93562c56-6202-4ca4-a4de-cadf1aa59ee5","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Boolq: Exploring the surprising difficulty of natural yes/no questions","work_id":"873a4f9b-c4f6-4b39-8ae6-2abd7ce8e5ea","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Variance- reduced zeroth-order methods for fine-tuning language models.arXiv preprint arXiv:2404.08080","work_id":"781b7a33-777b-4e7b-8974-383845c8da8a","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"ARO : A New Lens On Matrix Optimization For Large Models","work_id":"0b1b41e4-a2fe-43e0-a97d-1b3ba9fbf25c","year":null}],"snapshot_sha256":"1ec1e7da4dddbe9d5ad8efa056ebdc2d257ab7925755acfb3b132a6f730c2107"},"source":{"id":"2605.09034","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-19T17:29:05.372590Z","id":"f74a27e0-851c-469c-8b5e-23f6c21784bd","model_set":{"reader":"grok-4.3"},"one_line_summary":"ZO-MOPI accelerates zeroth-order LLM fine-tuning by applying partial spectral orthogonalization from power iteration inside a momentum-projected subspace to reduce variance and exploit dominant directions.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Partial orthogonalization via power iteration accelerates zeroth-order fine-tuning of large language models.","strongest_claim":"Experiments on LLM fine-tuning show that our method can achieve from 1.5x to 4x the convergence speed of ZO-Muon, the current SOTA algorithm, across SuperGlue datasets in the OPT-13B model.","weakest_assumption":"The assumption that projecting the search space onto a momentum-derived subspace sufficiently lowers gradient variance to stabilize the streaming power-iteration procedure and enable effective partial orthogonalization in the zeroth-order regime (abstract, paragraph on streaming variant)."}},"verdict_id":"f74a27e0-851c-469c-8b5e-23f6c21784bd"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:5c96af07cf6c55c7491ef8291091b9fae0894fe80cd1c6e484652aed3375223f","target":"record","created_at":"2026-05-20T00:01:43Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"bb98591811764327e9dcc45e04ece45f1493c8f7f3df164546662746b696236a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T16:16:45Z","title_canon_sha256":"99cf8b2f472b177bbce07c93deb1fc36c779117d99a8bd71f72d650e982582fe"},"schema_version":"1.0","source":{"id":"2605.09034","kind":"arxiv","version":2}},"canonical_sha256":"b61d39a19f8b93cba02691d34a257de1c5155204de04de9c2df55619a9e876da","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b61d39a19f8b93cba02691d34a257de1c5155204de04de9c2df55619a9e876da","first_computed_at":"2026-05-20T00:01:43.050728Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:01:43.050728Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"OTP+0y0m/M93qyTSH/F4hjfqBkpBl2LsjPRJV/AEvRwcMzVCG48GVy4k9evpsZt/eWrmv6Pip0UtcoJW1x7xAw==","signature_status":"signed_v1","signed_at":"2026-05-20T00:01:43.051519Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.09034","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5c96af07cf6c55c7491ef8291091b9fae0894fe80cd1c6e484652aed3375223f","sha256:e86d3fdb746a737e700c895546ee716cbfae98427e0ca9423744e00003491190"],"state_sha256":"aa884ffca6636dd059bde3326befd0610073cf33c1b2df04b0c9414938e94263"}