{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZS2SKAOOPR7PRYBOJLEYQNVX2I","short_pith_number":"pith:ZS2SKAOO","schema_version":"1.0","canonical_sha256":"ccb52501ce7c7ef8e02e4ac98836b7d209df2144765ef2366c12dc9fb3ae21d1","source":{"kind":"arxiv","id":"2604.14669","version":2},"attestation_state":"computed","paper":{"title":"Zeroth-Order Optimization at the Edge of Stability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Zeroth-order methods remain mean-square stable only when their step size satisfies a bound that depends on the full Hessian spectrum rather than its largest eigenvalue alone.","cross_cats":["math.DS","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bingcong Li, Liang Zhang, Michael Muehlebach, Minhak Song, Niao He, Sewoong Oh","submitted_at":"2026-04-16T06:23:18Z","abstract_excerpt":"Zeroth-order (ZO) methods are widely used when gradients are unavailable or prohibitively expensive, including black-box learning and memory-efficient fine-tuning of large models, yet their optimization dynamics in deep learning remain underexplored. In this work, we provide an explicit step size condition that exactly captures the (mean-square) linear stability of a family of ZO methods based on the standard two-point estimator. Our characterization reveals a sharp contrast with first-order (FO) methods: whereas FO stability is governed solely by the largest Hessian eigenvalue, mean-square st"},"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":"2604.14669","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-16T06:23:18Z","cross_cats_sorted":["math.DS","math.OC","stat.ML"],"title_canon_sha256":"579a22f9f3d9dd0870be92dac9c81190f376974fd9f771bf37ac8126983d8a4f","abstract_canon_sha256":"1e3b5b5f0d85cc6d5838d159f8430a92acd19e59916cfe0e1b08dc4e9ac8fc9f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-02T01:17:31.260011Z","signature_b64":"rKPFTcflGMyZ8SM351XvgJFvZAdW8YQU1wEdZA57rR9ry20NAkyVq7q4wmizj3pfjcRdIzAwskYiAU6sDd0CBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ccb52501ce7c7ef8e02e4ac98836b7d209df2144765ef2366c12dc9fb3ae21d1","last_reissued_at":"2026-07-02T01:17:31.259570Z","signature_status":"signed_v1","first_computed_at":"2026-07-02T01:17:31.259570Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Zeroth-Order Optimization at the Edge of Stability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Zeroth-order methods remain mean-square stable only when their step size satisfies a bound that depends on the full Hessian spectrum rather than its largest eigenvalue alone.","cross_cats":["math.DS","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bingcong Li, Liang Zhang, Michael Muehlebach, Minhak Song, Niao He, Sewoong Oh","submitted_at":"2026-04-16T06:23:18Z","abstract_excerpt":"Zeroth-order (ZO) methods are widely used when gradients are unavailable or prohibitively expensive, including black-box learning and memory-efficient fine-tuning of large models, yet their optimization dynamics in deep learning remain underexplored. In this work, we provide an explicit step size condition that exactly captures the (mean-square) linear stability of a family of ZO methods based on the standard two-point estimator. Our characterization reveals a sharp contrast with first-order (FO) methods: whereas FO stability is governed solely by the largest Hessian eigenvalue, mean-square st"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we provide an explicit step size condition that exactly captures the (mean-square) linear stability of a family of ZO methods based on the standard two-point estimator... mean-square stability of ZO methods depends on the entire Hessian spectrum.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The linear stability analysis assumes the loss can be locally approximated by a quadratic form whose Hessian is constant over the relevant trajectory, and that the two-point estimator is used without additional smoothing or variance reduction.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Zeroth-order methods achieve mean-square stability when the step size satisfies a condition involving the entire Hessian spectrum, with full-batch ZO optimizers operating at the edge of stability and large steps regularizing the Hessian trace.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Zeroth-order methods remain mean-square stable only when their step size satisfies a bound that depends on the full Hessian spectrum rather than its largest eigenvalue alone.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"532643a344f2c4571bf4ea935d01ee819ff3b696b8d96b7c29d30a87f1eb086c"},"source":{"id":"2604.14669","kind":"arxiv","version":2},"verdict":{"id":"64c66983-492d-41d4-bcd9-fbb0eec76e7d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T11:15:27.370755Z","strongest_claim":"we provide an explicit step size condition that exactly captures the (mean-square) linear stability of a family of ZO methods based on the standard two-point estimator... mean-square stability of ZO methods depends on the entire Hessian spectrum.","one_line_summary":"Zeroth-order methods achieve mean-square stability when the step size satisfies a condition involving the entire Hessian spectrum, with full-batch ZO optimizers operating at the edge of stability and large steps regularizing the Hessian trace.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The linear stability analysis assumes the loss can be locally approximated by a quadratic form whose Hessian is constant over the relevant trajectory, and that the two-point estimator is used without additional smoothing or variance reduction.","pith_extraction_headline":"Zeroth-order methods remain mean-square stable only when their step size satisfies a bound that depends on the full Hessian spectrum rather than its largest eigenvalue alone."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.14669/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":"2604.14669","created_at":"2026-07-02T01:17:31.259626+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.14669v2","created_at":"2026-07-02T01:17:31.259626+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.14669","created_at":"2026-07-02T01:17:31.259626+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZS2SKAOOPR7P","created_at":"2026-07-02T01:17:31.259626+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZS2SKAOOPR7PRYBO","created_at":"2026-07-02T01:17:31.259626+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZS2SKAOO","created_at":"2026-07-02T01:17:31.259626+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/ZS2SKAOOPR7PRYBOJLEYQNVX2I","json":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I.json","graph_json":"https://pith.science/api/pith-number/ZS2SKAOOPR7PRYBOJLEYQNVX2I/graph.json","events_json":"https://pith.science/api/pith-number/ZS2SKAOOPR7PRYBOJLEYQNVX2I/events.json","paper":"https://pith.science/paper/ZS2SKAOO"},"agent_actions":{"view_html":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I","download_json":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I.json","view_paper":"https://pith.science/paper/ZS2SKAOO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.14669&json=true","fetch_graph":"https://pith.science/api/pith-number/ZS2SKAOOPR7PRYBOJLEYQNVX2I/graph.json","fetch_events":"https://pith.science/api/pith-number/ZS2SKAOOPR7PRYBOJLEYQNVX2I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I/action/storage_attestation","attest_author":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I/action/author_attestation","sign_citation":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I/action/citation_signature","submit_replication":"https://pith.science/pith/ZS2SKAOOPR7PRYBOJLEYQNVX2I/action/replication_record"}},"created_at":"2026-07-02T01:17:31.259626+00:00","updated_at":"2026-07-02T01:17:31.259626+00:00"}