{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:W4JVQ3RAM7EHHMIBEVN64VKDLL","short_pith_number":"pith:W4JVQ3RA","schema_version":"1.0","canonical_sha256":"b713586e2067c873b101255bee55435afbbb2c674d0a13a4043588cbfa1ce179","source":{"kind":"arxiv","id":"2602.16340","version":3},"attestation_state":"computed","paper":{"title":"The Implicit Bias of Adam and Muon on Smooth Homogeneous Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Eitan Gronich, Gal Vardi","submitted_at":"2026-02-18T10:25:07Z","abstract_excerpt":"We study the implicit bias of momentum-based optimizers on smooth homogeneous models. We show that \\textit{momentum steepest descent} algorithms like Muon (spectral norm), MomentumGD ($\\ell_2$ norm), and Signum ($\\ell_\\infty$ norm) are \\textit{approximate} steepest descent trajectories under a decaying learning rate schedule, proving that these algorithms have a bias towards KKT points of the corresponding margin maximization problem. We extend the analysis to Adam (without the stability constant), which maximizes the $\\ell_\\infty$ margin, and to Muon-Signum and Muon-Adam, which maximize a hyb"},"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":"2602.16340","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-18T10:25:07Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"77999e7dca449a26457d7817214516f3d8a2b366e518e26fc4eace031f243a20","abstract_canon_sha256":"aff8fd30733152405ac5eea2f574b759fba6d2e1abf24aca4dff4f8fb395f03d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:25.555748Z","signature_b64":"yGyITFaXPtTUzGZ30XCFVFbAhNboqzSHpbhw6FOKy+xvKnB6M3Df0NN/01gujR5GPfONhxQsC4Ld5M3V4dANDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b713586e2067c873b101255bee55435afbbb2c674d0a13a4043588cbfa1ce179","last_reissued_at":"2026-05-26T01:03:25.555034Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:25.555034Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Implicit Bias of Adam and Muon on Smooth Homogeneous Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Eitan Gronich, Gal Vardi","submitted_at":"2026-02-18T10:25:07Z","abstract_excerpt":"We study the implicit bias of momentum-based optimizers on smooth homogeneous models. We show that \\textit{momentum steepest descent} algorithms like Muon (spectral norm), MomentumGD ($\\ell_2$ norm), and Signum ($\\ell_\\infty$ norm) are \\textit{approximate} steepest descent trajectories under a decaying learning rate schedule, proving that these algorithms have a bias towards KKT points of the corresponding margin maximization problem. We extend the analysis to Adam (without the stability constant), which maximizes the $\\ell_\\infty$ margin, and to Muon-Signum and Muon-Adam, which maximize a hyb"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.16340","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/2602.16340/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":"2602.16340","created_at":"2026-05-26T01:03:25.555145+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.16340v3","created_at":"2026-05-26T01:03:25.555145+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.16340","created_at":"2026-05-26T01:03:25.555145+00:00"},{"alias_kind":"pith_short_12","alias_value":"W4JVQ3RAM7EH","created_at":"2026-05-26T01:03:25.555145+00:00"},{"alias_kind":"pith_short_16","alias_value":"W4JVQ3RAM7EHHMIB","created_at":"2026-05-26T01:03:25.555145+00:00"},{"alias_kind":"pith_short_8","alias_value":"W4JVQ3RA","created_at":"2026-05-26T01:03:25.555145+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.26977","citing_title":"Convergence of Spectral Descent for Non-smooth Optimization","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18528","citing_title":"Scale-Invariant Neural Network Optimization: Norm Geometry and Heavy-Tailed Noise","ref_index":150,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL","json":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL.json","graph_json":"https://pith.science/api/pith-number/W4JVQ3RAM7EHHMIBEVN64VKDLL/graph.json","events_json":"https://pith.science/api/pith-number/W4JVQ3RAM7EHHMIBEVN64VKDLL/events.json","paper":"https://pith.science/paper/W4JVQ3RA"},"agent_actions":{"view_html":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL","download_json":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL.json","view_paper":"https://pith.science/paper/W4JVQ3RA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.16340&json=true","fetch_graph":"https://pith.science/api/pith-number/W4JVQ3RAM7EHHMIBEVN64VKDLL/graph.json","fetch_events":"https://pith.science/api/pith-number/W4JVQ3RAM7EHHMIBEVN64VKDLL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL/action/storage_attestation","attest_author":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL/action/author_attestation","sign_citation":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL/action/citation_signature","submit_replication":"https://pith.science/pith/W4JVQ3RAM7EHHMIBEVN64VKDLL/action/replication_record"}},"created_at":"2026-05-26T01:03:25.555145+00:00","updated_at":"2026-05-26T01:03:25.555145+00:00"}