{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AZK4VD3SA66SNL3ITM7CLBMS2Y","short_pith_number":"pith:AZK4VD3S","schema_version":"1.0","canonical_sha256":"0655ca8f7207bd26af689b3e258592d62921ac3bc309ccb67bcc31fcddfc5f03","source":{"kind":"arxiv","id":"2605.19811","version":1},"attestation_state":"computed","paper":{"title":"LionMuon: Alternating Spectral and Sign Descent for Efficient Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aleksandr Beznosikov, Andrey Veprikov, Arman Bolatov, Artem Riabinin, Martin Tak\\'a\\v{c}, Nikita Kornilov, Samuel Horv\\'ath","submitted_at":"2026-05-19T13:07:59Z","abstract_excerpt":"In large-scale optimization, the cheapness and effectiveness of update steps are the most crucial factors for a successful optimizer. Sign-based optimizers like Lion or Signum produce cheap per-step updates, whereas Muon's spectral matrix-sign update gives a much stronger direction at a substantially higher per-step cost. In this work, we propose LionMuon, which retains the effectiveness of Muon steps while considerably cutting the averaged iteration cost, similar to sign-based methods. It alternates between Lion's and Muon's updates on a fixed period P, sharing a single dual-EMA momentum buff"},"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":"2605.19811","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-19T13:07:59Z","cross_cats_sorted":[],"title_canon_sha256":"0e3709532fd091b629b1a33bcb25af3566b866cdd1843c5cf3963c295fadd5b3","abstract_canon_sha256":"697d9176a6a9d9c6dadd3d854432c1e3bfb2a297514f87d2c88d4ce453552229"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:06:15.723935Z","signature_b64":"84Rs9bRRBu8rXD1kbCjZx4RryzA7H4WQsZlxuudLah8s/v5fc6lyrny3gtf2HjIr1ETRGoaLwHtxKl32/3QxCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0655ca8f7207bd26af689b3e258592d62921ac3bc309ccb67bcc31fcddfc5f03","last_reissued_at":"2026-05-20T01:06:15.723272Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:06:15.723272Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LionMuon: Alternating Spectral and Sign Descent for Efficient Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aleksandr Beznosikov, Andrey Veprikov, Arman Bolatov, Artem Riabinin, Martin Tak\\'a\\v{c}, Nikita Kornilov, Samuel Horv\\'ath","submitted_at":"2026-05-19T13:07:59Z","abstract_excerpt":"In large-scale optimization, the cheapness and effectiveness of update steps are the most crucial factors for a successful optimizer. Sign-based optimizers like Lion or Signum produce cheap per-step updates, whereas Muon's spectral matrix-sign update gives a much stronger direction at a substantially higher per-step cost. In this work, we propose LionMuon, which retains the effectiveness of Muon steps while considerably cutting the averaged iteration cost, similar to sign-based methods. It alternates between Lion's and Muon's updates on a fixed period P, sharing a single dual-EMA momentum buff"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19811","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/2605.19811/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":"2605.19811","created_at":"2026-05-20T01:06:15.723378+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19811v1","created_at":"2026-05-20T01:06:15.723378+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19811","created_at":"2026-05-20T01:06:15.723378+00:00"},{"alias_kind":"pith_short_12","alias_value":"AZK4VD3SA66S","created_at":"2026-05-20T01:06:15.723378+00:00"},{"alias_kind":"pith_short_16","alias_value":"AZK4VD3SA66SNL3I","created_at":"2026-05-20T01:06:15.723378+00:00"},{"alias_kind":"pith_short_8","alias_value":"AZK4VD3S","created_at":"2026-05-20T01:06:15.723378+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/AZK4VD3SA66SNL3ITM7CLBMS2Y","json":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y.json","graph_json":"https://pith.science/api/pith-number/AZK4VD3SA66SNL3ITM7CLBMS2Y/graph.json","events_json":"https://pith.science/api/pith-number/AZK4VD3SA66SNL3ITM7CLBMS2Y/events.json","paper":"https://pith.science/paper/AZK4VD3S"},"agent_actions":{"view_html":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y","download_json":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y.json","view_paper":"https://pith.science/paper/AZK4VD3S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19811&json=true","fetch_graph":"https://pith.science/api/pith-number/AZK4VD3SA66SNL3ITM7CLBMS2Y/graph.json","fetch_events":"https://pith.science/api/pith-number/AZK4VD3SA66SNL3ITM7CLBMS2Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y/action/storage_attestation","attest_author":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y/action/author_attestation","sign_citation":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y/action/citation_signature","submit_replication":"https://pith.science/pith/AZK4VD3SA66SNL3ITM7CLBMS2Y/action/replication_record"}},"created_at":"2026-05-20T01:06:15.723378+00:00","updated_at":"2026-05-20T01:06:15.723378+00:00"}