{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MJNSP2MSW7S5WYDCPRP2Z3DLAH","short_pith_number":"pith:MJNSP2MS","schema_version":"1.0","canonical_sha256":"625b27e992b7e5db60627c5facec6b01eacc3d118d3dd3814847872d736fce3c","source":{"kind":"arxiv","id":"2506.15025","version":1},"attestation_state":"computed","paper":{"title":"Optimal Embedding Learning Rate in LLMs: The Effect of Vocabulary Size","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Liyuan Liu, Soufiane Hayou","submitted_at":"2025-06-17T23:57:30Z","abstract_excerpt":"Pretraining large language models is a costly process. To make this process more efficient, several methods have been proposed to optimize model architecture/parametrization and hardware use. On the parametrization side, $\\mu P$ (Maximal Update Parametrization) parametrizes model weights and learning rate (LR) in a way that makes hyperparameters (HPs) transferable with width (embedding dimension): HPs can be tuned for a small model and used for larger models without additional tuning. While $\\mu$P showed impressive results in practice, recent empirical studies have reported conflicting observa"},"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":"2506.15025","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-17T23:57:30Z","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"title_canon_sha256":"313722b1ffa0f2b5135ac071d3359579e03a450aeeaa1ded1c14f3bccb69d09b","abstract_canon_sha256":"9d91185da0c7126caae37684885d3f2673b6bff43f69cfc3bffab73acc8cf3da"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:23:35.735414Z","signature_b64":"4UZtgR+AD2O/ktAByQdU77IhepoZ8ExRkLwnaK2xeF64CWCrijQcb8HdSaP5USyg+FoG1+oHuWGC1jehWasGAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"625b27e992b7e5db60627c5facec6b01eacc3d118d3dd3814847872d736fce3c","last_reissued_at":"2026-07-05T11:23:35.734835Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:23:35.734835Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimal Embedding Learning Rate in LLMs: The Effect of Vocabulary Size","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Liyuan Liu, Soufiane Hayou","submitted_at":"2025-06-17T23:57:30Z","abstract_excerpt":"Pretraining large language models is a costly process. To make this process more efficient, several methods have been proposed to optimize model architecture/parametrization and hardware use. On the parametrization side, $\\mu P$ (Maximal Update Parametrization) parametrizes model weights and learning rate (LR) in a way that makes hyperparameters (HPs) transferable with width (embedding dimension): HPs can be tuned for a small model and used for larger models without additional tuning. While $\\mu$P showed impressive results in practice, recent empirical studies have reported conflicting observa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.15025","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/2506.15025/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":"2506.15025","created_at":"2026-07-05T11:23:35.734901+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.15025v1","created_at":"2026-07-05T11:23:35.734901+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15025","created_at":"2026-07-05T11:23:35.734901+00:00"},{"alias_kind":"pith_short_12","alias_value":"MJNSP2MSW7S5","created_at":"2026-07-05T11:23:35.734901+00:00"},{"alias_kind":"pith_short_16","alias_value":"MJNSP2MSW7S5WYDC","created_at":"2026-07-05T11:23:35.734901+00:00"},{"alias_kind":"pith_short_8","alias_value":"MJNSP2MS","created_at":"2026-07-05T11:23:35.734901+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22297","citing_title":"One LR Doesn't Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMs","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22297","citing_title":"One LR Doesn't Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMs","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21486","citing_title":"Quantifying Hyperparameter Transfer and the Importance of Embedding Layer Learning Rate","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH","json":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH.json","graph_json":"https://pith.science/api/pith-number/MJNSP2MSW7S5WYDCPRP2Z3DLAH/graph.json","events_json":"https://pith.science/api/pith-number/MJNSP2MSW7S5WYDCPRP2Z3DLAH/events.json","paper":"https://pith.science/paper/MJNSP2MS"},"agent_actions":{"view_html":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH","download_json":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH.json","view_paper":"https://pith.science/paper/MJNSP2MS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.15025&json=true","fetch_graph":"https://pith.science/api/pith-number/MJNSP2MSW7S5WYDCPRP2Z3DLAH/graph.json","fetch_events":"https://pith.science/api/pith-number/MJNSP2MSW7S5WYDCPRP2Z3DLAH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH/action/storage_attestation","attest_author":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH/action/author_attestation","sign_citation":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH/action/citation_signature","submit_replication":"https://pith.science/pith/MJNSP2MSW7S5WYDCPRP2Z3DLAH/action/replication_record"}},"created_at":"2026-07-05T11:23:35.734901+00:00","updated_at":"2026-07-05T11:23:35.734901+00:00"}