{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NFZGS7BDFWNR2UBDYKPXZDGOM7","short_pith_number":"pith:NFZGS7BD","schema_version":"1.0","canonical_sha256":"6972697c232d9b1d5023c29f7c8cce67e803eafdd945f09277c80697cd0a201f","source":{"kind":"arxiv","id":"2409.19913","version":3},"attestation_state":"computed","paper":{"title":"Scaling Optimal LR Across Token Horizons","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alon Benhaim, Furu Wei, Johan Bjorck, Vishrav Chaudhary, Xia Song","submitted_at":"2024-09-30T03:32:02Z","abstract_excerpt":"State-of-the-art LLMs are powered by scaling -- scaling model size, dataset size and cluster size. It is economically infeasible to extensively tune hyperparameter for the largest runs. Instead, approximately optimal hyperparameters must be inferred or \\textit{transferred} from smaller experiments. Hyperparameter transfer across model sizes has been studied in Yang et al. However, hyperparameter transfer across dataset size -- or token horizon -- has not been studied yet. To remedy this we conduct a large scale empirical study on how optimal learning rate (LR) depends on token horizon in LLM t"},"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":"2409.19913","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-09-30T03:32:02Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"3964b1083d4e42bc457b3c238c32b1e756bdec30479658ad126f7c9874fca040","abstract_canon_sha256":"40283f870db5d43f03bc5bc9c34d1cd4c38294eeff187af29bc7e24e02db7b4a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:42.087033Z","signature_b64":"Xyyu+RX47yok/CNhn0gcF3twMwuLfWxNTqI5VTKdOq1bGmbWuQgiXgNoVNU5Iqyc6Npyi0eBO/VHMu7CDSRZAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6972697c232d9b1d5023c29f7c8cce67e803eafdd945f09277c80697cd0a201f","last_reissued_at":"2026-07-05T10:18:42.086203Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:42.086203Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Optimal LR Across Token Horizons","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alon Benhaim, Furu Wei, Johan Bjorck, Vishrav Chaudhary, Xia Song","submitted_at":"2024-09-30T03:32:02Z","abstract_excerpt":"State-of-the-art LLMs are powered by scaling -- scaling model size, dataset size and cluster size. It is economically infeasible to extensively tune hyperparameter for the largest runs. Instead, approximately optimal hyperparameters must be inferred or \\textit{transferred} from smaller experiments. Hyperparameter transfer across model sizes has been studied in Yang et al. However, hyperparameter transfer across dataset size -- or token horizon -- has not been studied yet. To remedy this we conduct a large scale empirical study on how optimal learning rate (LR) depends on token horizon in LLM t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.19913","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/2409.19913/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":"2409.19913","created_at":"2026-07-05T10:18:42.086643+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.19913v3","created_at":"2026-07-05T10:18:42.086643+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.19913","created_at":"2026-07-05T10:18:42.086643+00:00"},{"alias_kind":"pith_short_12","alias_value":"NFZGS7BDFWNR","created_at":"2026-07-05T10:18:42.086643+00:00"},{"alias_kind":"pith_short_16","alias_value":"NFZGS7BDFWNR2UBD","created_at":"2026-07-05T10:18:42.086643+00:00"},{"alias_kind":"pith_short_8","alias_value":"NFZGS7BD","created_at":"2026-07-05T10:18:42.086643+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01487","citing_title":"How to Allocate Your Tokens? Scaling Laws with Training Steps and Batch Size","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2602.04774","citing_title":"Theory of Optimal Learning Rate Schedules and Scaling Laws for a Random Feature Model","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7","json":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7.json","graph_json":"https://pith.science/api/pith-number/NFZGS7BDFWNR2UBDYKPXZDGOM7/graph.json","events_json":"https://pith.science/api/pith-number/NFZGS7BDFWNR2UBDYKPXZDGOM7/events.json","paper":"https://pith.science/paper/NFZGS7BD"},"agent_actions":{"view_html":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7","download_json":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7.json","view_paper":"https://pith.science/paper/NFZGS7BD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.19913&json=true","fetch_graph":"https://pith.science/api/pith-number/NFZGS7BDFWNR2UBDYKPXZDGOM7/graph.json","fetch_events":"https://pith.science/api/pith-number/NFZGS7BDFWNR2UBDYKPXZDGOM7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7/action/storage_attestation","attest_author":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7/action/author_attestation","sign_citation":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7/action/citation_signature","submit_replication":"https://pith.science/pith/NFZGS7BDFWNR2UBDYKPXZDGOM7/action/replication_record"}},"created_at":"2026-07-05T10:18:42.086643+00:00","updated_at":"2026-07-05T10:18:42.086643+00:00"}