{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ETUQABQPPCYAFG6SZGRJS5GND6","short_pith_number":"pith:ETUQABQP","schema_version":"1.0","canonical_sha256":"24e900060f78b0029bd2c9a29974cd1fb82947970b42ee9febfeadd0a83ed499","source":{"kind":"arxiv","id":"2408.11029","version":2},"attestation_state":"computed","paper":{"title":"Scaling Law with Learning Rate Annealing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Howe Tissue, Lu Wang, Venus Wang","submitted_at":"2024-08-20T17:30:48Z","abstract_excerpt":"We find that the cross-entropy loss curves of neural language models empirically adhere to a scaling law with learning rate (LR) annealing over training steps: $$L(s) = L_0 + A\\cdot S_1^{-\\alpha} - C\\cdot S_2,$$ where $L(s)$ is the validation loss at step $s$, $S_1$ is the area under the LR curve, $S_2$ is the LR annealing area, and $L_0$, $A$, $C$, $\\alpha$ are constant parameters. This formulation takes into account two factors: (1) power-law scaling over data size, and (2) the additional loss reduction during LR annealing. Therefore, this formulation can describe the full loss curve at each"},"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":"2408.11029","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-08-20T17:30:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"0785522b6b06b36dfef8c2aa0e4cbb336c6c4a3c19744f38bf5efdd76084051f","abstract_canon_sha256":"d056d2b87b9bdc7a6127f6b0e48bbdb19a743c00d2cd10f709e16ac2ae19b9c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:25:46.566721Z","signature_b64":"xLc6eoH5cV3hwrE49ze5ZbS6NsMBjaDiwDnxjuskGAj/mTeyxlLr3uuoF7v1lXofzbWh0tCyQuPfUGt2EtadCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"24e900060f78b0029bd2c9a29974cd1fb82947970b42ee9febfeadd0a83ed499","last_reissued_at":"2026-07-05T09:25:46.566202Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:25:46.566202Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Law with Learning Rate Annealing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Howe Tissue, Lu Wang, Venus Wang","submitted_at":"2024-08-20T17:30:48Z","abstract_excerpt":"We find that the cross-entropy loss curves of neural language models empirically adhere to a scaling law with learning rate (LR) annealing over training steps: $$L(s) = L_0 + A\\cdot S_1^{-\\alpha} - C\\cdot S_2,$$ where $L(s)$ is the validation loss at step $s$, $S_1$ is the area under the LR curve, $S_2$ is the LR annealing area, and $L_0$, $A$, $C$, $\\alpha$ are constant parameters. This formulation takes into account two factors: (1) power-law scaling over data size, and (2) the additional loss reduction during LR annealing. Therefore, this formulation can describe the full loss curve at each"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.11029","kind":"arxiv","version":2},"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/2408.11029/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":"2408.11029","created_at":"2026-07-05T09:25:46.566260+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.11029v2","created_at":"2026-07-05T09:25:46.566260+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.11029","created_at":"2026-07-05T09:25:46.566260+00:00"},{"alias_kind":"pith_short_12","alias_value":"ETUQABQPPCYA","created_at":"2026-07-05T09:25:46.566260+00:00"},{"alias_kind":"pith_short_16","alias_value":"ETUQABQPPCYAFG6S","created_at":"2026-07-05T09:25:46.566260+00:00"},{"alias_kind":"pith_short_8","alias_value":"ETUQABQP","created_at":"2026-07-05T09:25:46.566260+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.29158","citing_title":"On the Nonlinearity of Learning Rate Scaling for LLM Training","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6","json":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6.json","graph_json":"https://pith.science/api/pith-number/ETUQABQPPCYAFG6SZGRJS5GND6/graph.json","events_json":"https://pith.science/api/pith-number/ETUQABQPPCYAFG6SZGRJS5GND6/events.json","paper":"https://pith.science/paper/ETUQABQP"},"agent_actions":{"view_html":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6","download_json":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6.json","view_paper":"https://pith.science/paper/ETUQABQP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.11029&json=true","fetch_graph":"https://pith.science/api/pith-number/ETUQABQPPCYAFG6SZGRJS5GND6/graph.json","fetch_events":"https://pith.science/api/pith-number/ETUQABQPPCYAFG6SZGRJS5GND6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6/action/storage_attestation","attest_author":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6/action/author_attestation","sign_citation":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6/action/citation_signature","submit_replication":"https://pith.science/pith/ETUQABQPPCYAFG6SZGRJS5GND6/action/replication_record"}},"created_at":"2026-07-05T09:25:46.566260+00:00","updated_at":"2026-07-05T09:25:46.566260+00:00"}