{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4WJ5AIG3BWWAHI2HCE2BQ7ZBVV","short_pith_number":"pith:4WJ5AIG3","schema_version":"1.0","canonical_sha256":"e593d020db0dac03a3471134187f21ad620eae03488d5db0fd24c4db2ccafcd4","source":{"kind":"arxiv","id":"2506.21570","version":1},"attestation_state":"computed","paper":{"title":"Random Initialization Can't Catch Up: The Advantage of Language Model Transfer for Time Series Forecasting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Alexis Roger, Andrew R. Williams, Arjun Ashok, Irina Rish, Kashif Rasul, Prateek Humane, Roland Riachi, Yuriy Nevmyvaka","submitted_at":"2025-06-12T18:39:38Z","abstract_excerpt":"Recent works have demonstrated the effectiveness of adapting pre-trained language models (LMs) for forecasting time series in the low-data regime. We build upon these findings by analyzing the effective transfer from language models to time series forecasting under various design choices including upstream post-training, time series tokenizer and language backbone size. In the low-data regime, these design choices have a significant impact on the validation loss, with clear-cut choices that outperform others. Contrary to Hernandez et al. (2021), we observe that the validation loss of the LMs c"},"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":"2506.21570","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-12T18:39:38Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"54a44dbaa922b0b225dc84cccb0726ac34de290c96ab628ee7adbf7a7d82e45e","abstract_canon_sha256":"65faefcafc1814761456715250fe6057f4880e0b2fb0848934985b8ea6a77172"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:27:56.570029Z","signature_b64":"LN0wkdXNZxyOH3n83TekIjvqonY5YWIuxNfTfZ+nYr5Dg6VDjEEUKRfFRBpBgo2M69y3xhwnicoLHsSPXXkDAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e593d020db0dac03a3471134187f21ad620eae03488d5db0fd24c4db2ccafcd4","last_reissued_at":"2026-07-05T11:27:56.569508Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:27:56.569508Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Random Initialization Can't Catch Up: The Advantage of Language Model Transfer for Time Series Forecasting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Alexis Roger, Andrew R. Williams, Arjun Ashok, Irina Rish, Kashif Rasul, Prateek Humane, Roland Riachi, Yuriy Nevmyvaka","submitted_at":"2025-06-12T18:39:38Z","abstract_excerpt":"Recent works have demonstrated the effectiveness of adapting pre-trained language models (LMs) for forecasting time series in the low-data regime. We build upon these findings by analyzing the effective transfer from language models to time series forecasting under various design choices including upstream post-training, time series tokenizer and language backbone size. In the low-data regime, these design choices have a significant impact on the validation loss, with clear-cut choices that outperform others. Contrary to Hernandez et al. (2021), we observe that the validation loss of the LMs c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.21570","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/2506.21570/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":"2506.21570","created_at":"2026-07-05T11:27:56.569569+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.21570v1","created_at":"2026-07-05T11:27:56.569569+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.21570","created_at":"2026-07-05T11:27:56.569569+00:00"},{"alias_kind":"pith_short_12","alias_value":"4WJ5AIG3BWWA","created_at":"2026-07-05T11:27:56.569569+00:00"},{"alias_kind":"pith_short_16","alias_value":"4WJ5AIG3BWWAHI2H","created_at":"2026-07-05T11:27:56.569569+00:00"},{"alias_kind":"pith_short_8","alias_value":"4WJ5AIG3","created_at":"2026-07-05T11:27:56.569569+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/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV","json":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV.json","graph_json":"https://pith.science/api/pith-number/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/graph.json","events_json":"https://pith.science/api/pith-number/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/events.json","paper":"https://pith.science/paper/4WJ5AIG3"},"agent_actions":{"view_html":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV","download_json":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV.json","view_paper":"https://pith.science/paper/4WJ5AIG3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.21570&json=true","fetch_graph":"https://pith.science/api/pith-number/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/graph.json","fetch_events":"https://pith.science/api/pith-number/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/action/storage_attestation","attest_author":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/action/author_attestation","sign_citation":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/action/citation_signature","submit_replication":"https://pith.science/pith/4WJ5AIG3BWWAHI2HCE2BQ7ZBVV/action/replication_record"}},"created_at":"2026-07-05T11:27:56.569569+00:00","updated_at":"2026-07-05T11:27:56.569569+00:00"}