{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WM4B4WWBN6NKFDFYJTMDCEZXLK","short_pith_number":"pith:WM4B4WWB","schema_version":"1.0","canonical_sha256":"b3381e5ac16f9aa28cb84cd83113375ab715969d6d9be532e20d7c9678960801","source":{"kind":"arxiv","id":"2411.12986","version":2},"attestation_state":"computed","paper":{"title":"Training Bilingual LMs with Data Constraints in the Targeted Language","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"David Grangier, Maartje ter Hoeve, Natalie Schluter, Richard He Bai, Skyler Seto","submitted_at":"2024-11-20T02:27:40Z","abstract_excerpt":"Large language models are trained on massive scrapes of the web, as required by current scaling laws. Most progress is made for English, given its abundance of high-quality pretraining data. For most other languages, however, such high quality pretraining data is unavailable. In this work, we study how to boost pretrained model performance in a target language with insufficient pretraining data for training a high performing language model, by enlisting data from an auxiliary language for which high quality data is available. We study this by quantifying the performance gap between training wi"},"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":"2411.12986","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-11-20T02:27:40Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a49cd33cbc76f777a0d42a1997feca230c31ba5fe0d927b5f97048f64c4fa183","abstract_canon_sha256":"a97426accf99f08c6914fdcab8a30122e30fc39392f0e3c462ba875142ac083f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:10:09.815065Z","signature_b64":"loBh4EjSIUjiSDs29ZeOlRQhOsQPEdbeUvxh1HJWE33pw/QJQvXBo6udOxT5afZXHBjrh6GNioBSKj2upuMhDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b3381e5ac16f9aa28cb84cd83113375ab715969d6d9be532e20d7c9678960801","last_reissued_at":"2026-07-05T10:10:09.814509Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:10:09.814509Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training Bilingual LMs with Data Constraints in the Targeted Language","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"David Grangier, Maartje ter Hoeve, Natalie Schluter, Richard He Bai, Skyler Seto","submitted_at":"2024-11-20T02:27:40Z","abstract_excerpt":"Large language models are trained on massive scrapes of the web, as required by current scaling laws. Most progress is made for English, given its abundance of high-quality pretraining data. For most other languages, however, such high quality pretraining data is unavailable. In this work, we study how to boost pretrained model performance in a target language with insufficient pretraining data for training a high performing language model, by enlisting data from an auxiliary language for which high quality data is available. We study this by quantifying the performance gap between training wi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.12986","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/2411.12986/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":"2411.12986","created_at":"2026-07-05T10:10:09.814581+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.12986v2","created_at":"2026-07-05T10:10:09.814581+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.12986","created_at":"2026-07-05T10:10:09.814581+00:00"},{"alias_kind":"pith_short_12","alias_value":"WM4B4WWBN6NK","created_at":"2026-07-05T10:10:09.814581+00:00"},{"alias_kind":"pith_short_16","alias_value":"WM4B4WWBN6NKFDFY","created_at":"2026-07-05T10:10:09.814581+00:00"},{"alias_kind":"pith_short_8","alias_value":"WM4B4WWB","created_at":"2026-07-05T10:10:09.814581+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.13225","citing_title":"Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK","json":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK.json","graph_json":"https://pith.science/api/pith-number/WM4B4WWBN6NKFDFYJTMDCEZXLK/graph.json","events_json":"https://pith.science/api/pith-number/WM4B4WWBN6NKFDFYJTMDCEZXLK/events.json","paper":"https://pith.science/paper/WM4B4WWB"},"agent_actions":{"view_html":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK","download_json":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK.json","view_paper":"https://pith.science/paper/WM4B4WWB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.12986&json=true","fetch_graph":"https://pith.science/api/pith-number/WM4B4WWBN6NKFDFYJTMDCEZXLK/graph.json","fetch_events":"https://pith.science/api/pith-number/WM4B4WWBN6NKFDFYJTMDCEZXLK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK/action/storage_attestation","attest_author":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK/action/author_attestation","sign_citation":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK/action/citation_signature","submit_replication":"https://pith.science/pith/WM4B4WWBN6NKFDFYJTMDCEZXLK/action/replication_record"}},"created_at":"2026-07-05T10:10:09.814581+00:00","updated_at":"2026-07-05T10:10:09.814581+00:00"}