{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XPEQR4F5ZI6LXFRWNNKK47VW32","short_pith_number":"pith:XPEQR4F5","schema_version":"1.0","canonical_sha256":"bbc908f0bdca3cbb96366b54ae7eb6de90eddc9dfdfdc9c49e0fa15cf8e16346","source":{"kind":"arxiv","id":"2410.12883","version":2},"attestation_state":"computed","paper":{"title":"Scaling Laws for Multilingual Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Alon Benhaim, Barun Patra, Han Zhao, Parul Chopra, Praneetha Vaddamanu, Sanchit Ahuja, Vishrav Chaudhary, Xia Song, Yifei He","submitted_at":"2024-10-15T20:29:38Z","abstract_excerpt":"We propose a novel scaling law for general-purpose decoder-only language models (LMs) trained on multilingual data, tackling the problem of balancing languages during multilingual pretraining. A primary challenge in studying multilingual scaling is the difficulty of analyzing individual language performance due to cross-lingual transfer. To address this, we shift the focus from individual languages to language families. We introduce and validate a hypothesis that the test cross-entropy loss for each language family is determined solely by its own sampling ratio, independent of other languages "},"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":"2410.12883","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-15T20:29:38Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"bb93e407977b58be9301f0166f828c8bc8c85fa95a0f0fdfc8798d8bff9087d1","abstract_canon_sha256":"ced7ada17b545eadb50253a253ef920c82394c77f35a083e1bbdd068392de60c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:44:15.619418Z","signature_b64":"zSr88iQhDL1H+pEAOH3Dh6HkO7+vE0//OV/mFooiQdDTYNmwYrCrpei8WmmhR4MtTf3k2NNeiQMuEXrSMWTqBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bbc908f0bdca3cbb96366b54ae7eb6de90eddc9dfdfdc9c49e0fa15cf8e16346","last_reissued_at":"2026-07-05T09:44:15.618931Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:44:15.618931Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Laws for Multilingual Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Alon Benhaim, Barun Patra, Han Zhao, Parul Chopra, Praneetha Vaddamanu, Sanchit Ahuja, Vishrav Chaudhary, Xia Song, Yifei He","submitted_at":"2024-10-15T20:29:38Z","abstract_excerpt":"We propose a novel scaling law for general-purpose decoder-only language models (LMs) trained on multilingual data, tackling the problem of balancing languages during multilingual pretraining. A primary challenge in studying multilingual scaling is the difficulty of analyzing individual language performance due to cross-lingual transfer. To address this, we shift the focus from individual languages to language families. We introduce and validate a hypothesis that the test cross-entropy loss for each language family is determined solely by its own sampling ratio, independent of other languages "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.12883","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/2410.12883/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":"2410.12883","created_at":"2026-07-05T09:44:15.618989+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.12883v2","created_at":"2026-07-05T09:44:15.618989+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.12883","created_at":"2026-07-05T09:44:15.618989+00:00"},{"alias_kind":"pith_short_12","alias_value":"XPEQR4F5ZI6L","created_at":"2026-07-05T09:44:15.618989+00:00"},{"alias_kind":"pith_short_16","alias_value":"XPEQR4F5ZI6LXFRW","created_at":"2026-07-05T09:44:15.618989+00:00"},{"alias_kind":"pith_short_8","alias_value":"XPEQR4F5","created_at":"2026-07-05T09:44:15.618989+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"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":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02881","citing_title":"One Model to Translate Them All? A Journey to Mount Doom for Multilingual Model Merging","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32","json":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32.json","graph_json":"https://pith.science/api/pith-number/XPEQR4F5ZI6LXFRWNNKK47VW32/graph.json","events_json":"https://pith.science/api/pith-number/XPEQR4F5ZI6LXFRWNNKK47VW32/events.json","paper":"https://pith.science/paper/XPEQR4F5"},"agent_actions":{"view_html":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32","download_json":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32.json","view_paper":"https://pith.science/paper/XPEQR4F5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.12883&json=true","fetch_graph":"https://pith.science/api/pith-number/XPEQR4F5ZI6LXFRWNNKK47VW32/graph.json","fetch_events":"https://pith.science/api/pith-number/XPEQR4F5ZI6LXFRWNNKK47VW32/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32/action/storage_attestation","attest_author":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32/action/author_attestation","sign_citation":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32/action/citation_signature","submit_replication":"https://pith.science/pith/XPEQR4F5ZI6LXFRWNNKK47VW32/action/replication_record"}},"created_at":"2026-07-05T09:44:15.618989+00:00","updated_at":"2026-07-05T09:44:15.618989+00:00"}