{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BY5D5WVDTAYPXXVFUFA45Y27QT","short_pith_number":"pith:BY5D5WVD","schema_version":"1.0","canonical_sha256":"0e3a3edaa39830fbdea5a141cee35f84e483cfdb4a12a10e2d9635d6136327b4","source":{"kind":"arxiv","id":"2504.01801","version":2},"attestation_state":"computed","paper":{"title":"Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chao Deng, Hao Zhou, Jiahuan Li, Jingang Wang, Junlan Feng, Rongxiang Weng, Shujian Huang, Xin Huang, Xue Han, Zhijun Wang","submitted_at":"2025-04-02T15:09:58Z","abstract_excerpt":"Large language models (LLMs) exhibit remarkable multilingual capabilities despite the extreme language imbalance in the pre-training data. In this paper, we closely examine the reasons behind this phenomenon, focusing on the pre-training corpus. We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities. We conduct an analysis to investigate code-switching in the pre-training corpus, examining its presence and categorizing it into four types within two quadrants. We then assess its impact on multilingual performan"},"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":"2504.01801","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-02T15:09:58Z","cross_cats_sorted":[],"title_canon_sha256":"9efaf846901979f4f85485ea3a716840687766f6356c7f8b985674d71febff3f","abstract_canon_sha256":"cb9aef835e1f443b62e25ae842e5ebbf15a57b6a8ee60239496972180e1f3b5b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:52:21.693400Z","signature_b64":"kkixvAwnW+HJ+dqraFxNS7rOttRTQPlkytvG3AcNAPTK+T0McEhlaulvV07Ztm/i29oQlLl3ZWvP5EKJuQIrCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0e3a3edaa39830fbdea5a141cee35f84e483cfdb4a12a10e2d9635d6136327b4","last_reissued_at":"2026-07-05T10:52:21.692925Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:52:21.692925Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chao Deng, Hao Zhou, Jiahuan Li, Jingang Wang, Junlan Feng, Rongxiang Weng, Shujian Huang, Xin Huang, Xue Han, Zhijun Wang","submitted_at":"2025-04-02T15:09:58Z","abstract_excerpt":"Large language models (LLMs) exhibit remarkable multilingual capabilities despite the extreme language imbalance in the pre-training data. In this paper, we closely examine the reasons behind this phenomenon, focusing on the pre-training corpus. We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities. We conduct an analysis to investigate code-switching in the pre-training corpus, examining its presence and categorizing it into four types within two quadrants. We then assess its impact on multilingual performan"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.01801","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/2504.01801/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":"2504.01801","created_at":"2026-07-05T10:52:21.692994+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.01801v2","created_at":"2026-07-05T10:52:21.692994+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.01801","created_at":"2026-07-05T10:52:21.692994+00:00"},{"alias_kind":"pith_short_12","alias_value":"BY5D5WVDTAYP","created_at":"2026-07-05T10:52:21.692994+00:00"},{"alias_kind":"pith_short_16","alias_value":"BY5D5WVDTAYPXXVF","created_at":"2026-07-05T10:52:21.692994+00:00"},{"alias_kind":"pith_short_8","alias_value":"BY5D5WVD","created_at":"2026-07-05T10:52:21.692994+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.19116","citing_title":"Controlling Language Confusion in Multilingual LLMs","ref_index":23,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT","json":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT.json","graph_json":"https://pith.science/api/pith-number/BY5D5WVDTAYPXXVFUFA45Y27QT/graph.json","events_json":"https://pith.science/api/pith-number/BY5D5WVDTAYPXXVFUFA45Y27QT/events.json","paper":"https://pith.science/paper/BY5D5WVD"},"agent_actions":{"view_html":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT","download_json":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT.json","view_paper":"https://pith.science/paper/BY5D5WVD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.01801&json=true","fetch_graph":"https://pith.science/api/pith-number/BY5D5WVDTAYPXXVFUFA45Y27QT/graph.json","fetch_events":"https://pith.science/api/pith-number/BY5D5WVDTAYPXXVFUFA45Y27QT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT/action/storage_attestation","attest_author":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT/action/author_attestation","sign_citation":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT/action/citation_signature","submit_replication":"https://pith.science/pith/BY5D5WVDTAYPXXVFUFA45Y27QT/action/replication_record"}},"created_at":"2026-07-05T10:52:21.692994+00:00","updated_at":"2026-07-05T10:52:21.692994+00:00"}