{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BCWLLW7TFK4AJS5XXM5OE2LXQR","short_pith_number":"pith:BCWLLW7T","schema_version":"1.0","canonical_sha256":"08acb5dbf32ab804cbb7bb3ae269778452d04cd8aeabc340845403bfaf2d41ca","source":{"kind":"arxiv","id":"2404.17790","version":1},"attestation_state":"computed","paper":{"title":"Continual Pre-Training for Cross-Lingual LLM Adaptation: Enhancing Japanese Language Capabilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Hirai Shota, Hiroki Iida, Kakeru Hattori, Kazuki Fujii, Masanari Ohi, Mengsay Loem, Naoaki Okazaki, Rio Yokota, Sakae Mizuki, Taishi Nakamura","submitted_at":"2024-04-27T06:07:55Z","abstract_excerpt":"Cross-lingual continual pre-training of large language models (LLMs) initially trained on English corpus allows us to leverage the vast amount of English language resources and reduce the pre-training cost. In this study, we constructed Swallow, an LLM with enhanced Japanese capability, by extending the vocabulary of Llama 2 to include Japanese characters and conducting continual pre-training on a large Japanese web corpus. Experimental results confirmed that the performance on Japanese tasks drastically improved through continual pre-training, and the performance monotonically increased with "},"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":"2404.17790","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-27T06:07:55Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"600e1964d629da32441c1a2aa631ab3a49dc6dc3eac92044d81adf610ffb1dac","abstract_canon_sha256":"e3dc41edaa013c99e82fcf04d6f63fa3d71127954164f6cf7cce348d148472c3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:12:54.220356Z","signature_b64":"u394pHorc/rgPEHZcQNK54/p50rwMePlhwn3eRnWxo1ZqIhzd/SdRd0HZExYA4aw4Ql2zTulpjwXcgokXCVoDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"08acb5dbf32ab804cbb7bb3ae269778452d04cd8aeabc340845403bfaf2d41ca","last_reissued_at":"2026-07-05T08:12:54.219866Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:12:54.219866Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Continual Pre-Training for Cross-Lingual LLM Adaptation: Enhancing Japanese Language Capabilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Hirai Shota, Hiroki Iida, Kakeru Hattori, Kazuki Fujii, Masanari Ohi, Mengsay Loem, Naoaki Okazaki, Rio Yokota, Sakae Mizuki, Taishi Nakamura","submitted_at":"2024-04-27T06:07:55Z","abstract_excerpt":"Cross-lingual continual pre-training of large language models (LLMs) initially trained on English corpus allows us to leverage the vast amount of English language resources and reduce the pre-training cost. In this study, we constructed Swallow, an LLM with enhanced Japanese capability, by extending the vocabulary of Llama 2 to include Japanese characters and conducting continual pre-training on a large Japanese web corpus. Experimental results confirmed that the performance on Japanese tasks drastically improved through continual pre-training, and the performance monotonically increased with "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.17790","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/2404.17790/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":"2404.17790","created_at":"2026-07-05T08:12:54.219924+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.17790v1","created_at":"2026-07-05T08:12:54.219924+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.17790","created_at":"2026-07-05T08:12:54.219924+00:00"},{"alias_kind":"pith_short_12","alias_value":"BCWLLW7TFK4A","created_at":"2026-07-05T08:12:54.219924+00:00"},{"alias_kind":"pith_short_16","alias_value":"BCWLLW7TFK4AJS5X","created_at":"2026-07-05T08:12:54.219924+00:00"},{"alias_kind":"pith_short_8","alias_value":"BCWLLW7T","created_at":"2026-07-05T08:12:54.219924+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07422","citing_title":"The Masked Advantage: Uncovering Local-Language Access to Cultural Knowledge in LLMs","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11255","citing_title":"HEBATRON: A Hebrew-Specialized Open-Weight Mixture-of-Experts Language Model","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15732","citing_title":"Accuracy Is Speed: Towards Long-Context-Aware Routing for Distributed LLM Serving","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25486","citing_title":"ReTokSync: Self-Synchronizing Tokenization Disambiguation for Generative Linguistic Steganography","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR","json":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR.json","graph_json":"https://pith.science/api/pith-number/BCWLLW7TFK4AJS5XXM5OE2LXQR/graph.json","events_json":"https://pith.science/api/pith-number/BCWLLW7TFK4AJS5XXM5OE2LXQR/events.json","paper":"https://pith.science/paper/BCWLLW7T"},"agent_actions":{"view_html":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR","download_json":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR.json","view_paper":"https://pith.science/paper/BCWLLW7T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.17790&json=true","fetch_graph":"https://pith.science/api/pith-number/BCWLLW7TFK4AJS5XXM5OE2LXQR/graph.json","fetch_events":"https://pith.science/api/pith-number/BCWLLW7TFK4AJS5XXM5OE2LXQR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR/action/storage_attestation","attest_author":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR/action/author_attestation","sign_citation":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR/action/citation_signature","submit_replication":"https://pith.science/pith/BCWLLW7TFK4AJS5XXM5OE2LXQR/action/replication_record"}},"created_at":"2026-07-05T08:12:54.219924+00:00","updated_at":"2026-07-05T08:12:54.219924+00:00"}