{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TOEZ7OURKGZE22DC4ABKSOJJY4","short_pith_number":"pith:TOEZ7OUR","schema_version":"1.0","canonical_sha256":"9b899fba9151b24d6862e002a93929c7248208252343b6c307e44ea008f86058","source":{"kind":"arxiv","id":"2504.15431","version":1},"attestation_state":"computed","paper":{"title":"Trillion 7B Technical Report","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hyungguk Kim, Jamin Shin (Trillion Labs), Juyoung Suk, Kyuseok Kim, Seungtaek Choi, Sungjun Han, Suyeong An, Wonsuk Yang","submitted_at":"2025-04-21T20:54:44Z","abstract_excerpt":"We introduce Trillion-7B, the most token-efficient Korean-centric multilingual LLM available. Our novel Cross-lingual Document Attention (XLDA) mechanism enables highly efficient and effective knowledge transfer from English to target languages like Korean and Japanese. Combined with optimized data mixtures, language-specific filtering, and tailored tokenizer construction, Trillion-7B achieves competitive performance while dedicating only 10\\% of its 2T training tokens to multilingual data and requiring just 59.4K H100 GPU hours (\\$148K) for full training. Comprehensive evaluations across 27 b"},"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":"2504.15431","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-21T20:54:44Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"5adc6968214305a465d38686485940c920c673cfc1b396f3c2deb91ebbb3e0f7","abstract_canon_sha256":"3dccae00d961fe4cf6f24a6f8736e9a0d0acbf878784f7b1e291faf4477121d1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:52:24.505800Z","signature_b64":"t1yM/jy6vPUHy0z188l8o1A04c3X2bXjBINt5JAPFRyl4p6lqAnKsPkNLsFH8hz3hdIymRqSxLYPVfgGm6xVCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9b899fba9151b24d6862e002a93929c7248208252343b6c307e44ea008f86058","last_reissued_at":"2026-07-05T10:52:24.505269Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:52:24.505269Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Trillion 7B Technical Report","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Hyungguk Kim, Jamin Shin (Trillion Labs), Juyoung Suk, Kyuseok Kim, Seungtaek Choi, Sungjun Han, Suyeong An, Wonsuk Yang","submitted_at":"2025-04-21T20:54:44Z","abstract_excerpt":"We introduce Trillion-7B, the most token-efficient Korean-centric multilingual LLM available. Our novel Cross-lingual Document Attention (XLDA) mechanism enables highly efficient and effective knowledge transfer from English to target languages like Korean and Japanese. Combined with optimized data mixtures, language-specific filtering, and tailored tokenizer construction, Trillion-7B achieves competitive performance while dedicating only 10\\% of its 2T training tokens to multilingual data and requiring just 59.4K H100 GPU hours (\\$148K) for full training. Comprehensive evaluations across 27 b"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.15431","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/2504.15431/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":"2504.15431","created_at":"2026-07-05T10:52:24.505347+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.15431v1","created_at":"2026-07-05T10:52:24.505347+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.15431","created_at":"2026-07-05T10:52:24.505347+00:00"},{"alias_kind":"pith_short_12","alias_value":"TOEZ7OURKGZE","created_at":"2026-07-05T10:52:24.505347+00:00"},{"alias_kind":"pith_short_16","alias_value":"TOEZ7OURKGZE22DC","created_at":"2026-07-05T10:52:24.505347+00:00"},{"alias_kind":"pith_short_8","alias_value":"TOEZ7OUR","created_at":"2026-07-05T10:52:24.505347+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.16235","citing_title":"Optimizing Korean-Centric LLMs via Token Pruning","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4","json":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4.json","graph_json":"https://pith.science/api/pith-number/TOEZ7OURKGZE22DC4ABKSOJJY4/graph.json","events_json":"https://pith.science/api/pith-number/TOEZ7OURKGZE22DC4ABKSOJJY4/events.json","paper":"https://pith.science/paper/TOEZ7OUR"},"agent_actions":{"view_html":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4","download_json":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4.json","view_paper":"https://pith.science/paper/TOEZ7OUR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.15431&json=true","fetch_graph":"https://pith.science/api/pith-number/TOEZ7OURKGZE22DC4ABKSOJJY4/graph.json","fetch_events":"https://pith.science/api/pith-number/TOEZ7OURKGZE22DC4ABKSOJJY4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4/action/storage_attestation","attest_author":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4/action/author_attestation","sign_citation":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4/action/citation_signature","submit_replication":"https://pith.science/pith/TOEZ7OURKGZE22DC4ABKSOJJY4/action/replication_record"}},"created_at":"2026-07-05T10:52:24.505347+00:00","updated_at":"2026-07-05T10:52:24.505347+00:00"}