{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:V3KE6ABM7MU2WKVB6XSBV4T4ZF","short_pith_number":"pith:V3KE6ABM","schema_version":"1.0","canonical_sha256":"aed44f002cfb29ab2aa1f5e41af27cc95c377a70514ea34497ad7223ad3b79f4","source":{"kind":"arxiv","id":"2605.18083","version":1},"attestation_state":"computed","paper":{"title":"A Data-Efficient Path to Multilingual LLMs: Language Expansion via Post-training PARAM$\\Delta$ Integration into Upcycled MoE","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Baosong Yang, Hao-Ran Wei, Hao Zhou, Jiajun Chen, Linjuan Wu, Shuaijie She, Shujian Huang, Tianhao Li, Zhijun Wang","submitted_at":"2026-05-18T08:59:08Z","abstract_excerpt":"Expanding Large Language Models~(LLMs) to new languages is a costly endeavor, demanding extensive Continued Pre-Training~(CPT) and data-intensive alignment. While recent data-free merging techniques attempt to bypass alignment by fusing a multilingual CPT-enhanced model with its instruct counterpart, they are plagued by a critical trade-off: mitigating parameter conflicts to preserve original abilities inevitably dilutes new language acquisition, and vice-versa. To resolve this conflict, we introduce \\method, which upcycles a dense model into a Mixture-of-Experts~(MoE) architecture, allocating"},"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":"2605.18083","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-18T08:59:08Z","cross_cats_sorted":[],"title_canon_sha256":"890d7396d492d64ee7f0a2cc1597605b350fc8e72e0bbfe89891dd743f8461c4","abstract_canon_sha256":"9e469e9c2647a03f497d4d8b626ab89f9db2cb2f07118fcb859a085289b93344"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:15.190498Z","signature_b64":"H+KOKDVKWA928JkLhPdPeLOquttixSE4/lMrD9uSzPSRMqmF1nUvCXC3+5rUKD96zCy2u43VDX8ynzYRni5QBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aed44f002cfb29ab2aa1f5e41af27cc95c377a70514ea34497ad7223ad3b79f4","last_reissued_at":"2026-05-20T00:05:15.189583Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:15.189583Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Data-Efficient Path to Multilingual LLMs: Language Expansion via Post-training PARAM$\\Delta$ Integration into Upcycled MoE","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Baosong Yang, Hao-Ran Wei, Hao Zhou, Jiajun Chen, Linjuan Wu, Shuaijie She, Shujian Huang, Tianhao Li, Zhijun Wang","submitted_at":"2026-05-18T08:59:08Z","abstract_excerpt":"Expanding Large Language Models~(LLMs) to new languages is a costly endeavor, demanding extensive Continued Pre-Training~(CPT) and data-intensive alignment. While recent data-free merging techniques attempt to bypass alignment by fusing a multilingual CPT-enhanced model with its instruct counterpart, they are plagued by a critical trade-off: mitigating parameter conflicts to preserve original abilities inevitably dilutes new language acquisition, and vice-versa. To resolve this conflict, we introduce \\method, which upcycles a dense model into a Mixture-of-Experts~(MoE) architecture, allocating"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18083","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/2605.18083/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-19T23:41:59.222701Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T23:33:35.449389Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"dbe1af9199807909ac8aa1bd8786b18df9ea03f5412ba6f4c11905915638a60f"},"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":"2605.18083","created_at":"2026-05-20T00:05:15.189740+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.18083v1","created_at":"2026-05-20T00:05:15.189740+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18083","created_at":"2026-05-20T00:05:15.189740+00:00"},{"alias_kind":"pith_short_12","alias_value":"V3KE6ABM7MU2","created_at":"2026-05-20T00:05:15.189740+00:00"},{"alias_kind":"pith_short_16","alias_value":"V3KE6ABM7MU2WKVB","created_at":"2026-05-20T00:05:15.189740+00:00"},{"alias_kind":"pith_short_8","alias_value":"V3KE6ABM","created_at":"2026-05-20T00:05:15.189740+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF","json":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF.json","graph_json":"https://pith.science/api/pith-number/V3KE6ABM7MU2WKVB6XSBV4T4ZF/graph.json","events_json":"https://pith.science/api/pith-number/V3KE6ABM7MU2WKVB6XSBV4T4ZF/events.json","paper":"https://pith.science/paper/V3KE6ABM"},"agent_actions":{"view_html":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF","download_json":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF.json","view_paper":"https://pith.science/paper/V3KE6ABM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.18083&json=true","fetch_graph":"https://pith.science/api/pith-number/V3KE6ABM7MU2WKVB6XSBV4T4ZF/graph.json","fetch_events":"https://pith.science/api/pith-number/V3KE6ABM7MU2WKVB6XSBV4T4ZF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF/action/storage_attestation","attest_author":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF/action/author_attestation","sign_citation":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF/action/citation_signature","submit_replication":"https://pith.science/pith/V3KE6ABM7MU2WKVB6XSBV4T4ZF/action/replication_record"}},"created_at":"2026-05-20T00:05:15.189740+00:00","updated_at":"2026-05-20T00:05:15.189740+00:00"}