{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:IMPSEHPVVPKKESQWD3PJ76RXNS","short_pith_number":"pith:IMPSEHPV","schema_version":"1.0","canonical_sha256":"431f221df5abd4a24a161ede9ffa376cb506bd237749d45f5fe90b30faffa52e","source":{"kind":"arxiv","id":"2605.24681","version":1},"attestation_state":"computed","paper":{"title":"Mix-MoE: Improving Multilingual Machine Translation of Large Language Models through Mixed MoEs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bo Li, Deyi Xiong, Shaolin Zhu, Tianyu Dong","submitted_at":"2026-05-23T17:35:43Z","abstract_excerpt":"Large Language Models (LLMs) have shown great promise in multilingual machine translation (MT), even with limited bilingual supervision. However, fine-tuning LLMs with parallel corpora presents major challenges, namely parameter interference. To address these issues, we propose Mix-MoE, a mixed Mixture-of-Experts framework designed to train LLMs for multilingual MT. Our framework operates in two distinct stages: (1) post-pretraining with MoE on monolingual corpora, and (2) post-pretraining with MoE on parallel corpora. Crucially, we divide the MoE layers into two specialized groups: Language M"},"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":"2605.24681","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-23T17:35:43Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"5089606d5201af76b827a190272207a8747fc0acf9f9a69fa7a152cca1a09090","abstract_canon_sha256":"670defdb05abfe65c954da77933a9f02d3ee3dc3aec016e722b8cc964872b149"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:53.007674Z","signature_b64":"05KeqEIim/YP/JVa5dbduvvck+/+5L0TZsj7/3Yh3x2FypM6AoFz8tlJ+2+c0INurx58IhphezOfJg0QXAcUCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"431f221df5abd4a24a161ede9ffa376cb506bd237749d45f5fe90b30faffa52e","last_reissued_at":"2026-05-26T01:03:53.006925Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:53.006925Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mix-MoE: Improving Multilingual Machine Translation of Large Language Models through Mixed MoEs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bo Li, Deyi Xiong, Shaolin Zhu, Tianyu Dong","submitted_at":"2026-05-23T17:35:43Z","abstract_excerpt":"Large Language Models (LLMs) have shown great promise in multilingual machine translation (MT), even with limited bilingual supervision. However, fine-tuning LLMs with parallel corpora presents major challenges, namely parameter interference. To address these issues, we propose Mix-MoE, a mixed Mixture-of-Experts framework designed to train LLMs for multilingual MT. Our framework operates in two distinct stages: (1) post-pretraining with MoE on monolingual corpora, and (2) post-pretraining with MoE on parallel corpora. Crucially, we divide the MoE layers into two specialized groups: Language M"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24681","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/2605.24681/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":"2605.24681","created_at":"2026-05-26T01:03:53.007043+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.24681v1","created_at":"2026-05-26T01:03:53.007043+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24681","created_at":"2026-05-26T01:03:53.007043+00:00"},{"alias_kind":"pith_short_12","alias_value":"IMPSEHPVVPKK","created_at":"2026-05-26T01:03:53.007043+00:00"},{"alias_kind":"pith_short_16","alias_value":"IMPSEHPVVPKKESQW","created_at":"2026-05-26T01:03:53.007043+00:00"},{"alias_kind":"pith_short_8","alias_value":"IMPSEHPV","created_at":"2026-05-26T01:03:53.007043+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/IMPSEHPVVPKKESQWD3PJ76RXNS","json":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS.json","graph_json":"https://pith.science/api/pith-number/IMPSEHPVVPKKESQWD3PJ76RXNS/graph.json","events_json":"https://pith.science/api/pith-number/IMPSEHPVVPKKESQWD3PJ76RXNS/events.json","paper":"https://pith.science/paper/IMPSEHPV"},"agent_actions":{"view_html":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS","download_json":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS.json","view_paper":"https://pith.science/paper/IMPSEHPV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.24681&json=true","fetch_graph":"https://pith.science/api/pith-number/IMPSEHPVVPKKESQWD3PJ76RXNS/graph.json","fetch_events":"https://pith.science/api/pith-number/IMPSEHPVVPKKESQWD3PJ76RXNS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS/action/storage_attestation","attest_author":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS/action/author_attestation","sign_citation":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS/action/citation_signature","submit_replication":"https://pith.science/pith/IMPSEHPVVPKKESQWD3PJ76RXNS/action/replication_record"}},"created_at":"2026-05-26T01:03:53.007043+00:00","updated_at":"2026-05-26T01:03:53.007043+00:00"}