{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PVM4J7MLWXZLZ7PM5FOR3LAQSD","short_pith_number":"pith:PVM4J7ML","schema_version":"1.0","canonical_sha256":"7d59c4fd8bb5f2bcfdece95d1dac1090e301aa3e5a06fb9eb0d188a9d9899e69","source":{"kind":"arxiv","id":"2501.05122","version":1},"attestation_state":"computed","paper":{"title":"Centurio: On Drivers of Multilingual Ability of Large Vision-Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Anne Lauscher, Carolin Holtermann, Chris Biemann, Florian Schneider, Goran Glava\\v{s}, Gregor Geigle, Radu Timofte","submitted_at":"2025-01-09T10:26:14Z","abstract_excerpt":"Most Large Vision-Language Models (LVLMs) to date are trained predominantly on English data, which makes them struggle to understand non-English input and fail to generate output in the desired target language. Existing efforts mitigate these issues by adding multilingual training data, but do so in a largely ad-hoc manner, lacking insight into how different training mixes tip the scale for different groups of languages. In this work, we present a comprehensive investigation into the training strategies for massively multilingual LVLMs. First, we conduct a series of multi-stage experiments spa"},"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":"2501.05122","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-09T10:26:14Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"71b18e38f195627809a2029ad0a1381875bec8c2525be01fac8304956de66aef","abstract_canon_sha256":"0ce010e385fe97f7aa95535bcad0b96fbed59ea6c9bfdf1dc33fa0fd8316c74e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:59:04.746806Z","signature_b64":"D4VssX/Q7FIBR9wFwRkodPS3bTwo/cxD3Fu71jb8lp9IBCQLK2HqINIJmD0NIWPRexqPuBzJIpN/P44PXDX4Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7d59c4fd8bb5f2bcfdece95d1dac1090e301aa3e5a06fb9eb0d188a9d9899e69","last_reissued_at":"2026-07-05T09:59:04.746392Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:59:04.746392Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Centurio: On Drivers of Multilingual Ability of Large Vision-Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Anne Lauscher, Carolin Holtermann, Chris Biemann, Florian Schneider, Goran Glava\\v{s}, Gregor Geigle, Radu Timofte","submitted_at":"2025-01-09T10:26:14Z","abstract_excerpt":"Most Large Vision-Language Models (LVLMs) to date are trained predominantly on English data, which makes them struggle to understand non-English input and fail to generate output in the desired target language. Existing efforts mitigate these issues by adding multilingual training data, but do so in a largely ad-hoc manner, lacking insight into how different training mixes tip the scale for different groups of languages. In this work, we present a comprehensive investigation into the training strategies for massively multilingual LVLMs. First, we conduct a series of multi-stage experiments spa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.05122","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/2501.05122/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":"2501.05122","created_at":"2026-07-05T09:59:04.746448+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.05122v1","created_at":"2026-07-05T09:59:04.746448+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.05122","created_at":"2026-07-05T09:59:04.746448+00:00"},{"alias_kind":"pith_short_12","alias_value":"PVM4J7MLWXZL","created_at":"2026-07-05T09:59:04.746448+00:00"},{"alias_kind":"pith_short_16","alias_value":"PVM4J7MLWXZLZ7PM","created_at":"2026-07-05T09:59:04.746448+00:00"},{"alias_kind":"pith_short_8","alias_value":"PVM4J7ML","created_at":"2026-07-05T09:59:04.746448+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.22123","citing_title":"Multilingual Vision-Language Models, A Survey","ref_index":54,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD","json":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD.json","graph_json":"https://pith.science/api/pith-number/PVM4J7MLWXZLZ7PM5FOR3LAQSD/graph.json","events_json":"https://pith.science/api/pith-number/PVM4J7MLWXZLZ7PM5FOR3LAQSD/events.json","paper":"https://pith.science/paper/PVM4J7ML"},"agent_actions":{"view_html":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD","download_json":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD.json","view_paper":"https://pith.science/paper/PVM4J7ML","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.05122&json=true","fetch_graph":"https://pith.science/api/pith-number/PVM4J7MLWXZLZ7PM5FOR3LAQSD/graph.json","fetch_events":"https://pith.science/api/pith-number/PVM4J7MLWXZLZ7PM5FOR3LAQSD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD/action/storage_attestation","attest_author":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD/action/author_attestation","sign_citation":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD/action/citation_signature","submit_replication":"https://pith.science/pith/PVM4J7MLWXZLZ7PM5FOR3LAQSD/action/replication_record"}},"created_at":"2026-07-05T09:59:04.746448+00:00","updated_at":"2026-07-05T09:59:04.746448+00:00"}