{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QBKFGKTQHONZHJRQADQUBRM2YW","short_pith_number":"pith:QBKFGKTQ","schema_version":"1.0","canonical_sha256":"8054532a703b9b93a63000e140c59ac5b0c077b44685cac3b3a863f6e9230b4d","source":{"kind":"arxiv","id":"2406.02539","version":3},"attestation_state":"computed","paper":{"title":"Parrot: Multilingual Visual Instruction Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Chao Yi, Da-Wei Zhou, De-Chuan Zhan, Hai-Long Sun, Han-Jia Ye, Kaifu Zhang, Qing-Guo Chen, Shiyin Lu, Weihua Luo, Yang Li, Zhao Xu","submitted_at":"2024-06-04T17:56:28Z","abstract_excerpt":"The rapid development of Multimodal Large Language Models (MLLMs), such as GPT-4o, marks a significant step toward artificial general intelligence. Existing methods typically align vision encoders with LLMs via supervised fine-tuning (SFT), but this often deteriorates their ability to handle multiple languages as training progresses. We empirically observe that imbalanced SFT datasets, largely English-centric, degrade performance on non-English languages due to the failure in multilingual token alignment. To address this, we propose PARROT, a novel approach that leverages textual guidance for "},"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":"2406.02539","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-06-04T17:56:28Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"title_canon_sha256":"f3e85675bbc0dc2444aeca8525cadac008662c7f282a1dbf2385e06eae280263","abstract_canon_sha256":"7878edc82b271df0b1481973062c75800115db5c78a43553229da13c6b7a3d5c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:57.979493Z","signature_b64":"vpBiryX8XisImY9ZNJe5+oi/cuEUJ41GdPD0WWW/Yqfx4igz+GRs+G9aoGASMK+9ssfgF2bSf5Ef96Qv/sXFCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8054532a703b9b93a63000e140c59ac5b0c077b44685cac3b3a863f6e9230b4d","last_reissued_at":"2026-07-05T11:08:57.978811Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:57.978811Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Parrot: Multilingual Visual Instruction Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Chao Yi, Da-Wei Zhou, De-Chuan Zhan, Hai-Long Sun, Han-Jia Ye, Kaifu Zhang, Qing-Guo Chen, Shiyin Lu, Weihua Luo, Yang Li, Zhao Xu","submitted_at":"2024-06-04T17:56:28Z","abstract_excerpt":"The rapid development of Multimodal Large Language Models (MLLMs), such as GPT-4o, marks a significant step toward artificial general intelligence. Existing methods typically align vision encoders with LLMs via supervised fine-tuning (SFT), but this often deteriorates their ability to handle multiple languages as training progresses. We empirically observe that imbalanced SFT datasets, largely English-centric, degrade performance on non-English languages due to the failure in multilingual token alignment. To address this, we propose PARROT, a novel approach that leverages textual guidance for "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.02539","kind":"arxiv","version":3},"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/2406.02539/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":"2406.02539","created_at":"2026-07-05T11:08:57.978874+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.02539v3","created_at":"2026-07-05T11:08:57.978874+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.02539","created_at":"2026-07-05T11:08:57.978874+00:00"},{"alias_kind":"pith_short_12","alias_value":"QBKFGKTQHONZ","created_at":"2026-07-05T11:08:57.978874+00:00"},{"alias_kind":"pith_short_16","alias_value":"QBKFGKTQHONZHJRQ","created_at":"2026-07-05T11:08:57.978874+00:00"},{"alias_kind":"pith_short_8","alias_value":"QBKFGKTQ","created_at":"2026-07-05T11:08:57.978874+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.28551","citing_title":"DataComp-VLM: Improved Open Datasets for Vision-Language Models","ref_index":271,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03793","citing_title":"Exploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Models","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28551","citing_title":"DataComp-VLM: Improved Open Datasets for Vision-Language Models","ref_index":271,"is_internal_anchor":false},{"citing_arxiv_id":"2509.10026","citing_title":"LaV-CoT: Language-Aware Visual CoT with Multi-Aspect Reward Optimization for Real-World Multilingual VQA","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2504.10479","citing_title":"InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models","ref_index":110,"is_internal_anchor":false},{"citing_arxiv_id":"2412.05271","citing_title":"Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling","ref_index":219,"is_internal_anchor":false},{"citing_arxiv_id":"2508.18265","citing_title":"InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency","ref_index":114,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW","json":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW.json","graph_json":"https://pith.science/api/pith-number/QBKFGKTQHONZHJRQADQUBRM2YW/graph.json","events_json":"https://pith.science/api/pith-number/QBKFGKTQHONZHJRQADQUBRM2YW/events.json","paper":"https://pith.science/paper/QBKFGKTQ"},"agent_actions":{"view_html":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW","download_json":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW.json","view_paper":"https://pith.science/paper/QBKFGKTQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.02539&json=true","fetch_graph":"https://pith.science/api/pith-number/QBKFGKTQHONZHJRQADQUBRM2YW/graph.json","fetch_events":"https://pith.science/api/pith-number/QBKFGKTQHONZHJRQADQUBRM2YW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW/action/storage_attestation","attest_author":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW/action/author_attestation","sign_citation":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW/action/citation_signature","submit_replication":"https://pith.science/pith/QBKFGKTQHONZHJRQADQUBRM2YW/action/replication_record"}},"created_at":"2026-07-05T11:08:57.978874+00:00","updated_at":"2026-07-05T11:08:57.978874+00:00"}