{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PWUPYNX5G372REPQYNU72QTD7J","short_pith_number":"pith:PWUPYNX5","schema_version":"1.0","canonical_sha256":"7da8fc36fd36ffa891f0c369fd4263fa58ebebd7f518da3ef1f1951b830a31e1","source":{"kind":"arxiv","id":"2412.14475","version":1},"attestation_state":"computed","paper":{"title":"MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bo Zhao, Chen Jason Zhang, Defu Lian, Junjie Zhou, Shitao Xiao, Yongping Xiong, Yueze Wang, Ze Liu, Zheng Liu","submitted_at":"2024-12-19T02:49:55Z","abstract_excerpt":"Despite the rapidly growing demand for multimodal retrieval, progress in this field remains severely constrained by a lack of training data. In this paper, we introduce MegaPairs, a novel data synthesis method that leverages vision language models (VLMs) and open-domain images, together with a massive synthetic dataset generated from this method. Our empirical analysis shows that MegaPairs generates high-quality data, enabling the multimodal retriever to significantly outperform the baseline model trained on 70$\\times$ more data from existing datasets. Moreover, since MegaPairs solely relies o"},"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":"2412.14475","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-19T02:49:55Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"b046f6de1d17714811a454541ae980a121839f1ee7a4204fbc76ac41235ee603","abstract_canon_sha256":"67a8010213e89bdd7248e00a7644f5ee772124b0205cdf9e4b1866ee0fce9ae1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:51:47.881875Z","signature_b64":"/5trIkIayYCwIxNHmqIWwytyWWTaLDjv6nsVUXWVszh10XG3TldUMnYcqU8kYoOJLIIVJvWzLMR+THn2RqrWDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7da8fc36fd36ffa891f0c369fd4263fa58ebebd7f518da3ef1f1951b830a31e1","last_reissued_at":"2026-07-05T09:51:47.881371Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:51:47.881371Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bo Zhao, Chen Jason Zhang, Defu Lian, Junjie Zhou, Shitao Xiao, Yongping Xiong, Yueze Wang, Ze Liu, Zheng Liu","submitted_at":"2024-12-19T02:49:55Z","abstract_excerpt":"Despite the rapidly growing demand for multimodal retrieval, progress in this field remains severely constrained by a lack of training data. In this paper, we introduce MegaPairs, a novel data synthesis method that leverages vision language models (VLMs) and open-domain images, together with a massive synthetic dataset generated from this method. Our empirical analysis shows that MegaPairs generates high-quality data, enabling the multimodal retriever to significantly outperform the baseline model trained on 70$\\times$ more data from existing datasets. Moreover, since MegaPairs solely relies o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.14475","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/2412.14475/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":"2412.14475","created_at":"2026-07-05T09:51:47.881432+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.14475v1","created_at":"2026-07-05T09:51:47.881432+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.14475","created_at":"2026-07-05T09:51:47.881432+00:00"},{"alias_kind":"pith_short_12","alias_value":"PWUPYNX5G372","created_at":"2026-07-05T09:51:47.881432+00:00"},{"alias_kind":"pith_short_16","alias_value":"PWUPYNX5G372REPQ","created_at":"2026-07-05T09:51:47.881432+00:00"},{"alias_kind":"pith_short_8","alias_value":"PWUPYNX5","created_at":"2026-07-05T09:51:47.881432+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18434","citing_title":"TIGER-FG: Text-Guided Implicit Fine-Grained Grounding for E-commerce Retrieval","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18095","citing_title":"MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24621","citing_title":"FreeRet: MLLMs as Training-Free Retrievers","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02073","citing_title":"PLUME: Latent Reasoning Based Universal Multimodal Embedding","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25273","citing_title":"Combating Visual Neglect and Semantic Drift in Large Multimodal Models for Enhanced Cross-Modal Retrieval","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07220","citing_title":"HIVE: Query, Hypothesize, Verify An LLM Framework for Multimodal Reasoning-Intensive Retrieval","ref_index":41,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J","json":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J.json","graph_json":"https://pith.science/api/pith-number/PWUPYNX5G372REPQYNU72QTD7J/graph.json","events_json":"https://pith.science/api/pith-number/PWUPYNX5G372REPQYNU72QTD7J/events.json","paper":"https://pith.science/paper/PWUPYNX5"},"agent_actions":{"view_html":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J","download_json":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J.json","view_paper":"https://pith.science/paper/PWUPYNX5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.14475&json=true","fetch_graph":"https://pith.science/api/pith-number/PWUPYNX5G372REPQYNU72QTD7J/graph.json","fetch_events":"https://pith.science/api/pith-number/PWUPYNX5G372REPQYNU72QTD7J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J/action/storage_attestation","attest_author":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J/action/author_attestation","sign_citation":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J/action/citation_signature","submit_replication":"https://pith.science/pith/PWUPYNX5G372REPQYNU72QTD7J/action/replication_record"}},"created_at":"2026-07-05T09:51:47.881432+00:00","updated_at":"2026-07-05T09:51:47.881432+00:00"}