{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NQS6LYOA7DVZ7ZBA3U6ZDVARX2","short_pith_number":"pith:NQS6LYOA","schema_version":"1.0","canonical_sha256":"6c25e5e1c0f8eb9fe420dd3d91d411bea7906835a336d3eea4d97b8787c66bf3","source":{"kind":"arxiv","id":"2505.19650","version":2},"attestation_state":"computed","paper":{"title":"Modality Curation: Building Universal Embeddings for Advanced Multimodal Information Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.IR","cs.MM"],"primary_cat":"cs.CV","authors_text":"Daling Wang, Fanheng Kong, Fuzheng Zhang, Guorui Zhou, Hongzhi Zhang, Jingyuan Zhang, Shi Feng, Victoria W., Xiaocui Yang, Yahui Liu, Yu Tian","submitted_at":"2025-05-26T08:09:44Z","abstract_excerpt":"Multimodal information retrieval (MIR) faces inherent challenges due to the heterogeneity of data sources and the complexity of cross-modal alignment. While previous studies have identified modal gaps in feature spaces, a systematic approach to address these challenges remains unexplored. In this work, we introduce UNITE, a universal framework that tackles these challenges through two critical yet underexplored aspects: data curation and modality-aware training configurations. Our work provides the first comprehensive analysis of how modality-specific data properties influence downstream task "},"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":"2505.19650","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-05-26T08:09:44Z","cross_cats_sorted":["cs.IR","cs.MM"],"title_canon_sha256":"ad3b256e5477b999825def95c7116ada10777170c698a11a621e0d212f356af2","abstract_canon_sha256":"f2b1ed0c5b5632354698e24a92efba0d8007023e580e8c9dc71e286fbf13d8e5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:10.218354Z","signature_b64":"n3gqDFrHUoYDSglOIt6p994dFJ8dq6zSQTyy5uX/dIWytLWWPceq57t9MqrmwuTpxnRNWXMdjtAOT+m+B6umCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6c25e5e1c0f8eb9fe420dd3d91d411bea7906835a336d3eea4d97b8787c66bf3","last_reissued_at":"2026-07-05T11:10:10.217860Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:10.217860Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Modality Curation: Building Universal Embeddings for Advanced Multimodal Information Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.IR","cs.MM"],"primary_cat":"cs.CV","authors_text":"Daling Wang, Fanheng Kong, Fuzheng Zhang, Guorui Zhou, Hongzhi Zhang, Jingyuan Zhang, Shi Feng, Victoria W., Xiaocui Yang, Yahui Liu, Yu Tian","submitted_at":"2025-05-26T08:09:44Z","abstract_excerpt":"Multimodal information retrieval (MIR) faces inherent challenges due to the heterogeneity of data sources and the complexity of cross-modal alignment. While previous studies have identified modal gaps in feature spaces, a systematic approach to address these challenges remains unexplored. In this work, we introduce UNITE, a universal framework that tackles these challenges through two critical yet underexplored aspects: data curation and modality-aware training configurations. Our work provides the first comprehensive analysis of how modality-specific data properties influence downstream task "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.19650","kind":"arxiv","version":2},"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/2505.19650/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":"2505.19650","created_at":"2026-07-05T11:10:10.217922+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.19650v2","created_at":"2026-07-05T11:10:10.217922+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.19650","created_at":"2026-07-05T11:10:10.217922+00:00"},{"alias_kind":"pith_short_12","alias_value":"NQS6LYOA7DVZ","created_at":"2026-07-05T11:10:10.217922+00:00"},{"alias_kind":"pith_short_16","alias_value":"NQS6LYOA7DVZ7ZBA","created_at":"2026-07-05T11:10:10.217922+00:00"},{"alias_kind":"pith_short_8","alias_value":"NQS6LYOA","created_at":"2026-07-05T11:10:10.217922+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20280","citing_title":"ELVA: Exploring Ranking-Driven Universal Multimodal Retrieval","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18095","citing_title":"MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24621","citing_title":"FreeRet: MLLMs as Training-Free Retrievers","ref_index":10,"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":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22280","citing_title":"Beyond Chain-of-Thought: Rewrite as a Universal Interface for Generative Multimodal Embeddings","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2","json":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2.json","graph_json":"https://pith.science/api/pith-number/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/graph.json","events_json":"https://pith.science/api/pith-number/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/events.json","paper":"https://pith.science/paper/NQS6LYOA"},"agent_actions":{"view_html":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2","download_json":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2.json","view_paper":"https://pith.science/paper/NQS6LYOA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.19650&json=true","fetch_graph":"https://pith.science/api/pith-number/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/graph.json","fetch_events":"https://pith.science/api/pith-number/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/action/storage_attestation","attest_author":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/action/author_attestation","sign_citation":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/action/citation_signature","submit_replication":"https://pith.science/pith/NQS6LYOA7DVZ7ZBA3U6ZDVARX2/action/replication_record"}},"created_at":"2026-07-05T11:10:10.217922+00:00","updated_at":"2026-07-05T11:10:10.217922+00:00"}