{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2KSABC4PB6Q3Y2QTZAFL6BGG6J","short_pith_number":"pith:2KSABC4P","schema_version":"1.0","canonical_sha256":"d2a4008b8f0fa1bc6a13c80abf04c6f24ba84faec9b11b6dccf1af116d5c8c55","source":{"kind":"arxiv","id":"2506.18902","version":3},"attestation_state":"computed","paper":{"title":"jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.IR"],"primary_cat":"cs.AI","authors_text":"Andrei Ungureanu, Bo Wang, Han Xiao, Isabelle Mohr, Maximilian Werk, Michael G\\\"unther, Mohammad Kalim Akram, Nan Wang, Saba Sturua, Scott Martens, Sedigheh Eslami","submitted_at":"2025-06-23T17:59:55Z","abstract_excerpt":"We introduce jina-embeddings-v4, a 3.8 billion parameter multimodal embedding model that unifies text and image representations through a novel architecture supporting both single-vector and multi-vector embeddings in the late interaction style. The model incorporates task-specific Low-Rank Adaptation (LoRA) adapters to optimize performance across diverse retrieval scenarios, including query-document retrieval, semantic text similarity, and code search. Comprehensive evaluations demonstrate that jina-embeddings-v4 achieves state-of-the-art performance on both single-modal and cross-modal retri"},"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":"2506.18902","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-06-23T17:59:55Z","cross_cats_sorted":["cs.CL","cs.IR"],"title_canon_sha256":"535d92a8c6d2b55081763a5dbb17969f7293c17a34b2dfe1bcba2672f035c348","abstract_canon_sha256":"ea593f231e45389baaabc72612ffb4ac97e406a08b550bccd28f061db600d1aa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:33:05.725311Z","signature_b64":"20wXdOSnoHTpIuGZ/oTwqwrurSk7zdsI3S28k9By4PI40cwAVHco9pNvt0dmukN/0lz2NgiW4cnnRNtfSf/2Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d2a4008b8f0fa1bc6a13c80abf04c6f24ba84faec9b11b6dccf1af116d5c8c55","last_reissued_at":"2026-07-05T11:33:05.724751Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:33:05.724751Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.IR"],"primary_cat":"cs.AI","authors_text":"Andrei Ungureanu, Bo Wang, Han Xiao, Isabelle Mohr, Maximilian Werk, Michael G\\\"unther, Mohammad Kalim Akram, Nan Wang, Saba Sturua, Scott Martens, Sedigheh Eslami","submitted_at":"2025-06-23T17:59:55Z","abstract_excerpt":"We introduce jina-embeddings-v4, a 3.8 billion parameter multimodal embedding model that unifies text and image representations through a novel architecture supporting both single-vector and multi-vector embeddings in the late interaction style. The model incorporates task-specific Low-Rank Adaptation (LoRA) adapters to optimize performance across diverse retrieval scenarios, including query-document retrieval, semantic text similarity, and code search. Comprehensive evaluations demonstrate that jina-embeddings-v4 achieves state-of-the-art performance on both single-modal and cross-modal retri"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.18902","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/2506.18902/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":"2506.18902","created_at":"2026-07-05T11:33:05.724822+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.18902v3","created_at":"2026-07-05T11:33:05.724822+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.18902","created_at":"2026-07-05T11:33:05.724822+00:00"},{"alias_kind":"pith_short_12","alias_value":"2KSABC4PB6Q3","created_at":"2026-07-05T11:33:05.724822+00:00"},{"alias_kind":"pith_short_16","alias_value":"2KSABC4PB6Q3Y2QT","created_at":"2026-07-05T11:33:05.724822+00:00"},{"alias_kind":"pith_short_8","alias_value":"2KSABC4P","created_at":"2026-07-05T11:33:05.724822+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06036","citing_title":"Learn to Pool: Lightweight Fine-Tuning for Flexible Multi-Vector Compression","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2605.24938","citing_title":"Your Embedding Model is SMARTer Than You Think","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30027","citing_title":"DocRetriever: A Plug-and-Play Framework for Multimodal Document Retrieval with Comprehensive Benchmark","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2510.05038","citing_title":"Guided Query Refinement: Multimodal Hybrid Retrieval with Test-Time Optimization","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2510.15946","citing_title":"Fall into a Pit, Gain in a Wit: Cognitive-Guided Harmful Meme Detection via Misjudgment Risk Pattern Retrieval","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2512.04847","citing_title":"Language Models as Semantic Teachers: Post-Training Alignment for Medical Audio Understanding","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2603.12572","citing_title":"LMEB: Long-horizon Memory Embedding Benchmark","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2603.19339","citing_title":"Spectral Tempering for Embedding Compression in Dense Passage Retrieval","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2601.04720","citing_title":"Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08421","citing_title":"Beyond Bag-of-Patches: Learning Global Layout via Textual Supervision for Late-Interaction Visual Document Retrieval","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10167","citing_title":"Visual Late Chunking: An Empirical Study of Contextual Chunking for Efficient Visual Document Retrieval","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12081","citing_title":"Human-Inspired Context-Selective Multimodal Memory for Social Robots","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J","json":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J.json","graph_json":"https://pith.science/api/pith-number/2KSABC4PB6Q3Y2QTZAFL6BGG6J/graph.json","events_json":"https://pith.science/api/pith-number/2KSABC4PB6Q3Y2QTZAFL6BGG6J/events.json","paper":"https://pith.science/paper/2KSABC4P"},"agent_actions":{"view_html":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J","download_json":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J.json","view_paper":"https://pith.science/paper/2KSABC4P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.18902&json=true","fetch_graph":"https://pith.science/api/pith-number/2KSABC4PB6Q3Y2QTZAFL6BGG6J/graph.json","fetch_events":"https://pith.science/api/pith-number/2KSABC4PB6Q3Y2QTZAFL6BGG6J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J/action/storage_attestation","attest_author":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J/action/author_attestation","sign_citation":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J/action/citation_signature","submit_replication":"https://pith.science/pith/2KSABC4PB6Q3Y2QTZAFL6BGG6J/action/replication_record"}},"created_at":"2026-07-05T11:33:05.724822+00:00","updated_at":"2026-07-05T11:33:05.724822+00:00"}