{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:VPRIFEC477XGEA2NIGFUFFO6YV","short_pith_number":"pith:VPRIFEC4","schema_version":"1.0","canonical_sha256":"abe282905cffee62034d418b4295dec566bd67aebe54788bf6bbefe15d6b7a28","source":{"kind":"arxiv","id":"2405.20204","version":2},"attestation_state":"computed","paper":{"title":"Jina CLIP: Your CLIP Model Is Also Your Text Retriever","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV","cs.IR"],"primary_cat":"cs.CL","authors_text":"Andreas Koukounas, Bo Wang, Georgios Mastrapas, Han Xiao, Isabelle Mohr, Joan Fontanals Mart\\'inez, Maximilian Werk, Michael G\\\"unther, Mohammad Kalim Akram, Nan Wang, Saahil Ognawala, Saba Sturua, Scott Martens, Susana Guzman","submitted_at":"2024-05-30T16:07:54Z","abstract_excerpt":"Contrastive Language-Image Pretraining (CLIP) is widely used to train models to align images and texts in a common embedding space by mapping them to fixed-sized vectors. These models are key to multimodal information retrieval and related tasks. However, CLIP models generally underperform in text-only tasks compared to specialized text models. This creates inefficiencies for information retrieval systems that keep separate embeddings and models for text-only and multimodal tasks. We propose a novel, multi-task contrastive training method to address this issue, which we use to train the jina-c"},"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":"2405.20204","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-30T16:07:54Z","cross_cats_sorted":["cs.AI","cs.CV","cs.IR"],"title_canon_sha256":"654a1314abdce96dbba9d83fc14cc7ba6e9981c550e05f6ca11c0691c2802e5a","abstract_canon_sha256":"4250b1c314083999839bfd17a09bd5c7a777198a75935e94a2914dc2122697ad"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:36:48.149917Z","signature_b64":"cBJeh7nu7xcXaYD+og9TKmPFX2kCbeBHGHTl1Ifs0TnsZMuRAjWewo5UfFD4WExc/JmuJLlJm5FBPIuHmMXUBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"abe282905cffee62034d418b4295dec566bd67aebe54788bf6bbefe15d6b7a28","last_reissued_at":"2026-07-05T08:36:48.149403Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:36:48.149403Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Jina CLIP: Your CLIP Model Is Also Your Text Retriever","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV","cs.IR"],"primary_cat":"cs.CL","authors_text":"Andreas Koukounas, Bo Wang, Georgios Mastrapas, Han Xiao, Isabelle Mohr, Joan Fontanals Mart\\'inez, Maximilian Werk, Michael G\\\"unther, Mohammad Kalim Akram, Nan Wang, Saahil Ognawala, Saba Sturua, Scott Martens, Susana Guzman","submitted_at":"2024-05-30T16:07:54Z","abstract_excerpt":"Contrastive Language-Image Pretraining (CLIP) is widely used to train models to align images and texts in a common embedding space by mapping them to fixed-sized vectors. These models are key to multimodal information retrieval and related tasks. However, CLIP models generally underperform in text-only tasks compared to specialized text models. This creates inefficiencies for information retrieval systems that keep separate embeddings and models for text-only and multimodal tasks. We propose a novel, multi-task contrastive training method to address this issue, which we use to train the jina-c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.20204","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/2405.20204/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":"2405.20204","created_at":"2026-07-05T08:36:48.149468+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.20204v2","created_at":"2026-07-05T08:36:48.149468+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.20204","created_at":"2026-07-05T08:36:48.149468+00:00"},{"alias_kind":"pith_short_12","alias_value":"VPRIFEC477XG","created_at":"2026-07-05T08:36:48.149468+00:00"},{"alias_kind":"pith_short_16","alias_value":"VPRIFEC477XGEA2N","created_at":"2026-07-05T08:36:48.149468+00:00"},{"alias_kind":"pith_short_8","alias_value":"VPRIFEC4","created_at":"2026-07-05T08:36:48.149468+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26794","citing_title":"ReasonCLIP-58M: Visually Grounded Commonsense Reasoning Supervision for CLIP","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13288","citing_title":"Cross-Modal Masked Compositional Concept Modeling for Enhancing Visio-Linguistic Compositionality","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08384","citing_title":"jina-embeddings-v5-omni: Geometry-preserving Embeddings via Locked Aligned Towers","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08384","citing_title":"jina-embeddings-v5-omni: Geometry-preserving Embeddings via Locked Aligned Towers","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08384","citing_title":"jina-embeddings-v5-omni: Geometry-preserving Embeddings via Locked Aligned Towers","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25102","citing_title":"One Perturbation, Two Failure Modes: Probing VLM Safety via Embedding-Guided Typographic Perturbations","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12371","citing_title":"Reading Between the Pixels: Linking Text-Image Embedding Alignment to Typographic Attack Success on Vision-Language Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07201","citing_title":"BRIDGE: Multimodal-to-Text Retrieval via Reinforcement-Learned Query Alignment","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07079","citing_title":"MARVEL: Multimodal Adaptive Reasoning-intensiVe Expand-rerank and retrievaL","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03045","citing_title":"TCD-Arena: Assessing Robustness of Time Series Causal Discovery Methods Against Assumption Violations","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV","json":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV.json","graph_json":"https://pith.science/api/pith-number/VPRIFEC477XGEA2NIGFUFFO6YV/graph.json","events_json":"https://pith.science/api/pith-number/VPRIFEC477XGEA2NIGFUFFO6YV/events.json","paper":"https://pith.science/paper/VPRIFEC4"},"agent_actions":{"view_html":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV","download_json":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV.json","view_paper":"https://pith.science/paper/VPRIFEC4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.20204&json=true","fetch_graph":"https://pith.science/api/pith-number/VPRIFEC477XGEA2NIGFUFFO6YV/graph.json","fetch_events":"https://pith.science/api/pith-number/VPRIFEC477XGEA2NIGFUFFO6YV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV/action/storage_attestation","attest_author":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV/action/author_attestation","sign_citation":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV/action/citation_signature","submit_replication":"https://pith.science/pith/VPRIFEC477XGEA2NIGFUFFO6YV/action/replication_record"}},"created_at":"2026-07-05T08:36:48.149468+00:00","updated_at":"2026-07-05T08:36:48.149468+00:00"}