{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5RQZVUR6SVFKFO4BPGDBZJM6QN","short_pith_number":"pith:5RQZVUR6","schema_version":"1.0","canonical_sha256":"ec619ad23e954aa2bb8179861ca59e835201365af357c693efa83074afa5a80b","source":{"kind":"arxiv","id":"2507.22062","version":3},"attestation_state":"computed","paper":{"title":"Meta CLIP 2: A Worldwide Scaling Recipe","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Ching-Feng Yeh, Dong Wang, Hu Xu, James Glass, Jason Weston, Kehan Lyu, Lifei Huang, Luke Zettlemoyer, Ramya Raghavendra, Saining Xie, Shang-Wen Li, Wen-tau Yih, Xinlei Chen, Yang Li, Yung-Sung Chuang, Zhuang Liu","submitted_at":"2025-07-29T17:59:58Z","abstract_excerpt":"Contrastive Language-Image Pretraining (CLIP) is a popular foundation model, supporting from zero-shot classification, retrieval to encoders for multimodal large language models (MLLMs). Although CLIP is successfully trained on billion-scale image-text pairs from the English world, scaling CLIP's training further to learning from the worldwide web data is still challenging: (1) no curation method is available to handle data points from non-English world; (2) the English performance from existing multilingual CLIP is worse than its English-only counterpart, i.e., \"curse of multilinguality\" that"},"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":"2507.22062","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-07-29T17:59:58Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"4b7a1f5155ed61daf34ce36a2c788297c056dc9e28c054131eec9944d3c05b1e","abstract_canon_sha256":"16fba1b2fcb35df4ce2a18afa2baccfda11d326ef12af62494035e6e6a1b1e93"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:46:46.096139Z","signature_b64":"Rj6MzgIJ9vdVaMQ3bhF53oIdT1GUN4x5wJaRqTME4pNu9amCEAriAeF6RxZtsDkdy/5Y+p213sAMgZSOxmlLCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ec619ad23e954aa2bb8179861ca59e835201365af357c693efa83074afa5a80b","last_reissued_at":"2026-07-05T11:46:46.095683Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:46:46.095683Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Meta CLIP 2: A Worldwide Scaling Recipe","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Ching-Feng Yeh, Dong Wang, Hu Xu, James Glass, Jason Weston, Kehan Lyu, Lifei Huang, Luke Zettlemoyer, Ramya Raghavendra, Saining Xie, Shang-Wen Li, Wen-tau Yih, Xinlei Chen, Yang Li, Yung-Sung Chuang, Zhuang Liu","submitted_at":"2025-07-29T17:59:58Z","abstract_excerpt":"Contrastive Language-Image Pretraining (CLIP) is a popular foundation model, supporting from zero-shot classification, retrieval to encoders for multimodal large language models (MLLMs). Although CLIP is successfully trained on billion-scale image-text pairs from the English world, scaling CLIP's training further to learning from the worldwide web data is still challenging: (1) no curation method is available to handle data points from non-English world; (2) the English performance from existing multilingual CLIP is worse than its English-only counterpart, i.e., \"curse of multilinguality\" that"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.22062","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/2507.22062/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":"2507.22062","created_at":"2026-07-05T11:46:46.095741+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.22062v3","created_at":"2026-07-05T11:46:46.095741+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.22062","created_at":"2026-07-05T11:46:46.095741+00:00"},{"alias_kind":"pith_short_12","alias_value":"5RQZVUR6SVFK","created_at":"2026-07-05T11:46:46.095741+00:00"},{"alias_kind":"pith_short_16","alias_value":"5RQZVUR6SVFKFO4B","created_at":"2026-07-05T11:46:46.095741+00:00"},{"alias_kind":"pith_short_8","alias_value":"5RQZVUR6","created_at":"2026-07-05T11:46:46.095741+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26794","citing_title":"ReasonCLIP-58M: Visually Grounded Commonsense Reasoning Supervision for CLIP","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00684","citing_title":"AdaBoosting Text Prompts for Vision-Language Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04613","citing_title":"Beyond Symmetric Alignment: Spectral Diagnostics of Modality Imbalance in Vision-Language Models in the Medical Domain","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2509.18095","citing_title":"MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2509.23370","citing_title":"GRAPE: Let GRPO Supervise Query Rewriting by Ranking for Retrieval","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2511.23170","citing_title":"PowerCLIP: Powerset Alignment for Contrastive Pre-Training","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2602.01738","citing_title":"Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2603.27759","citing_title":"When Surfaces Lie: Exploiting Wrinkle-Induced Attention Shift to Attack Vision-Language Models","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12650","citing_title":"CRAFT: Clinical Reward-Aligned Finetuning for Medical Image Synthesis","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03555","citing_title":"HEDGE: Heterogeneous Ensemble for Detection of AI-GEnerated Images in the Wild","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03558","citing_title":"LOGER: Local--Global Ensemble for Robust Deepfake Detection in the Wild","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10228","citing_title":"FLARE: Full-Modality Long-Video Audiovisual Retrieval Benchmark with User-Simulated Queries","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21681","citing_title":"Sapiens2","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12307","citing_title":"Boosting Robust AIGI Detection with LoRA-based Pairwise Training","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13313","citing_title":"Concrete Jungle: Towards Concreteness Paved Contrastive Negative Mining for Compositional Understanding","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08008","citing_title":"SearchAD: Large-Scale Rare Image Retrieval Dataset for Autonomous Driving","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN","json":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN.json","graph_json":"https://pith.science/api/pith-number/5RQZVUR6SVFKFO4BPGDBZJM6QN/graph.json","events_json":"https://pith.science/api/pith-number/5RQZVUR6SVFKFO4BPGDBZJM6QN/events.json","paper":"https://pith.science/paper/5RQZVUR6"},"agent_actions":{"view_html":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN","download_json":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN.json","view_paper":"https://pith.science/paper/5RQZVUR6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.22062&json=true","fetch_graph":"https://pith.science/api/pith-number/5RQZVUR6SVFKFO4BPGDBZJM6QN/graph.json","fetch_events":"https://pith.science/api/pith-number/5RQZVUR6SVFKFO4BPGDBZJM6QN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN/action/storage_attestation","attest_author":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN/action/author_attestation","sign_citation":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN/action/citation_signature","submit_replication":"https://pith.science/pith/5RQZVUR6SVFKFO4BPGDBZJM6QN/action/replication_record"}},"created_at":"2026-07-05T11:46:46.095741+00:00","updated_at":"2026-07-05T11:46:46.095741+00:00"}