{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WYGFUPKUZQ45Z2WFOAI2V52JF4","short_pith_number":"pith:WYGFUPKU","schema_version":"1.0","canonical_sha256":"b60c5a3d54cc39dceac57011aaf7492f09f45cd0c411098ea8c2a99701e9f1f5","source":{"kind":"arxiv","id":"2402.04252","version":1},"attestation_state":"computed","paper":{"title":"EVA-CLIP-18B: Scaling CLIP to 18 Billion Parameters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fan Zhang, Jinsheng Wang, Qiying Yu, Quan Sun, Xiaosong Zhang, Xinlong Wang, Yufeng Cui","submitted_at":"2024-02-06T18:59:48Z","abstract_excerpt":"Scaling up contrastive language-image pretraining (CLIP) is critical for empowering both vision and multimodal models. We present EVA-CLIP-18B, the largest and most powerful open-source CLIP model to date, with 18-billion parameters. With only 6-billion training samples seen, EVA-CLIP-18B achieves an exceptional 80.7% zero-shot top-1 accuracy averaged across 27 widely recognized image classification benchmarks, outperforming its forerunner EVA-CLIP (5-billion parameters) and other open-source CLIP models by a large margin. Remarkably, we observe a consistent performance improvement with the mo"},"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":"2402.04252","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-02-06T18:59:48Z","cross_cats_sorted":[],"title_canon_sha256":"250f2b45e4d706aff8e0908abc9f3d6a873f5221311df2b9d088e6d3f97b997a","abstract_canon_sha256":"14d9356cb803af4cb67f6aecd2e4ab32ca5ad50a259236d822f6d8bbee258f7e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:42:05.422899Z","signature_b64":"s7Cq61wYQHFlIXb841xJvSmHDYbo5DQ0Q7Hu2QJy7EEXtiCtgWU++gQCjINbM8N5KnL3It+PNrcHlPpi4rJNBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b60c5a3d54cc39dceac57011aaf7492f09f45cd0c411098ea8c2a99701e9f1f5","last_reissued_at":"2026-07-05T07:42:05.422376Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:42:05.422376Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EVA-CLIP-18B: Scaling CLIP to 18 Billion Parameters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fan Zhang, Jinsheng Wang, Qiying Yu, Quan Sun, Xiaosong Zhang, Xinlong Wang, Yufeng Cui","submitted_at":"2024-02-06T18:59:48Z","abstract_excerpt":"Scaling up contrastive language-image pretraining (CLIP) is critical for empowering both vision and multimodal models. We present EVA-CLIP-18B, the largest and most powerful open-source CLIP model to date, with 18-billion parameters. With only 6-billion training samples seen, EVA-CLIP-18B achieves an exceptional 80.7% zero-shot top-1 accuracy averaged across 27 widely recognized image classification benchmarks, outperforming its forerunner EVA-CLIP (5-billion parameters) and other open-source CLIP models by a large margin. Remarkably, we observe a consistent performance improvement with the mo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.04252","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/2402.04252/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":"2402.04252","created_at":"2026-07-05T07:42:05.422435+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.04252v1","created_at":"2026-07-05T07:42:05.422435+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.04252","created_at":"2026-07-05T07:42:05.422435+00:00"},{"alias_kind":"pith_short_12","alias_value":"WYGFUPKUZQ45","created_at":"2026-07-05T07:42:05.422435+00:00"},{"alias_kind":"pith_short_16","alias_value":"WYGFUPKUZQ45Z2WF","created_at":"2026-07-05T07:42:05.422435+00:00"},{"alias_kind":"pith_short_8","alias_value":"WYGFUPKU","created_at":"2026-07-05T07:42:05.422435+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26794","citing_title":"ReasonCLIP-58M: Visually Grounded Commonsense Reasoning Supervision for CLIP","ref_index":80,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23881","citing_title":"Ground Then Rank: Revisiting Knowledge-Based VQA with Training-Free Entity Identification","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20280","citing_title":"ELVA: Exploring Ranking-Driven Universal Multimodal Retrieval","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20104","citing_title":"Sensorimotor World Models: Perception for Action via Inverse Dynamics","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00159","citing_title":"Identifying and Resolving Pitfalls of Knowledge-Based VQA Benchmarks: Auditing, Repairing, and Augmenting","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2603.11689","citing_title":"Explicit Logic Channel for Validation and Enhancement of MLLMs on Zero-Shot Tasks","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24943","citing_title":"Perceive, Verify and Understand Long Video: Multi-Granular Perception and Active Verification via Interactive Agents","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2601.13856","citing_title":"QKVQA: Question-Focused Filtering for Knowledge-based VQA","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2504.13181","citing_title":"Perception Encoder: The best visual embeddings are not at the output of the network","ref_index":130,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03231","citing_title":"CoME-VL: Scaling Complementary Multi-Encoder Vision-Language Learning","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04969","citing_title":"MG$^2$-RAG: Multi-Granularity Graph for Multimodal Retrieval-Augmented Generation","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21444","citing_title":"HiCrew: Hierarchical Reasoning for Long-Form Video Understanding via Question-Aware Multi-Agent Collaboration","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12033","citing_title":"Benchmarking Deflection and Hallucination in Large Vision-Language Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12391","citing_title":"Chain-of-Models Pre-Training: Rethinking Training Acceleration of Vision Foundation Models","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20760","citing_title":"Exploring High-Order Self-Similarity for Video Understanding","ref_index":74,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4","json":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4.json","graph_json":"https://pith.science/api/pith-number/WYGFUPKUZQ45Z2WFOAI2V52JF4/graph.json","events_json":"https://pith.science/api/pith-number/WYGFUPKUZQ45Z2WFOAI2V52JF4/events.json","paper":"https://pith.science/paper/WYGFUPKU"},"agent_actions":{"view_html":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4","download_json":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4.json","view_paper":"https://pith.science/paper/WYGFUPKU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.04252&json=true","fetch_graph":"https://pith.science/api/pith-number/WYGFUPKUZQ45Z2WFOAI2V52JF4/graph.json","fetch_events":"https://pith.science/api/pith-number/WYGFUPKUZQ45Z2WFOAI2V52JF4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4/action/storage_attestation","attest_author":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4/action/author_attestation","sign_citation":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4/action/citation_signature","submit_replication":"https://pith.science/pith/WYGFUPKUZQ45Z2WFOAI2V52JF4/action/replication_record"}},"created_at":"2026-07-05T07:42:05.422435+00:00","updated_at":"2026-07-05T07:42:05.422435+00:00"}