{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:CEOA7KHFZCCHHQRCGKIYSSOYZD","short_pith_number":"pith:CEOA7KHF","schema_version":"1.0","canonical_sha256":"111c0fa8e5c88473c22232918949d8c8c2ca4faf75ef1331e2fabee532b74f37","source":{"kind":"arxiv","id":"2211.10411","version":1},"attestation_state":"computed","paper":{"title":"CITADEL: Conditional Token Interaction via Dynamic Lexical Routing for Efficient and Effective Multi-Vector Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.IR","authors_text":"Asish Ghoshal, Barlas Oguz, Jimmy Lin, Minghan Li, Sheng-Chieh Lin, Wen-tau Yih, Xilun Chen, Yashar Mehdad","submitted_at":"2022-11-18T18:27:35Z","abstract_excerpt":"Multi-vector retrieval methods combine the merits of sparse (e.g. BM25) and dense (e.g. DPR) retrievers and have achieved state-of-the-art performance on various retrieval tasks. These methods, however, are orders of magnitude slower and need much more space to store their indices compared to their single-vector counterparts. In this paper, we unify different multi-vector retrieval models from a token routing viewpoint and propose conditional token interaction via dynamic lexical routing, namely CITADEL, for efficient and effective multi-vector retrieval. CITADEL learns to route different toke"},"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":"2211.10411","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2022-11-18T18:27:35Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"e2e906aca2fb93cd0705c1095ed64c30d57569c159b2e7288ccb5bd97d31e265","abstract_canon_sha256":"8b36ee1e3f26ddb60fc697797eec136d05ea4939dbc82edd42998546a92da85e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:17:21.815592Z","signature_b64":"lnB8vCcgkWVI7bpBHGyiepiSl+S8gbIud1V39oqcCHhti6ZwhYXkCkqCESJJz92Bxq/xAjYFmhu/cX43cDewAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"111c0fa8e5c88473c22232918949d8c8c2ca4faf75ef1331e2fabee532b74f37","last_reissued_at":"2026-07-05T05:17:21.815169Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:17:21.815169Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CITADEL: Conditional Token Interaction via Dynamic Lexical Routing for Efficient and Effective Multi-Vector Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.IR","authors_text":"Asish Ghoshal, Barlas Oguz, Jimmy Lin, Minghan Li, Sheng-Chieh Lin, Wen-tau Yih, Xilun Chen, Yashar Mehdad","submitted_at":"2022-11-18T18:27:35Z","abstract_excerpt":"Multi-vector retrieval methods combine the merits of sparse (e.g. BM25) and dense (e.g. DPR) retrievers and have achieved state-of-the-art performance on various retrieval tasks. These methods, however, are orders of magnitude slower and need much more space to store their indices compared to their single-vector counterparts. In this paper, we unify different multi-vector retrieval models from a token routing viewpoint and propose conditional token interaction via dynamic lexical routing, namely CITADEL, for efficient and effective multi-vector retrieval. CITADEL learns to route different toke"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2211.10411","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/2211.10411/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":"2211.10411","created_at":"2026-07-05T05:17:21.815226+00:00"},{"alias_kind":"arxiv_version","alias_value":"2211.10411v1","created_at":"2026-07-05T05:17:21.815226+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2211.10411","created_at":"2026-07-05T05:17:21.815226+00:00"},{"alias_kind":"pith_short_12","alias_value":"CEOA7KHFZCCH","created_at":"2026-07-05T05:17:21.815226+00:00"},{"alias_kind":"pith_short_16","alias_value":"CEOA7KHFZCCHHQRC","created_at":"2026-07-05T05:17:21.815226+00:00"},{"alias_kind":"pith_short_8","alias_value":"CEOA7KHF","created_at":"2026-07-05T05:17:21.815226+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.00041","citing_title":"Decoding Dense Embeddings: Sparse Autoencoders for Interpreting and Discretizing Dense Retrieval","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD","json":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD.json","graph_json":"https://pith.science/api/pith-number/CEOA7KHFZCCHHQRCGKIYSSOYZD/graph.json","events_json":"https://pith.science/api/pith-number/CEOA7KHFZCCHHQRCGKIYSSOYZD/events.json","paper":"https://pith.science/paper/CEOA7KHF"},"agent_actions":{"view_html":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD","download_json":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD.json","view_paper":"https://pith.science/paper/CEOA7KHF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2211.10411&json=true","fetch_graph":"https://pith.science/api/pith-number/CEOA7KHFZCCHHQRCGKIYSSOYZD/graph.json","fetch_events":"https://pith.science/api/pith-number/CEOA7KHFZCCHHQRCGKIYSSOYZD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD/action/storage_attestation","attest_author":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD/action/author_attestation","sign_citation":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD/action/citation_signature","submit_replication":"https://pith.science/pith/CEOA7KHFZCCHHQRCGKIYSSOYZD/action/replication_record"}},"created_at":"2026-07-05T05:17:21.815226+00:00","updated_at":"2026-07-05T05:17:21.815226+00:00"}