{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:URUH7CGY5JSYYLUFN72GO7KPYG","short_pith_number":"pith:URUH7CGY","schema_version":"1.0","canonical_sha256":"a4687f88d8ea658c2e856ff4677d4fc1a9b6b0c91ec3b57072c2704a10da90eb","source":{"kind":"arxiv","id":"2605.17640","version":1},"attestation_state":"computed","paper":{"title":"MARQUIS: A Three-Stage Pipeline for Video Retrieval-Augmented Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MARQUIS is a three-stage pipeline that lifts video retrieval-augmented generation performance on complex queries and long contexts.","cross_cats":["cs.CV"],"primary_cat":"cs.IR","authors_text":"Alexander Martin, Benjamin Van Durme, Debashish Chakraborty, Dengjia Zhang, Hanting Liu, Hanxiang Qin, Jialiang Jin, Katherine Guerrerio, Reno Kriz, Tyler Skow","submitted_at":"2026-05-17T20:19:04Z","abstract_excerpt":"Retrieval-augmented generation from videos requires systems to retrieve relevant audiovisual evidence from large corpora and synthesize it into coherent, attributed text. Current approaches struggle at both ends: retrieval methods fail on complex, multi-faceted queries that cannot be captured by a single embedding, while generation methods lack the high-level reasoning needed to synthesize across multiple videos and face memory constraints over long, multi-video contexts. We present MARQUIS: a three-stage pipeline that addresses these limitations through (1) query expansion, fusion, and rerank"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.17640","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-05-17T20:19:04Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"d7845a728df11776d8f1b0be42546af7c1e0a67091959524c8612dbf44164cd5","abstract_canon_sha256":"dbdcec7d7b0396ab89a39f08f277a62bae79d552ff17ac74ba103c97e80cd862"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:50.177758Z","signature_b64":"/u6fVElGjfAjQ+o2MM96eGH1AvotITXj9i4MDhCQiHTi37iF5O6EMScXIfIQDQAgleNgBKRDN13qd3kYOaKEAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a4687f88d8ea658c2e856ff4677d4fc1a9b6b0c91ec3b57072c2704a10da90eb","last_reissued_at":"2026-05-20T00:04:50.176880Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:50.176880Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MARQUIS: A Three-Stage Pipeline for Video Retrieval-Augmented Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MARQUIS is a three-stage pipeline that lifts video retrieval-augmented generation performance on complex queries and long contexts.","cross_cats":["cs.CV"],"primary_cat":"cs.IR","authors_text":"Alexander Martin, Benjamin Van Durme, Debashish Chakraborty, Dengjia Zhang, Hanting Liu, Hanxiang Qin, Jialiang Jin, Katherine Guerrerio, Reno Kriz, Tyler Skow","submitted_at":"2026-05-17T20:19:04Z","abstract_excerpt":"Retrieval-augmented generation from videos requires systems to retrieve relevant audiovisual evidence from large corpora and synthesize it into coherent, attributed text. Current approaches struggle at both ends: retrieval methods fail on complex, multi-faceted queries that cannot be captured by a single embedding, while generation methods lack the high-level reasoning needed to synthesize across multiple videos and face memory constraints over long, multi-video contexts. We present MARQUIS: a three-stage pipeline that addresses these limitations through (1) query expansion, fusion, and rerank"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On the MAGMaR2026 shared task, we improve retrieval performance from 0.195 to 0.759 (nDCG@10). For article generation, ITER-QA-BASE improves average human score from 3.09 to 3.83 over the CAG baseline, while MARQUIS-RLM achieves a human score of 3.30 and the strongest citation recall among non-QA systems.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The improvements are attributable to the three-stage pipeline design rather than implementation details, baseline choices, or task-specific tuning not described in the abstract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MARQUIS is a three-stage pipeline for video RAG that boosts retrieval nDCG@10 from 0.195 to 0.759 and generation human scores on the MAGMaR2026 shared task.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MARQUIS is a three-stage pipeline that lifts video retrieval-augmented generation performance on complex queries and long contexts.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"904bb27fdb0fac14a69bd5a3035c4a79d513aa3fd9e91240c04c9dbfb5cf4717"},"source":{"id":"2605.17640","kind":"arxiv","version":1},"verdict":{"id":"e481fff4-0d5a-4803-8a78-15654ab8ac84","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T22:34:18.815023Z","strongest_claim":"On the MAGMaR2026 shared task, we improve retrieval performance from 0.195 to 0.759 (nDCG@10). For article generation, ITER-QA-BASE improves average human score from 3.09 to 3.83 over the CAG baseline, while MARQUIS-RLM achieves a human score of 3.30 and the strongest citation recall among non-QA systems.","one_line_summary":"MARQUIS is a three-stage pipeline for video RAG that boosts retrieval nDCG@10 from 0.195 to 0.759 and generation human scores on the MAGMaR2026 shared task.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The improvements are attributable to the three-stage pipeline design rather than implementation details, baseline choices, or task-specific tuning not described in the abstract.","pith_extraction_headline":"MARQUIS is a three-stage pipeline that lifts video retrieval-augmented generation performance on complex queries and long contexts."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.17640/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T23:01:19.419980Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:40:54.013250Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T22:22:49.593847Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.553066Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.474662Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"d043bf1eb29d6f5527902c711f89d94c0eabfa0eb97e5375b106cd6c07d3c1fc"},"references":{"count":65,"sample":[{"doi":"","year":2018,"title":"Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks","work_id":"27eaec54-c105-4969-8188-da5f0fca3688","ref_index":1,"cited_arxiv_id":"2005.11401","is_internal_anchor":true},{"doi":"","year":2022,"title":"Streamlining evaluation with ir-measures. In Advances in Information Retrieval - 44th European Conference on IR Research, ECIR 2022, Stavanger , Norway, April 10-14, 2022, Proceedings, Part II, volume","work_id":"4caa9f2b-6109-43b9-b68d-b9e6d7a89e9d","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Alec Radford, Jong Wook Kim, Tao Xu, Greg Brock- man, Christine Mcleavey, and Ilya Sutskever","work_id":"d96fb1d9-0710-444d-844a-c3fa30480a4a","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"arXiv preprint arXiv:2509.23040 , year=","work_id":"74006986-b24a-48c9-a5c0-d5cfbc5a76ab","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Do not merge separate information needs into one sub-query","work_id":"a8e93cf7-6a17-459e-97b1-6f4c86d815b9","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":65,"snapshot_sha256":"0764efb5f9ce85d3f77ac3110f4f8b877ef9462cadeacd382490dd0f94746562","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"567e977a63a2b9527cb78563645db86a21ef925cadda70bd7ac9be622d869b5c"},"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":"2605.17640","created_at":"2026-05-20T00:04:50.177022+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.17640v1","created_at":"2026-05-20T00:04:50.177022+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17640","created_at":"2026-05-20T00:04:50.177022+00:00"},{"alias_kind":"pith_short_12","alias_value":"URUH7CGY5JSY","created_at":"2026-05-20T00:04:50.177022+00:00"},{"alias_kind":"pith_short_16","alias_value":"URUH7CGY5JSYYLUF","created_at":"2026-05-20T00:04:50.177022+00:00"},{"alias_kind":"pith_short_8","alias_value":"URUH7CGY","created_at":"2026-05-20T00:04:50.177022+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG","json":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG.json","graph_json":"https://pith.science/api/pith-number/URUH7CGY5JSYYLUFN72GO7KPYG/graph.json","events_json":"https://pith.science/api/pith-number/URUH7CGY5JSYYLUFN72GO7KPYG/events.json","paper":"https://pith.science/paper/URUH7CGY"},"agent_actions":{"view_html":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG","download_json":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG.json","view_paper":"https://pith.science/paper/URUH7CGY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.17640&json=true","fetch_graph":"https://pith.science/api/pith-number/URUH7CGY5JSYYLUFN72GO7KPYG/graph.json","fetch_events":"https://pith.science/api/pith-number/URUH7CGY5JSYYLUFN72GO7KPYG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG/action/storage_attestation","attest_author":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG/action/author_attestation","sign_citation":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG/action/citation_signature","submit_replication":"https://pith.science/pith/URUH7CGY5JSYYLUFN72GO7KPYG/action/replication_record"}},"created_at":"2026-05-20T00:04:50.177022+00:00","updated_at":"2026-05-20T00:04:50.177022+00:00"}