{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:USIBBF6IGE6XFU36LO6RM6S66X","short_pith_number":"pith:USIBBF6I","schema_version":"1.0","canonical_sha256":"a4901097c8313d72d37e5bbd167a5ef5fc2b61d4a7ca97bc2d4c960693d6d3a3","source":{"kind":"arxiv","id":"2403.05676","version":1},"attestation_state":"computed","paper":{"title":"PipeRAG: Fast Retrieval-Augmented Generation via Algorithm-System Co-design","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bernie Wang, Boran Han, Jie Wang, Shuai Zhang, Tim Kraska, Wenqi Jiang","submitted_at":"2024-03-08T21:09:20Z","abstract_excerpt":"Retrieval-augmented generation (RAG) can enhance the generation quality of large language models (LLMs) by incorporating external token databases. However, retrievals from large databases can constitute a substantial portion of the overall generation time, particularly when retrievals are periodically performed to align the retrieved content with the latest states of generation. In this paper, we introduce PipeRAG, a novel algorithm-system co-design approach to reduce generation latency and enhance generation quality. PipeRAG integrates (1) pipeline parallelism to enable concurrent retrieval a"},"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":"2403.05676","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-08T21:09:20Z","cross_cats_sorted":[],"title_canon_sha256":"7be99a32088f6b6419de026043c3212d126bfe328d255064934244f04ca0bef8","abstract_canon_sha256":"9e929c16108640b68e5d1126683f2a7e402037eae1dd92abe3345a6a700d6c64"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:54:11.550024Z","signature_b64":"A2yLTXPuH893zoR4lU+yEhiL9nuI4x1isR0KvZWFlniU0n4wTAtHgUDoNxx/669tL+EMS2POBY/ZoVmBezqWDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a4901097c8313d72d37e5bbd167a5ef5fc2b61d4a7ca97bc2d4c960693d6d3a3","last_reissued_at":"2026-07-05T07:54:11.549611Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:54:11.549611Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PipeRAG: Fast Retrieval-Augmented Generation via Algorithm-System Co-design","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bernie Wang, Boran Han, Jie Wang, Shuai Zhang, Tim Kraska, Wenqi Jiang","submitted_at":"2024-03-08T21:09:20Z","abstract_excerpt":"Retrieval-augmented generation (RAG) can enhance the generation quality of large language models (LLMs) by incorporating external token databases. However, retrievals from large databases can constitute a substantial portion of the overall generation time, particularly when retrievals are periodically performed to align the retrieved content with the latest states of generation. In this paper, we introduce PipeRAG, a novel algorithm-system co-design approach to reduce generation latency and enhance generation quality. PipeRAG integrates (1) pipeline parallelism to enable concurrent retrieval a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.05676","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/2403.05676/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":"2403.05676","created_at":"2026-07-05T07:54:11.549675+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.05676v1","created_at":"2026-07-05T07:54:11.549675+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.05676","created_at":"2026-07-05T07:54:11.549675+00:00"},{"alias_kind":"pith_short_12","alias_value":"USIBBF6IGE6X","created_at":"2026-07-05T07:54:11.549675+00:00"},{"alias_kind":"pith_short_16","alias_value":"USIBBF6IGE6XFU36","created_at":"2026-07-05T07:54:11.549675+00:00"},{"alias_kind":"pith_short_8","alias_value":"USIBBF6I","created_at":"2026-07-05T07:54:11.549675+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09441","citing_title":"SIFT: Selective-Index For Fast Compute of RAG Prefill by Exploiting Attention Invariance","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2404.10981","citing_title":"A Survey on Retrieval-Augmented Text Generation for Large Language Models","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2407.13193","citing_title":"Retrieval-Augmented Generation for Natural Language Processing: A Survey","ref_index":85,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X","json":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X.json","graph_json":"https://pith.science/api/pith-number/USIBBF6IGE6XFU36LO6RM6S66X/graph.json","events_json":"https://pith.science/api/pith-number/USIBBF6IGE6XFU36LO6RM6S66X/events.json","paper":"https://pith.science/paper/USIBBF6I"},"agent_actions":{"view_html":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X","download_json":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X.json","view_paper":"https://pith.science/paper/USIBBF6I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.05676&json=true","fetch_graph":"https://pith.science/api/pith-number/USIBBF6IGE6XFU36LO6RM6S66X/graph.json","fetch_events":"https://pith.science/api/pith-number/USIBBF6IGE6XFU36LO6RM6S66X/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X/action/timestamp_anchor","attest_storage":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X/action/storage_attestation","attest_author":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X/action/author_attestation","sign_citation":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X/action/citation_signature","submit_replication":"https://pith.science/pith/USIBBF6IGE6XFU36LO6RM6S66X/action/replication_record"}},"created_at":"2026-07-05T07:54:11.549675+00:00","updated_at":"2026-07-05T07:54:11.549675+00:00"}