{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:IDK5M4ZPLPSKZV5ICZQHMUQNQD","short_pith_number":"pith:IDK5M4ZP","schema_version":"1.0","canonical_sha256":"40d5d6732f5be4acd7a8166076520d80e201af3c36c1c7fc78c5c818da20d099","source":{"kind":"arxiv","id":"2606.09441","version":1},"attestation_state":"computed","paper":{"title":"SIFT: Selective-Index For Fast Compute of RAG Prefill by Exploiting Attention Invariance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AR"],"primary_cat":"cs.AI","authors_text":"Hritvik Taneja, Moinuddin Qureshi, Rya Sanovar, Srikant Bharadwaj","submitted_at":"2026-06-08T12:50:13Z","abstract_excerpt":"Retrieval-Augmented Generation (RAG) injects LLM queries with relevant documents to improve response quality. This injection increases prompt length and slows time to first token (TTFT). Unlike standard queries, RAG queries have a unique property of context reuse where the same documents recur across user queries. Thus, fully recomputing documents for every RAG query does redundant compute and increases TTFT. Prior works precompute KV tensors of RAG documents offline and coarsely recompute some tokens during online prefill. However, such KV reuse is often slower than full recomputation on mode"},"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":"2606.09441","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-06-08T12:50:13Z","cross_cats_sorted":["cs.AR"],"title_canon_sha256":"3d75ba19bdbcb7552324f5427f674f747f09f1e53f2fd6b3a9463ca261cf5bf7","abstract_canon_sha256":"c24144b8e8dfdc020b4d720ab3035274c7670450b966f97b291ff3e079f12155"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:08:48.707386Z","signature_b64":"pdJAZ17aW9GT5W8hXpxV5bO09IWl9K7xCaAUZwS4R8UR1orPPGWFpwMy+Rw5v6VOku3Ir7zmIuqWCUTNwIQwCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"40d5d6732f5be4acd7a8166076520d80e201af3c36c1c7fc78c5c818da20d099","last_reissued_at":"2026-06-09T02:08:48.706482Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:08:48.706482Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SIFT: Selective-Index For Fast Compute of RAG Prefill by Exploiting Attention Invariance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AR"],"primary_cat":"cs.AI","authors_text":"Hritvik Taneja, Moinuddin Qureshi, Rya Sanovar, Srikant Bharadwaj","submitted_at":"2026-06-08T12:50:13Z","abstract_excerpt":"Retrieval-Augmented Generation (RAG) injects LLM queries with relevant documents to improve response quality. This injection increases prompt length and slows time to first token (TTFT). Unlike standard queries, RAG queries have a unique property of context reuse where the same documents recur across user queries. Thus, fully recomputing documents for every RAG query does redundant compute and increases TTFT. Prior works precompute KV tensors of RAG documents offline and coarsely recompute some tokens during online prefill. However, such KV reuse is often slower than full recomputation on mode"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.09441","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/2606.09441/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":"2606.09441","created_at":"2026-06-09T02:08:48.706627+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.09441v1","created_at":"2026-06-09T02:08:48.706627+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.09441","created_at":"2026-06-09T02:08:48.706627+00:00"},{"alias_kind":"pith_short_12","alias_value":"IDK5M4ZPLPSK","created_at":"2026-06-09T02:08:48.706627+00:00"},{"alias_kind":"pith_short_16","alias_value":"IDK5M4ZPLPSKZV5I","created_at":"2026-06-09T02:08:48.706627+00:00"},{"alias_kind":"pith_short_8","alias_value":"IDK5M4ZP","created_at":"2026-06-09T02:08:48.706627+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD","json":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD.json","graph_json":"https://pith.science/api/pith-number/IDK5M4ZPLPSKZV5ICZQHMUQNQD/graph.json","events_json":"https://pith.science/api/pith-number/IDK5M4ZPLPSKZV5ICZQHMUQNQD/events.json","paper":"https://pith.science/paper/IDK5M4ZP"},"agent_actions":{"view_html":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD","download_json":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD.json","view_paper":"https://pith.science/paper/IDK5M4ZP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.09441&json=true","fetch_graph":"https://pith.science/api/pith-number/IDK5M4ZPLPSKZV5ICZQHMUQNQD/graph.json","fetch_events":"https://pith.science/api/pith-number/IDK5M4ZPLPSKZV5ICZQHMUQNQD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD/action/storage_attestation","attest_author":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD/action/author_attestation","sign_citation":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD/action/citation_signature","submit_replication":"https://pith.science/pith/IDK5M4ZPLPSKZV5ICZQHMUQNQD/action/replication_record"}},"created_at":"2026-06-09T02:08:48.706627+00:00","updated_at":"2026-06-09T02:08:48.706627+00:00"}