{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LFL6VP4FSEEAJJ2XC7P7Z4WCKB","short_pith_number":"pith:LFL6VP4F","schema_version":"1.0","canonical_sha256":"5957eabf85910804a75717dffcf2c2505441430674ad82ecff918e6a4391da62","source":{"kind":"arxiv","id":"2410.07590","version":1},"attestation_state":"computed","paper":{"title":"TurboRAG: Accelerating Retrieval-Augmented Generation with Precomputed KV Caches for Chunked Text","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Hua Wang, Songshuo Lu, Yaohua Tang, Yutian Rong, Zhi Chen","submitted_at":"2024-10-10T03:52:54Z","abstract_excerpt":"Current Retrieval-Augmented Generation (RAG) systems concatenate and process numerous retrieved document chunks for prefill which requires a large volume of computation, therefore leading to significant latency in time-to-first-token (TTFT). To reduce the computation overhead as well as TTFT, we introduce TurboRAG, a novel RAG system that redesigns the inference paradigm of the current RAG system by first pre-computing and storing the key-value (KV) caches of documents offline, and then directly retrieving the saved KV cache for prefill. Hence, online computation of KV caches is eliminated dur"},"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":"2410.07590","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-10-10T03:52:54Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"bc1ae264a57e047c624558823c14b8da194ee324e60a0085a31fddd1a88a0a40","abstract_canon_sha256":"c5f75abcc3fc4d162f897e849835e4d7dbc8805cff197cd5f8b501443006c0cb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:18:40.081800Z","signature_b64":"po4Dw1gvzLAQYwscH7i3ZG/M8ypkl14dKDSEHyjiiRBXk/toHx8tEtPmq0k6LwyCTv16IQJcF31hjHdfJc6dAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5957eabf85910804a75717dffcf2c2505441430674ad82ecff918e6a4391da62","last_reissued_at":"2026-07-05T09:18:40.081284Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:18:40.081284Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TurboRAG: Accelerating Retrieval-Augmented Generation with Precomputed KV Caches for Chunked Text","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Hua Wang, Songshuo Lu, Yaohua Tang, Yutian Rong, Zhi Chen","submitted_at":"2024-10-10T03:52:54Z","abstract_excerpt":"Current Retrieval-Augmented Generation (RAG) systems concatenate and process numerous retrieved document chunks for prefill which requires a large volume of computation, therefore leading to significant latency in time-to-first-token (TTFT). To reduce the computation overhead as well as TTFT, we introduce TurboRAG, a novel RAG system that redesigns the inference paradigm of the current RAG system by first pre-computing and storing the key-value (KV) caches of documents offline, and then directly retrieving the saved KV cache for prefill. Hence, online computation of KV caches is eliminated dur"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.07590","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/2410.07590/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":"2410.07590","created_at":"2026-07-05T09:18:40.081359+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.07590v1","created_at":"2026-07-05T09:18:40.081359+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.07590","created_at":"2026-07-05T09:18:40.081359+00:00"},{"alias_kind":"pith_short_12","alias_value":"LFL6VP4FSEEA","created_at":"2026-07-05T09:18:40.081359+00:00"},{"alias_kind":"pith_short_16","alias_value":"LFL6VP4FSEEAJJ2X","created_at":"2026-07-05T09:18:40.081359+00:00"},{"alias_kind":"pith_short_8","alias_value":"LFL6VP4F","created_at":"2026-07-05T09:18:40.081359+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"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":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27494","citing_title":"Grounded Cache Routing for Retrieval-Augmented Generation: When Is It Safe to Reuse an Answer?","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13097","citing_title":"Functional Cache Grafting: Robust and Rapid Code-Policy Synthesis for Embodied Agents","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2510.10129","citing_title":"CacheClip: Accelerating RAG with Effective KV Cache Reuse","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08584","citing_title":"CSAttention: Centroid-Scoring Attention for Accelerating LLM Inference","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB","json":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB.json","graph_json":"https://pith.science/api/pith-number/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/graph.json","events_json":"https://pith.science/api/pith-number/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/events.json","paper":"https://pith.science/paper/LFL6VP4F"},"agent_actions":{"view_html":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB","download_json":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB.json","view_paper":"https://pith.science/paper/LFL6VP4F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.07590&json=true","fetch_graph":"https://pith.science/api/pith-number/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/graph.json","fetch_events":"https://pith.science/api/pith-number/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/action/storage_attestation","attest_author":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/action/author_attestation","sign_citation":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/action/citation_signature","submit_replication":"https://pith.science/pith/LFL6VP4FSEEAJJ2XC7P7Z4WCKB/action/replication_record"}},"created_at":"2026-07-05T09:18:40.081359+00:00","updated_at":"2026-07-05T09:18:40.081359+00:00"}