{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:L5WJP2YPS3T67AUPDYJN6GDT2G","short_pith_number":"pith:L5WJP2YP","schema_version":"1.0","canonical_sha256":"5f6c97eb0f96e7ef828f1e12df1873d1afac944c35e8472f5079a5ce3343e3ba","source":{"kind":"arxiv","id":"2403.19708","version":3},"attestation_state":"computed","paper":{"title":"Cost-Efficient Large Language Model Serving for Multi-turn Conversations with CachedAttention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Bin Gao, Djordje Jevdjic, Junbo Deng, Pengfei Zuo, Puru Sharma, Qingxuan Kang, Xingkun Yang, Zhou Yu, Zhuomin He","submitted_at":"2024-03-23T10:42:49Z","abstract_excerpt":"Interacting with humans through multi-turn conversations is a fundamental feature of large language models (LLMs). However, existing LLM serving engines executing multi-turn conversations are inefficient due to the need to repeatedly compute the key-value (KV) caches of historical tokens, incurring high serving costs. To address the problem, this paper proposes CachedAttention, a new attention mechanism that enables reuse of KV caches across multi-turn conversations, significantly reducing the repetitive computation overheads. CachedAttention maintains a hierarchical KV caching system that lev"},"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":"2403.19708","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-23T10:42:49Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"aaf35dcc51cd5df1174e1992dde2fb728c1010ef91b70c6d98717b5c58a04a42","abstract_canon_sha256":"038a86b101e9c62cf668c40270e5aa0cc2f98e89e1df085262abf0a111e2ca51"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:38:07.931776Z","signature_b64":"hhzTYnRBeUwDJVZPk5uCFsLaWgZTNXpTZHuXfowx8X2wbDo2cKgvfs9S6LYN4wDIu8EHGOK9RB4JjLA6lx6yCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5f6c97eb0f96e7ef828f1e12df1873d1afac944c35e8472f5079a5ce3343e3ba","last_reissued_at":"2026-07-05T08:38:07.931273Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:38:07.931273Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cost-Efficient Large Language Model Serving for Multi-turn Conversations with CachedAttention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Bin Gao, Djordje Jevdjic, Junbo Deng, Pengfei Zuo, Puru Sharma, Qingxuan Kang, Xingkun Yang, Zhou Yu, Zhuomin He","submitted_at":"2024-03-23T10:42:49Z","abstract_excerpt":"Interacting with humans through multi-turn conversations is a fundamental feature of large language models (LLMs). However, existing LLM serving engines executing multi-turn conversations are inefficient due to the need to repeatedly compute the key-value (KV) caches of historical tokens, incurring high serving costs. To address the problem, this paper proposes CachedAttention, a new attention mechanism that enables reuse of KV caches across multi-turn conversations, significantly reducing the repetitive computation overheads. CachedAttention maintains a hierarchical KV caching system that lev"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.19708","kind":"arxiv","version":3},"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/2403.19708/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":"2403.19708","created_at":"2026-07-05T08:38:07.931339+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.19708v3","created_at":"2026-07-05T08:38:07.931339+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.19708","created_at":"2026-07-05T08:38:07.931339+00:00"},{"alias_kind":"pith_short_12","alias_value":"L5WJP2YPS3T6","created_at":"2026-07-05T08:38:07.931339+00:00"},{"alias_kind":"pith_short_16","alias_value":"L5WJP2YPS3T67AUP","created_at":"2026-07-05T08:38:07.931339+00:00"},{"alias_kind":"pith_short_8","alias_value":"L5WJP2YP","created_at":"2026-07-05T08:38:07.931339+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08032","citing_title":"What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28565","citing_title":"KernelSight-LM: A Kernel-Level LLM Inference Simulator","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12556","citing_title":"ITME: Inference Tiered Memory Expansion with Disaggregated CXL-Hybrid Memories","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02964","citing_title":"Multi-Segment Attention: Enabling Efficient KV-Cache Management for Faster Large Language Model Serving","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28565","citing_title":"KernelSight-LM: A Kernel-Level LLM Inference Simulator","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02230","citing_title":"Continuum: Efficient and Robust Multi-Turn LLM Agent Scheduling with KV Cache Time-to-Live","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03143","citing_title":"TokenDance: Scaling Multi-Agent LLM Serving via Collective KV Cache Sharing","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11232","citing_title":"Rethinking LLMOps for Fraud and AML: Building a Compliance-Grade LLM Serving Stack","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03375","citing_title":"Tutti: Making SSD-Backed KV Cache Practical for Long-Context LLM Serving","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17353","citing_title":"Hive: A Multi-Agent Infrastructure for Algorithm- and Task-Level Scaling","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G","json":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G.json","graph_json":"https://pith.science/api/pith-number/L5WJP2YPS3T67AUPDYJN6GDT2G/graph.json","events_json":"https://pith.science/api/pith-number/L5WJP2YPS3T67AUPDYJN6GDT2G/events.json","paper":"https://pith.science/paper/L5WJP2YP"},"agent_actions":{"view_html":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G","download_json":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G.json","view_paper":"https://pith.science/paper/L5WJP2YP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.19708&json=true","fetch_graph":"https://pith.science/api/pith-number/L5WJP2YPS3T67AUPDYJN6GDT2G/graph.json","fetch_events":"https://pith.science/api/pith-number/L5WJP2YPS3T67AUPDYJN6GDT2G/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G/action/timestamp_anchor","attest_storage":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G/action/storage_attestation","attest_author":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G/action/author_attestation","sign_citation":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G/action/citation_signature","submit_replication":"https://pith.science/pith/L5WJP2YPS3T67AUPDYJN6GDT2G/action/replication_record"}},"created_at":"2026-07-05T08:38:07.931339+00:00","updated_at":"2026-07-05T08:38:07.931339+00:00"}