{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:7ADWNATDK6GTU7MRDOHH4OCCCO","short_pith_number":"pith:7ADWNATD","schema_version":"1.0","canonical_sha256":"f807668263578d3a7d911b8e7e3842139912d36330cd452e765c9ac3fd5f379e","source":{"kind":"arxiv","id":"2508.18736","version":1},"attestation_state":"computed","paper":{"title":"Rethinking Caching for LLM Serving Systems: Beyond Traditional Heuristics","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DB","authors_text":"Chanwoo Moon, Heejin Kim, Jaeheon Lee, Jungwoo Kim, Minsang Kim, Sungjin Lee, Taeho Hwang, Woosuk Chung, Yeseong Kim","submitted_at":"2025-08-26T07:09:09Z","abstract_excerpt":"Serving Large Language Models (LLMs) at scale requires meeting strict Service Level Objectives (SLOs) under severe computational and memory constraints. Nevertheless, traditional caching strategies fall short: exact-matching and prefix caches neglect query semantics, while state-of-the-art semantic caches remain confined to traditional intuitions, offering little conceptual departure. Building on this, we present SISO, a semantic caching system that redefines efficiency for LLM serving. SISO introduces centroid-based caching to maximize coverage with minimal memory, locality-aware replacement "},"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":"2508.18736","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DB","submitted_at":"2025-08-26T07:09:09Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"fead999f3a21bc2b57b1538030ff77681ac13cd31cd64ea1c1d7461e629ad2b7","abstract_canon_sha256":"6e675c77a7019f17e61a86ffde96bcb87010257c6ccc88e5d1044225ffc87fb4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:59:29.915399Z","signature_b64":"NAlcvoGn+9vv+U71YY5pkEC8ALireEY5AnKyuZJ4DuPnQlc1v9WFFX1FAMjdAHeH8lbhOOGT+2ubVWMp2R8TAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f807668263578d3a7d911b8e7e3842139912d36330cd452e765c9ac3fd5f379e","last_reissued_at":"2026-07-05T11:59:29.914943Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:59:29.914943Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking Caching for LLM Serving Systems: Beyond Traditional Heuristics","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DB","authors_text":"Chanwoo Moon, Heejin Kim, Jaeheon Lee, Jungwoo Kim, Minsang Kim, Sungjin Lee, Taeho Hwang, Woosuk Chung, Yeseong Kim","submitted_at":"2025-08-26T07:09:09Z","abstract_excerpt":"Serving Large Language Models (LLMs) at scale requires meeting strict Service Level Objectives (SLOs) under severe computational and memory constraints. Nevertheless, traditional caching strategies fall short: exact-matching and prefix caches neglect query semantics, while state-of-the-art semantic caches remain confined to traditional intuitions, offering little conceptual departure. Building on this, we present SISO, a semantic caching system that redefines efficiency for LLM serving. SISO introduces centroid-based caching to maximize coverage with minimal memory, locality-aware replacement "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.18736","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/2508.18736/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":"2508.18736","created_at":"2026-07-05T11:59:29.915000+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.18736v1","created_at":"2026-07-05T11:59:29.915000+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.18736","created_at":"2026-07-05T11:59:29.915000+00:00"},{"alias_kind":"pith_short_12","alias_value":"7ADWNATDK6GT","created_at":"2026-07-05T11:59:29.915000+00:00"},{"alias_kind":"pith_short_16","alias_value":"7ADWNATDK6GTU7MR","created_at":"2026-07-05T11:59:29.915000+00:00"},{"alias_kind":"pith_short_8","alias_value":"7ADWNATD","created_at":"2026-07-05T11:59:29.915000+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/7ADWNATDK6GTU7MRDOHH4OCCCO","json":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO.json","graph_json":"https://pith.science/api/pith-number/7ADWNATDK6GTU7MRDOHH4OCCCO/graph.json","events_json":"https://pith.science/api/pith-number/7ADWNATDK6GTU7MRDOHH4OCCCO/events.json","paper":"https://pith.science/paper/7ADWNATD"},"agent_actions":{"view_html":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO","download_json":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO.json","view_paper":"https://pith.science/paper/7ADWNATD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.18736&json=true","fetch_graph":"https://pith.science/api/pith-number/7ADWNATDK6GTU7MRDOHH4OCCCO/graph.json","fetch_events":"https://pith.science/api/pith-number/7ADWNATDK6GTU7MRDOHH4OCCCO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO/action/storage_attestation","attest_author":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO/action/author_attestation","sign_citation":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO/action/citation_signature","submit_replication":"https://pith.science/pith/7ADWNATDK6GTU7MRDOHH4OCCCO/action/replication_record"}},"created_at":"2026-07-05T11:59:29.915000+00:00","updated_at":"2026-07-05T11:59:29.915000+00:00"}