{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2XBLZNJP6WCTKEOQKZWODG333U","short_pith_number":"pith:2XBLZNJP","schema_version":"1.0","canonical_sha256":"d5c2bcb52ff5853511d0566ce19b7bdd3315db7b76625a95a6a46b23586b7696","source":{"kind":"arxiv","id":"2410.05004","version":1},"attestation_state":"computed","paper":{"title":"Fast State Restoration in LLM Serving with HCache","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Jiwu Shu, Shiwei Gao, Youmin Chen","submitted_at":"2024-10-07T13:03:45Z","abstract_excerpt":"The growing complexity of LLM usage today, e.g., multi-round conversation and retrieval-augmented generation (RAG), makes contextual states (i.e., KV cache) reusable across user requests. Given the capacity constraints of GPU memory, only a limited number of contexts can be cached on GPU for reusing. Existing inference systems typically evict part of the KV cache and restore it by recomputing it from the original tokens or offloading it to host storage for later retrieval, both of which introduce substantial computational or I/O overheads. We propose HCache, a novel LLM state restoration metho"},"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":"2410.05004","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.DC","submitted_at":"2024-10-07T13:03:45Z","cross_cats_sorted":[],"title_canon_sha256":"027a28004c45ba8df47894a1901b9b0f5f4bf4bd840065358285ea865f915e8f","abstract_canon_sha256":"0d9805ece667add639ef598377b05b3bfc73bc2dd9e7ab87c8b68edde063171e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:17:00.033528Z","signature_b64":"L1fc0ctGoSC3yroWpGPxmxjbSWZZlFtg9Yq89Maj+l90FVi6etOaOJW1sQut/T5b3sPfyzNPpMS7yzV/qCqLCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d5c2bcb52ff5853511d0566ce19b7bdd3315db7b76625a95a6a46b23586b7696","last_reissued_at":"2026-07-05T09:17:00.033073Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:17:00.033073Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fast State Restoration in LLM Serving with HCache","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Jiwu Shu, Shiwei Gao, Youmin Chen","submitted_at":"2024-10-07T13:03:45Z","abstract_excerpt":"The growing complexity of LLM usage today, e.g., multi-round conversation and retrieval-augmented generation (RAG), makes contextual states (i.e., KV cache) reusable across user requests. Given the capacity constraints of GPU memory, only a limited number of contexts can be cached on GPU for reusing. Existing inference systems typically evict part of the KV cache and restore it by recomputing it from the original tokens or offloading it to host storage for later retrieval, both of which introduce substantial computational or I/O overheads. We propose HCache, a novel LLM state restoration metho"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.05004","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/2410.05004/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":"2410.05004","created_at":"2026-07-05T09:17:00.033133+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.05004v1","created_at":"2026-07-05T09:17:00.033133+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.05004","created_at":"2026-07-05T09:17:00.033133+00:00"},{"alias_kind":"pith_short_12","alias_value":"2XBLZNJP6WCT","created_at":"2026-07-05T09:17:00.033133+00:00"},{"alias_kind":"pith_short_16","alias_value":"2XBLZNJP6WCTKEOQ","created_at":"2026-07-05T09:17:00.033133+00:00"},{"alias_kind":"pith_short_8","alias_value":"2XBLZNJP","created_at":"2026-07-05T09:17:00.033133+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2504.15965","citing_title":"From Human Memory to AI Memory: A Survey on Memory Mechanisms in the Era of LLMs","ref_index":134,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U","json":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U.json","graph_json":"https://pith.science/api/pith-number/2XBLZNJP6WCTKEOQKZWODG333U/graph.json","events_json":"https://pith.science/api/pith-number/2XBLZNJP6WCTKEOQKZWODG333U/events.json","paper":"https://pith.science/paper/2XBLZNJP"},"agent_actions":{"view_html":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U","download_json":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U.json","view_paper":"https://pith.science/paper/2XBLZNJP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.05004&json=true","fetch_graph":"https://pith.science/api/pith-number/2XBLZNJP6WCTKEOQKZWODG333U/graph.json","fetch_events":"https://pith.science/api/pith-number/2XBLZNJP6WCTKEOQKZWODG333U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U/action/storage_attestation","attest_author":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U/action/author_attestation","sign_citation":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U/action/citation_signature","submit_replication":"https://pith.science/pith/2XBLZNJP6WCTKEOQKZWODG333U/action/replication_record"}},"created_at":"2026-07-05T09:17:00.033133+00:00","updated_at":"2026-07-05T09:17:00.033133+00:00"}