{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:EJJXK7XL3N7M7GDWBDZ7MLPYDM","short_pith_number":"pith:EJJXK7XL","schema_version":"1.0","canonical_sha256":"2253757eebdb7ecf987608f3f62df81b0a27fa4fb3f63a94c225875b5454ed31","source":{"kind":"arxiv","id":"2508.06526","version":3},"attestation_state":"computed","paper":{"title":"PiKV: KV Cache Management System for Mixture of Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.DC","authors_text":"Ben Lengerich, Dong Liu, Yanxuan Yu, Ying Nian Wu","submitted_at":"2025-08-02T03:50:14Z","abstract_excerpt":"As large-scale language models continue to scale up in both size and context length, the memory and communication cost of key-value (KV) cache storage has become a major bottleneck in multi-GPU and multi-node inference. While MoE-based architectures sparsify computation across experts, the corresponding KV caches remain dense and globally synchronized, resulting in significant overhead.\n  We introduce \\textbf{PiKV}, a parallel and distributed KV cache serving framework tailored for MoE architecture. PiKV leverages \\textit{expert-sharded KV storage} to partition caches across GPUs, \\textit{PiKV"},"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":"2508.06526","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2025-08-02T03:50:14Z","cross_cats_sorted":["cs.AI","cs.AR"],"title_canon_sha256":"da9bfb659a9835ada44123f57de71515b791b47b4617e3929d60cf24afa57bd7","abstract_canon_sha256":"5ca7ca2ff300b21e13c5575d68cf85e2a674b040b513885ff8a2d3279ee9202a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:34.663633Z","signature_b64":"zayKa/WxjLF2xA6N0oy+XY/byOt07EVordI03oN/oFOMIWJGhhmh3EARhHMC/EcqE08aGosvNz7BRMslZSg9AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2253757eebdb7ecf987608f3f62df81b0a27fa4fb3f63a94c225875b5454ed31","last_reissued_at":"2026-05-20T02:05:34.662750Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:34.662750Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PiKV: KV Cache Management System for Mixture of Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.DC","authors_text":"Ben Lengerich, Dong Liu, Yanxuan Yu, Ying Nian Wu","submitted_at":"2025-08-02T03:50:14Z","abstract_excerpt":"As large-scale language models continue to scale up in both size and context length, the memory and communication cost of key-value (KV) cache storage has become a major bottleneck in multi-GPU and multi-node inference. While MoE-based architectures sparsify computation across experts, the corresponding KV caches remain dense and globally synchronized, resulting in significant overhead.\n  We introduce \\textbf{PiKV}, a parallel and distributed KV cache serving framework tailored for MoE architecture. PiKV leverages \\textit{expert-sharded KV storage} to partition caches across GPUs, \\textit{PiKV"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.06526","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/2508.06526/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":"2508.06526","created_at":"2026-05-20T02:05:34.662889+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.06526v3","created_at":"2026-05-20T02:05:34.662889+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.06526","created_at":"2026-05-20T02:05:34.662889+00:00"},{"alias_kind":"pith_short_12","alias_value":"EJJXK7XL3N7M","created_at":"2026-05-20T02:05:34.662889+00:00"},{"alias_kind":"pith_short_16","alias_value":"EJJXK7XL3N7M7GDW","created_at":"2026-05-20T02:05:34.662889+00:00"},{"alias_kind":"pith_short_8","alias_value":"EJJXK7XL","created_at":"2026-05-20T02:05:34.662889+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/EJJXK7XL3N7M7GDWBDZ7MLPYDM","json":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM.json","graph_json":"https://pith.science/api/pith-number/EJJXK7XL3N7M7GDWBDZ7MLPYDM/graph.json","events_json":"https://pith.science/api/pith-number/EJJXK7XL3N7M7GDWBDZ7MLPYDM/events.json","paper":"https://pith.science/paper/EJJXK7XL"},"agent_actions":{"view_html":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM","download_json":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM.json","view_paper":"https://pith.science/paper/EJJXK7XL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.06526&json=true","fetch_graph":"https://pith.science/api/pith-number/EJJXK7XL3N7M7GDWBDZ7MLPYDM/graph.json","fetch_events":"https://pith.science/api/pith-number/EJJXK7XL3N7M7GDWBDZ7MLPYDM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM/action/storage_attestation","attest_author":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM/action/author_attestation","sign_citation":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM/action/citation_signature","submit_replication":"https://pith.science/pith/EJJXK7XL3N7M7GDWBDZ7MLPYDM/action/replication_record"}},"created_at":"2026-05-20T02:05:34.662889+00:00","updated_at":"2026-05-20T02:05:34.662889+00:00"}