{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UCOEM6P55VWBU5OUOMMNNA4UQF","short_pith_number":"pith:UCOEM6P5","schema_version":"1.0","canonical_sha256":"a09c4679fded6c1a75d47318d68394816f945115e1dc91908107eecb5d013c59","source":{"kind":"arxiv","id":"2605.18071","version":1},"attestation_state":"computed","paper":{"title":"KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Haodong Wang, Haodyue Zhang, Jian Lin, Jiazhi Mi, Peng Li, Qianli Liu, Song Guo, Zicong Hong","submitted_at":"2026-05-18T08:54:16Z","abstract_excerpt":"Supporting long-context LLMs is challenging due to the substantial memory demands of the key-value (KV) cache. Existing offloading systems store the full cache in host memory and selectively fetch critical entries during decoding, but this strategy quickly hits a ceiling: sparsity cannot be pushed further without degrading accuracy. As a result, when context length and batch size grow, the volume of KV transfers rises sharply and becomes the dominant source of decoding latency. We present KVDrive, a holistic multi-tier KV cache management system spanning GPU memory, host DRAM, and SSD. Unlike "},"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":"2605.18071","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-18T08:54:16Z","cross_cats_sorted":[],"title_canon_sha256":"eff93fb061dcce5910a83a50684edde44b74d747095f1ff5a9f91c1acf2c026c","abstract_canon_sha256":"e259d054ab39735960ca254de26aaf0f58cd2a9c032133067e07691b00e3146e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:14.516495Z","signature_b64":"UDTfEVcGNTGaYGhNFaoqgNWRAy3RXA65cTSrlJ9R4iBtP6YD1+x8UuP39ID8FCesh9N8qKidZbfKzcbTkj2pBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a09c4679fded6c1a75d47318d68394816f945115e1dc91908107eecb5d013c59","last_reissued_at":"2026-05-20T00:05:14.515674Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:14.515674Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Haodong Wang, Haodyue Zhang, Jian Lin, Jiazhi Mi, Peng Li, Qianli Liu, Song Guo, Zicong Hong","submitted_at":"2026-05-18T08:54:16Z","abstract_excerpt":"Supporting long-context LLMs is challenging due to the substantial memory demands of the key-value (KV) cache. Existing offloading systems store the full cache in host memory and selectively fetch critical entries during decoding, but this strategy quickly hits a ceiling: sparsity cannot be pushed further without degrading accuracy. As a result, when context length and batch size grow, the volume of KV transfers rises sharply and becomes the dominant source of decoding latency. We present KVDrive, a holistic multi-tier KV cache management system spanning GPU memory, host DRAM, and SSD. Unlike "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18071","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/2605.18071/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-19T23:41:59.248523Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T23:33:35.461161Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"a3520bb5392c3224fe3482e1d2830c3cde1855c2de1974ff94f68391aac87add"},"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":"2605.18071","created_at":"2026-05-20T00:05:14.515817+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.18071v1","created_at":"2026-05-20T00:05:14.515817+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18071","created_at":"2026-05-20T00:05:14.515817+00:00"},{"alias_kind":"pith_short_12","alias_value":"UCOEM6P55VWB","created_at":"2026-05-20T00:05:14.515817+00:00"},{"alias_kind":"pith_short_16","alias_value":"UCOEM6P55VWBU5OU","created_at":"2026-05-20T00:05:14.515817+00:00"},{"alias_kind":"pith_short_8","alias_value":"UCOEM6P5","created_at":"2026-05-20T00:05:14.515817+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/UCOEM6P55VWBU5OUOMMNNA4UQF","json":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF.json","graph_json":"https://pith.science/api/pith-number/UCOEM6P55VWBU5OUOMMNNA4UQF/graph.json","events_json":"https://pith.science/api/pith-number/UCOEM6P55VWBU5OUOMMNNA4UQF/events.json","paper":"https://pith.science/paper/UCOEM6P5"},"agent_actions":{"view_html":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF","download_json":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF.json","view_paper":"https://pith.science/paper/UCOEM6P5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.18071&json=true","fetch_graph":"https://pith.science/api/pith-number/UCOEM6P55VWBU5OUOMMNNA4UQF/graph.json","fetch_events":"https://pith.science/api/pith-number/UCOEM6P55VWBU5OUOMMNNA4UQF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF/action/storage_attestation","attest_author":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF/action/author_attestation","sign_citation":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF/action/citation_signature","submit_replication":"https://pith.science/pith/UCOEM6P55VWBU5OUOMMNNA4UQF/action/replication_record"}},"created_at":"2026-05-20T00:05:14.515817+00:00","updated_at":"2026-05-20T00:05:14.515817+00:00"}