{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EL6A4QOTPLYS3LFEIZGOL5QVTM","short_pith_number":"pith:EL6A4QOT","schema_version":"1.0","canonical_sha256":"22fc0e41d37af12daca4464ce5f6159b3b7926b56310533ae0f6d3b869338159","source":{"kind":"arxiv","id":"2606.26666","version":1},"attestation_state":"computed","paper":{"title":"PersistentKV: Page-Aware Decode Scheduling for Long-Context LLM Serving on Commodity GPUs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Muhammad Ahmed","submitted_at":"2026-06-25T06:56:43Z","abstract_excerpt":"Autoregressive large language model (LLM) serving is increasingly limited by key-value (KV) cache movement rather than dense matrix multiplication. Modern paged-attention systems reduce KV-cache fragmentation and mature kernels such as FlashInfer provide highly optimized native-paged decode attention. However, the best single-kernel implementation is not always the best serving schedule: low-active long-context decode can under-utilize commodity GPUs, while mixed sequence lengths introduce a tension between many exact-length launches and coarse padded batches. We present PersistentKV, a native"},"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":"2606.26666","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-25T06:56:43Z","cross_cats_sorted":[],"title_canon_sha256":"9d8c38979e0d02daad360388d7246e3f51d9bd088c2a886a7ffc1d58b95f9116","abstract_canon_sha256":"6586281ebd085d9f0661bdf4e48c8899bb25db46360150cc184174c62866661d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-26T01:15:56.296492Z","signature_b64":"iURSKD0bjYIRaZXYb6MLzIBS7G/8Gukz0Vl5XfHevf6Db05tBVD3j39rv6Sf7MNoDdbz5KOIBHI7F7UW13npBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"22fc0e41d37af12daca4464ce5f6159b3b7926b56310533ae0f6d3b869338159","last_reissued_at":"2026-06-26T01:15:56.296077Z","signature_status":"signed_v1","first_computed_at":"2026-06-26T01:15:56.296077Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PersistentKV: Page-Aware Decode Scheduling for Long-Context LLM Serving on Commodity GPUs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Muhammad Ahmed","submitted_at":"2026-06-25T06:56:43Z","abstract_excerpt":"Autoregressive large language model (LLM) serving is increasingly limited by key-value (KV) cache movement rather than dense matrix multiplication. Modern paged-attention systems reduce KV-cache fragmentation and mature kernels such as FlashInfer provide highly optimized native-paged decode attention. However, the best single-kernel implementation is not always the best serving schedule: low-active long-context decode can under-utilize commodity GPUs, while mixed sequence lengths introduce a tension between many exact-length launches and coarse padded batches. We present PersistentKV, a native"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.26666","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/2606.26666/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":"2606.26666","created_at":"2026-06-26T01:15:56.296134+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.26666v1","created_at":"2026-06-26T01:15:56.296134+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.26666","created_at":"2026-06-26T01:15:56.296134+00:00"},{"alias_kind":"pith_short_12","alias_value":"EL6A4QOTPLYS","created_at":"2026-06-26T01:15:56.296134+00:00"},{"alias_kind":"pith_short_16","alias_value":"EL6A4QOTPLYS3LFE","created_at":"2026-06-26T01:15:56.296134+00:00"},{"alias_kind":"pith_short_8","alias_value":"EL6A4QOT","created_at":"2026-06-26T01:15:56.296134+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/EL6A4QOTPLYS3LFEIZGOL5QVTM","json":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM.json","graph_json":"https://pith.science/api/pith-number/EL6A4QOTPLYS3LFEIZGOL5QVTM/graph.json","events_json":"https://pith.science/api/pith-number/EL6A4QOTPLYS3LFEIZGOL5QVTM/events.json","paper":"https://pith.science/paper/EL6A4QOT"},"agent_actions":{"view_html":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM","download_json":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM.json","view_paper":"https://pith.science/paper/EL6A4QOT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.26666&json=true","fetch_graph":"https://pith.science/api/pith-number/EL6A4QOTPLYS3LFEIZGOL5QVTM/graph.json","fetch_events":"https://pith.science/api/pith-number/EL6A4QOTPLYS3LFEIZGOL5QVTM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM/action/storage_attestation","attest_author":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM/action/author_attestation","sign_citation":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM/action/citation_signature","submit_replication":"https://pith.science/pith/EL6A4QOTPLYS3LFEIZGOL5QVTM/action/replication_record"}},"created_at":"2026-06-26T01:15:56.296134+00:00","updated_at":"2026-06-26T01:15:56.296134+00:00"}