{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:D43WT376NFJ6FYAJUONUGVGFFS","short_pith_number":"pith:D43WT376","schema_version":"1.0","canonical_sha256":"1f3769effe6953e2e009a39b4354c52c9cb81291e3ec2b990c6b686ac45e104e","source":{"kind":"arxiv","id":"2606.08635","version":1},"attestation_state":"computed","paper":{"title":"SpectrumKV: Per-Token Mixed-Precision KV Cache Transfer for Prefill-Decode Disaggregated LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Yang Pengju","submitted_at":"2026-06-07T13:57:05Z","abstract_excerpt":"Prefill-decode (PD) disaggregation decouples prompt processing from token generation, but it also turns the key-value (KV) cache into a network payload. Existing PD-side KV reduction methods are mostly binary: selected tokens are transmitted at full precision and the rest are not transmitted. This paper argues that binary selection leaves a useful design space unused. SpectrumKV assigns a precision level to each token instead: attention sinks and other high-importance tokens are protected at FP16, medium-importance tokens are sent at INT8, and low-importance tokens are sent at INT4 when the mo"},"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":"2606.08635","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-07T13:57:05Z","cross_cats_sorted":["cs.DC"],"title_canon_sha256":"3288c46811f2ad77f7df8cb1dab8b54f8ce9e6b61a346e7dcaa2c102f15a68a2","abstract_canon_sha256":"f6be326f0dab2d110fd0f03799687e49a1b38010051892af0ee83b10be61d38b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:05:42.065838Z","signature_b64":"0xazXoolUPuO5VEdFaCw7UihH/6WZxGeZPeEHBA9SU/PM5rwnVBx6sqaruDaEqA3QviMZerdE4dbN+0XdbD/Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1f3769effe6953e2e009a39b4354c52c9cb81291e3ec2b990c6b686ac45e104e","last_reissued_at":"2026-06-09T01:05:42.065443Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:05:42.065443Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpectrumKV: Per-Token Mixed-Precision KV Cache Transfer for Prefill-Decode Disaggregated LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.LG","authors_text":"Yang Pengju","submitted_at":"2026-06-07T13:57:05Z","abstract_excerpt":"Prefill-decode (PD) disaggregation decouples prompt processing from token generation, but it also turns the key-value (KV) cache into a network payload. Existing PD-side KV reduction methods are mostly binary: selected tokens are transmitted at full precision and the rest are not transmitted. This paper argues that binary selection leaves a useful design space unused. SpectrumKV assigns a precision level to each token instead: attention sinks and other high-importance tokens are protected at FP16, medium-importance tokens are sent at INT8, and low-importance tokens are sent at INT4 when the mo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.08635","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/2606.08635/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":"2606.08635","created_at":"2026-06-09T01:05:42.065506+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.08635v1","created_at":"2026-06-09T01:05:42.065506+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.08635","created_at":"2026-06-09T01:05:42.065506+00:00"},{"alias_kind":"pith_short_12","alias_value":"D43WT376NFJ6","created_at":"2026-06-09T01:05:42.065506+00:00"},{"alias_kind":"pith_short_16","alias_value":"D43WT376NFJ6FYAJ","created_at":"2026-06-09T01:05:42.065506+00:00"},{"alias_kind":"pith_short_8","alias_value":"D43WT376","created_at":"2026-06-09T01:05:42.065506+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/D43WT376NFJ6FYAJUONUGVGFFS","json":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS.json","graph_json":"https://pith.science/api/pith-number/D43WT376NFJ6FYAJUONUGVGFFS/graph.json","events_json":"https://pith.science/api/pith-number/D43WT376NFJ6FYAJUONUGVGFFS/events.json","paper":"https://pith.science/paper/D43WT376"},"agent_actions":{"view_html":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS","download_json":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS.json","view_paper":"https://pith.science/paper/D43WT376","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.08635&json=true","fetch_graph":"https://pith.science/api/pith-number/D43WT376NFJ6FYAJUONUGVGFFS/graph.json","fetch_events":"https://pith.science/api/pith-number/D43WT376NFJ6FYAJUONUGVGFFS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS/action/storage_attestation","attest_author":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS/action/author_attestation","sign_citation":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS/action/citation_signature","submit_replication":"https://pith.science/pith/D43WT376NFJ6FYAJUONUGVGFFS/action/replication_record"}},"created_at":"2026-06-09T01:05:42.065506+00:00","updated_at":"2026-06-09T01:05:42.065506+00:00"}