{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BP5OP5KMDYX3BPFN42TJLVV7KF","short_pith_number":"pith:BP5OP5KM","schema_version":"1.0","canonical_sha256":"0bfae7f54c1e2fb0bcade6a695d6bf516df546dd9760ec4d1f9de07c9629ae03","source":{"kind":"arxiv","id":"2507.06608","version":5},"attestation_state":"computed","paper":{"title":"Nexus:Proactive Intra-GPU Disaggregation of Prefill and Decode in LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Colin Cai, Junjia Du, Xiaoxiang Shi, Zhihao Jia","submitted_at":"2025-07-09T07:27:18Z","abstract_excerpt":"Monolithic serving with chunked prefill improves GPU utilization by batching prefill and decode together, but suffers from fine-grained phase interference. Engine-level prefill-decode (PD) disaggregation avoids interference but incurs higher hardware and coordination overhead. Prior intra-GPU disaggregation approaches multiplex prefill and decode within a single GPU, using SLO-based tuning guided by heuristics from offline profiling or reactive feedback loops. However, these methods respond reactively to performance issues rather than anticipating them, limiting adaptability under dynamic work"},"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":"2507.06608","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-07-09T07:27:18Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"1b80c5f03326e0b7c9fda5093a85a925b777fb4f38de62935a96833db011b3f5","abstract_canon_sha256":"a0a7658b7222745d17228df0e7eef9837c5d1fa8571bc8c49cdb79aa6c141ae7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:49:56.738059Z","signature_b64":"6DW2Gi3h8272vv536CGuAW1vvmRrrzMRhyOc8l/Nd/+6bPWmswwnCBMM4k/BDx0A/k7PmuS9BBuDPl1uyVfDAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0bfae7f54c1e2fb0bcade6a695d6bf516df546dd9760ec4d1f9de07c9629ae03","last_reissued_at":"2026-07-05T11:49:56.737545Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:49:56.737545Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Nexus:Proactive Intra-GPU Disaggregation of Prefill and Decode in LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Colin Cai, Junjia Du, Xiaoxiang Shi, Zhihao Jia","submitted_at":"2025-07-09T07:27:18Z","abstract_excerpt":"Monolithic serving with chunked prefill improves GPU utilization by batching prefill and decode together, but suffers from fine-grained phase interference. Engine-level prefill-decode (PD) disaggregation avoids interference but incurs higher hardware and coordination overhead. Prior intra-GPU disaggregation approaches multiplex prefill and decode within a single GPU, using SLO-based tuning guided by heuristics from offline profiling or reactive feedback loops. However, these methods respond reactively to performance issues rather than anticipating them, limiting adaptability under dynamic work"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.06608","kind":"arxiv","version":5},"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/2507.06608/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":"2507.06608","created_at":"2026-07-05T11:49:56.737611+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.06608v5","created_at":"2026-07-05T11:49:56.737611+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.06608","created_at":"2026-07-05T11:49:56.737611+00:00"},{"alias_kind":"pith_short_12","alias_value":"BP5OP5KMDYX3","created_at":"2026-07-05T11:49:56.737611+00:00"},{"alias_kind":"pith_short_16","alias_value":"BP5OP5KMDYX3BPFN","created_at":"2026-07-05T11:49:56.737611+00:00"},{"alias_kind":"pith_short_8","alias_value":"BP5OP5KM","created_at":"2026-07-05T11:49:56.737611+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.02043","citing_title":"Towards Load-Aware Prefill Deflection for Disaggregated LLM Serving","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04415","citing_title":"FlexNPU: Transparent NPU Virtualization for Dynamic LLM Prefill-Decode Co-location","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF","json":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF.json","graph_json":"https://pith.science/api/pith-number/BP5OP5KMDYX3BPFN42TJLVV7KF/graph.json","events_json":"https://pith.science/api/pith-number/BP5OP5KMDYX3BPFN42TJLVV7KF/events.json","paper":"https://pith.science/paper/BP5OP5KM"},"agent_actions":{"view_html":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF","download_json":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF.json","view_paper":"https://pith.science/paper/BP5OP5KM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.06608&json=true","fetch_graph":"https://pith.science/api/pith-number/BP5OP5KMDYX3BPFN42TJLVV7KF/graph.json","fetch_events":"https://pith.science/api/pith-number/BP5OP5KMDYX3BPFN42TJLVV7KF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF/action/storage_attestation","attest_author":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF/action/author_attestation","sign_citation":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF/action/citation_signature","submit_replication":"https://pith.science/pith/BP5OP5KMDYX3BPFN42TJLVV7KF/action/replication_record"}},"created_at":"2026-07-05T11:49:56.737611+00:00","updated_at":"2026-07-05T11:49:56.737611+00:00"}