{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2GKAZ4IYN7PMHOPQPZJZTKF72R","short_pith_number":"pith:2GKAZ4IY","schema_version":"1.0","canonical_sha256":"d1940cf1186fdec3b9f07e5399a8bfd46c243d059a9d53d8994bb98287c57d3c","source":{"kind":"arxiv","id":"2602.14516","version":2},"attestation_state":"computed","paper":{"title":"Efficient Multi-round LLM Inference over Disaggregated Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bin Cui, Eiko Yoneki, Fangcheng Fu, Penghao Zhao, Quanqing Xu, Wenhao He, Youhe Jiang","submitted_at":"2026-02-16T07:07:30Z","abstract_excerpt":"With the rapid evolution of Large Language Models (LLMs), multi-round workflows, such as autonomous agents and iterative retrieval, have become increasingly prevalent. However, this raises hurdles for serving LLMs under prefill-decode (PD) disaggregation, a widely adopted paradigm that separates the compute-bound prefill phase and memory-bound decode phase onto individual resources. Specifically, existing systems overlook the interleaved prefill-decode workload pattern in multi-round inference, leading to sub-optimal handling of the incremental prefill workloads and model deployment for the tw"},"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":"2602.14516","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2026-02-16T07:07:30Z","cross_cats_sorted":[],"title_canon_sha256":"1717358015268d85ec6b489458f3456103dc9c8b97b7fc65f0c4544875fd97dc","abstract_canon_sha256":"19f3f0199eab165084ece7d46dbefbfc84bc34c5b12e2ca14ee4549b200d577a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-22T01:22:20.526433Z","signature_b64":"ycIWOLCqalGyA6B1/lgT7lN0twiYIP8wZH9iGWlMCSFU47QSiw/LmiVA0RYjKaycD/IRwgtZNywi9JQbqwJjDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d1940cf1186fdec3b9f07e5399a8bfd46c243d059a9d53d8994bb98287c57d3c","last_reissued_at":"2026-07-22T01:22:20.525497Z","signature_status":"signed_v1","first_computed_at":"2026-07-22T01:22:20.525497Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Multi-round LLM Inference over Disaggregated Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Bin Cui, Eiko Yoneki, Fangcheng Fu, Penghao Zhao, Quanqing Xu, Wenhao He, Youhe Jiang","submitted_at":"2026-02-16T07:07:30Z","abstract_excerpt":"With the rapid evolution of Large Language Models (LLMs), multi-round workflows, such as autonomous agents and iterative retrieval, have become increasingly prevalent. However, this raises hurdles for serving LLMs under prefill-decode (PD) disaggregation, a widely adopted paradigm that separates the compute-bound prefill phase and memory-bound decode phase onto individual resources. Specifically, existing systems overlook the interleaved prefill-decode workload pattern in multi-round inference, leading to sub-optimal handling of the incremental prefill workloads and model deployment for the tw"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.14516","kind":"arxiv","version":2},"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/2602.14516/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":"2602.14516","created_at":"2026-07-22T01:22:20.525952+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.14516v2","created_at":"2026-07-22T01:22:20.525952+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.14516","created_at":"2026-07-22T01:22:20.525952+00:00"},{"alias_kind":"pith_short_12","alias_value":"2GKAZ4IYN7PM","created_at":"2026-07-22T01:22:20.525952+00:00"},{"alias_kind":"pith_short_16","alias_value":"2GKAZ4IYN7PMHOPQ","created_at":"2026-07-22T01:22:20.525952+00:00"},{"alias_kind":"pith_short_8","alias_value":"2GKAZ4IY","created_at":"2026-07-22T01:22:20.525952+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2606.19271","citing_title":"TurboServe: Serving Streaming Video Generation Efficiently and Economically","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2606.01839","citing_title":"Observation, Not Prediction: Conversation-Level Disaggregated Scheduling for Agentic Serving","ref_index":55,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16637","citing_title":"HexAGenT: Efficient Agentic LLM Serving via Workflow- and Heterogeneity-Aware Scheduling","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12555","citing_title":"DelAC: A Multi-agent Reinforcement Learning of Team-Symmetric Stochastic Games","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2604.07144","citing_title":"Autopoiesis: A Self-Evolving System Paradigm for LLM Serving Under Runtime Dynamics","ref_index":51,"is_internal_anchor":true},{"citing_arxiv_id":"2604.16682","citing_title":"KAIROS: Stateful, Context-Aware Power-Efficient Agentic Inference Serving","ref_index":19,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R","json":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R.json","graph_json":"https://pith.science/api/pith-number/2GKAZ4IYN7PMHOPQPZJZTKF72R/graph.json","events_json":"https://pith.science/api/pith-number/2GKAZ4IYN7PMHOPQPZJZTKF72R/events.json","paper":"https://pith.science/paper/2GKAZ4IY"},"agent_actions":{"view_html":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R","download_json":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R.json","view_paper":"https://pith.science/paper/2GKAZ4IY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.14516&json=true","fetch_graph":"https://pith.science/api/pith-number/2GKAZ4IYN7PMHOPQPZJZTKF72R/graph.json","fetch_events":"https://pith.science/api/pith-number/2GKAZ4IYN7PMHOPQPZJZTKF72R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R/action/storage_attestation","attest_author":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R/action/author_attestation","sign_citation":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R/action/citation_signature","submit_replication":"https://pith.science/pith/2GKAZ4IYN7PMHOPQPZJZTKF72R/action/replication_record"}},"created_at":"2026-07-22T01:22:20.525952+00:00","updated_at":"2026-07-22T01:22:20.525952+00:00"}