{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OSEMS3TZORLVJLMGUZRHSCNFI2","short_pith_number":"pith:OSEMS3TZ","schema_version":"1.0","canonical_sha256":"7488c96e79745754ad86a6627909a54683e1c9492d29204fd8335593f731fcab","source":{"kind":"arxiv","id":"2504.11320","version":3},"attestation_state":"computed","paper":{"title":"Optimizing LLM Inference: Fluid-Guided Online Scheduling with Memory Constraints","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.DC","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"David Simchi-Levi, Gan Luo, Ruicheng Ao, Xinshang Wang","submitted_at":"2025-04-15T16:00:21Z","abstract_excerpt":"Large language models now serve millions of users daily, with providers incurring costs exceeding $700,000 per day. Each request requires token-by-token inference, making GPU scheduling central to latency, capacity, and cost. The difficulty is endogenous memory growth: generated tokens expand the Key-Value (KV) cache, and overflow can evict in-progress requests and waste prior computation. We formulate inference as a multi-stage online scheduling problem with endogenous memory growth, linear iteration times, and GPU-resident KV-cache constraints. We introduce a fluid model that characterizes e"},"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":"2504.11320","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-15T16:00:21Z","cross_cats_sorted":["cs.AI","cs.DC","math.OC","stat.ML"],"title_canon_sha256":"d9595304d1dda16053f3ee96a4ab3a401fc28fc848c753d9819b5d0626132ef1","abstract_canon_sha256":"b0ac48d841165a30158effd629097aa7befbb1cce0b24754c4d8b1287306514f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:00:19.288648Z","signature_b64":"oPlEMD8keRnLB0Os7aoMhR7GNmtlRXVMY2vMctyoVtq1w2tqm1qjvV6sq8NVsQx7ZqGtH+ZgEhb8ZD7gk6lKCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7488c96e79745754ad86a6627909a54683e1c9492d29204fd8335593f731fcab","last_reissued_at":"2026-05-20T00:00:19.287871Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:00:19.287871Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing LLM Inference: Fluid-Guided Online Scheduling with Memory Constraints","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.DC","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"David Simchi-Levi, Gan Luo, Ruicheng Ao, Xinshang Wang","submitted_at":"2025-04-15T16:00:21Z","abstract_excerpt":"Large language models now serve millions of users daily, with providers incurring costs exceeding $700,000 per day. Each request requires token-by-token inference, making GPU scheduling central to latency, capacity, and cost. The difficulty is endogenous memory growth: generated tokens expand the Key-Value (KV) cache, and overflow can evict in-progress requests and waste prior computation. We formulate inference as a multi-stage online scheduling problem with endogenous memory growth, linear iteration times, and GPU-resident KV-cache constraints. We introduce a fluid model that characterizes e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.11320","kind":"arxiv","version":3},"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/2504.11320/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":"2504.11320","created_at":"2026-05-20T00:00:19.288010+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.11320v3","created_at":"2026-05-20T00:00:19.288010+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.11320","created_at":"2026-05-20T00:00:19.288010+00:00"},{"alias_kind":"pith_short_12","alias_value":"OSEMS3TZORLV","created_at":"2026-05-20T00:00:19.288010+00:00"},{"alias_kind":"pith_short_16","alias_value":"OSEMS3TZORLVJLMG","created_at":"2026-05-20T00:00:19.288010+00:00"},{"alias_kind":"pith_short_8","alias_value":"OSEMS3TZ","created_at":"2026-05-20T00:00:19.288010+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2601.20309","citing_title":"SuperInfer: SLO-Aware Rotary Scheduling and Memory Management for LLM Inference on Superchips","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":48,"is_internal_anchor":true},{"citing_arxiv_id":"2605.04595","citing_title":"A Queueing-Theoretic Framework for Stability Analysis of LLM Inference with KV Cache Memory Constraints","ref_index":22,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01280","citing_title":"Position: LLM Serving Needs Mathematical Optimization and Algorithmic Foundations, Not Just Heuristics","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2604.11001","citing_title":"Flow-Controlled Scheduling for LLM Inference with Provable Stability Guarantees","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":48,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2","json":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2.json","graph_json":"https://pith.science/api/pith-number/OSEMS3TZORLVJLMGUZRHSCNFI2/graph.json","events_json":"https://pith.science/api/pith-number/OSEMS3TZORLVJLMGUZRHSCNFI2/events.json","paper":"https://pith.science/paper/OSEMS3TZ"},"agent_actions":{"view_html":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2","download_json":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2.json","view_paper":"https://pith.science/paper/OSEMS3TZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.11320&json=true","fetch_graph":"https://pith.science/api/pith-number/OSEMS3TZORLVJLMGUZRHSCNFI2/graph.json","fetch_events":"https://pith.science/api/pith-number/OSEMS3TZORLVJLMGUZRHSCNFI2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2/action/storage_attestation","attest_author":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2/action/author_attestation","sign_citation":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2/action/citation_signature","submit_replication":"https://pith.science/pith/OSEMS3TZORLVJLMGUZRHSCNFI2/action/replication_record"}},"created_at":"2026-05-20T00:00:19.288010+00:00","updated_at":"2026-05-20T00:00:19.288010+00:00"}