{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IPJOALITFO363UOPAO7KWW2VOO","short_pith_number":"pith:IPJOALIT","schema_version":"1.0","canonical_sha256":"43d2e02d132bb7edd1cf03beab5b5573924a814430867296749b8a90e765d5c9","source":{"kind":"arxiv","id":"2408.15792","version":1},"attestation_state":"computed","paper":{"title":"Efficient LLM Scheduling by Learning to Rank","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aurick Qiao, Hao Zhang, Ion Stoica, Runlong Su, Siqi Zhu, Yichao Fu","submitted_at":"2024-08-28T13:35:54Z","abstract_excerpt":"In Large Language Model (LLM) inference, the output length of an LLM request is typically regarded as not known a priori. Consequently, most LLM serving systems employ a simple First-come-first-serve (FCFS) scheduling strategy, leading to Head-Of-Line (HOL) blocking and reduced throughput and service quality. In this paper, we reexamine this assumption -- we show that, although predicting the exact generation length of each request is infeasible, it is possible to predict the relative ranks of output lengths in a batch of requests, using learning to rank. The ranking information offers valuabl"},"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":"2408.15792","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-08-28T13:35:54Z","cross_cats_sorted":[],"title_canon_sha256":"3a57e154c9304648baedc2dd2b3a2d8fdabaf9a9b9044816d1ef823429497bca","abstract_canon_sha256":"83ec0f2aa5b39e72760c494fb54466a9848444a41a35753f1dfa29f834a630dd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:00:17.568018Z","signature_b64":"sFoaz5lpCXsstE7L2BaICS1bzTZ7E8FhQbw/07fs8b+cf8QHn3Oy64GMfVJsRqhh9KqPlAHSXgJRZe+EtOUzAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"43d2e02d132bb7edd1cf03beab5b5573924a814430867296749b8a90e765d5c9","last_reissued_at":"2026-07-05T09:00:17.567506Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:00:17.567506Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient LLM Scheduling by Learning to Rank","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aurick Qiao, Hao Zhang, Ion Stoica, Runlong Su, Siqi Zhu, Yichao Fu","submitted_at":"2024-08-28T13:35:54Z","abstract_excerpt":"In Large Language Model (LLM) inference, the output length of an LLM request is typically regarded as not known a priori. Consequently, most LLM serving systems employ a simple First-come-first-serve (FCFS) scheduling strategy, leading to Head-Of-Line (HOL) blocking and reduced throughput and service quality. In this paper, we reexamine this assumption -- we show that, although predicting the exact generation length of each request is infeasible, it is possible to predict the relative ranks of output lengths in a batch of requests, using learning to rank. The ranking information offers valuabl"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.15792","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/2408.15792/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":"2408.15792","created_at":"2026-07-05T09:00:17.567566+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.15792v1","created_at":"2026-07-05T09:00:17.567566+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.15792","created_at":"2026-07-05T09:00:17.567566+00:00"},{"alias_kind":"pith_short_12","alias_value":"IPJOALITFO36","created_at":"2026-07-05T09:00:17.567566+00:00"},{"alias_kind":"pith_short_16","alias_value":"IPJOALITFO363UOP","created_at":"2026-07-05T09:00:17.567566+00:00"},{"alias_kind":"pith_short_8","alias_value":"IPJOALIT","created_at":"2026-07-05T09:00:17.567566+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2412.03594","citing_title":"BatchLLM: Optimizing Large Batched LLM Inference with Global Prefix Sharing and Throughput-oriented Token Batching","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01214","citing_title":"Agentic AI Systems Should Be Designed as Marginal Token Allocators","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06113","citing_title":"Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO","json":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO.json","graph_json":"https://pith.science/api/pith-number/IPJOALITFO363UOPAO7KWW2VOO/graph.json","events_json":"https://pith.science/api/pith-number/IPJOALITFO363UOPAO7KWW2VOO/events.json","paper":"https://pith.science/paper/IPJOALIT"},"agent_actions":{"view_html":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO","download_json":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO.json","view_paper":"https://pith.science/paper/IPJOALIT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.15792&json=true","fetch_graph":"https://pith.science/api/pith-number/IPJOALITFO363UOPAO7KWW2VOO/graph.json","fetch_events":"https://pith.science/api/pith-number/IPJOALITFO363UOPAO7KWW2VOO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO/action/storage_attestation","attest_author":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO/action/author_attestation","sign_citation":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO/action/citation_signature","submit_replication":"https://pith.science/pith/IPJOALITFO363UOPAO7KWW2VOO/action/replication_record"}},"created_at":"2026-07-05T09:00:17.567566+00:00","updated_at":"2026-07-05T09:00:17.567566+00:00"}