{"id":"https://openalex.org/W7147119302","doi":"https://doi.org/10.48550/arxiv.2603.26796","title":"Robust Batch-Level Query Routing for Large Language Models under Cost and Capacity Constraints","display_name":"Robust Batch-Level Query Routing for Large Language Models under Cost and Capacity Constraints","publication_year":2026,"publication_date":"2026-03-25","ids":{"openalex":"https://openalex.org/W7147119302","doi":"https://doi.org/10.48550/arxiv.2603.26796"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2603.26796","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2603.26796","pdf_url":"https://arxiv.org/pdf/2603.26796","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2603.26796","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5154030895","display_name":"Jelena Markovic-Voronov","orcid":"https://orcid.org/0000-0002-0562-0386","observed_orcids":["https://orcid.org/0000-0002-0562-0386"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Markovic-Voronov, Jelena","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075423316","display_name":"Kayhan Behdin","orcid":"https://orcid.org/0000-0003-3482-0421","observed_orcids":["https://orcid.org/0000-0003-3482-0421"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Behdin, Kayhan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126942858","display_name":"Yuanda Xu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yuanda","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125510945","display_name":"Zhengze Zhou","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Zhengze","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136144997","display_name":"Zhipeng Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhipeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5045271820","display_name":"Rahul Mazumder","orcid":"https://orcid.org/0000-0003-1384-9743","observed_orcids":["https://orcid.org/0000-0003-1384-9743","https://orcid.org/0000-0003-4285-7400"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mazumder, Rahul","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10101","display_name":"Cloud Computing and Resource Management","score":0.42820000648498535,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10101","display_name":"Cloud Computing and Resource Management","score":0.42820000648498535,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.13650000095367432,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.11249999701976776,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/query-routing","display_name":"query routing","score":0.9160000085830688},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9860000014305115},{"id":"https://openalex.org/keywords/resource-aware-routing","display_name":"resource-aware routing","score":0.6919999718666077},{"id":"https://openalex.org/keywords/gpu-resource-allocation","display_name":"GPU resource allocation","score":0.7860000133514404},{"id":"https://openalex.org/keywords/robustness","display_name":"robustness","score":0.4959999918937683},{"id":"https://openalex.org/keywords/performance-uncertainty","display_name":"performance uncertainty","score":0.4690000116825104},{"id":"https://openalex.org/keywords/multimodel-inference","display_name":"multi-model inference","score":0.5149999856948853}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8222000002861023},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5892999768257141},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.5090000033378601},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.4796000123023987},{"id":"https://openalex.org/C204948658","wikidata":"https://www.wikidata.org/wiki/Q1119410","display_name":"Static routing","level":4,"score":0.4309000074863434},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.4198000133037567},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3869999945163727},{"id":"https://openalex.org/C76522221","wikidata":"https://www.wikidata.org/wiki/Q5035396","display_name":"Multipath routing","level":5,"score":0.3578999936580658},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.357699990272522},{"id":"https://openalex.org/C29202148","wikidata":"https://www.wikidata.org/wiki/Q287260","display_name":"Resource allocation","level":2,"score":0.3343000113964081},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.3165000081062317},{"id":"https://openalex.org/C193702766","wikidata":"https://www.wikidata.org/wiki/Q1414548","display_name":"Concurrency","level":2,"score":0.31150001287460327},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.29649999737739563},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.29600000381469727},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.2921000123023987},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.28790000081062317},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.2621999979019165},{"id":"https://openalex.org/C84511453","wikidata":"https://www.wikidata.org/wiki/Q2914952","display_name":"Concurrency control","level":3,"score":0.25859999656677246}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2603.26796","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2603.26796","pdf_url":"https://arxiv.org/pdf/2603.26796","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2603.26796","doi":"https://doi.org/10.48550/arxiv.2603.26796","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26796","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2603.26796","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2603.26796","pdf_url":"https://arxiv.org/pdf/2603.26796","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7147119302.pdf","grobid_xml":"https://content.openalex.org/works/W7147119302.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7162670086","https://openalex.org/W7171523407","https://openalex.org/W7117870265","https://openalex.org/W4416506165","https://openalex.org/W7134854408","https://openalex.org/W7170072729","https://openalex.org/W7166272683","https://openalex.org/W4416042711","https://openalex.org/W4414591870","https://openalex.org/W4416712272"],"abstract_inverted_index":{"We":[0,60],"study":[1],"the":[2,107],"problem":[3],"of":[4,129],"routing":[5,21,43,111],"queries":[6],"to":[7,25,117,131,134],"large":[8],"language":[9],"models":[10],"(LLMs)":[11],"under":[12,30,119],"cost,":[13,28],"GPU":[14,144],"resources,":[15],"and":[16,56,84,122,143],"concurrency":[17],"constraints.":[18,146],"Prior":[19],"per-query":[20,113],"methods":[22,114],"often":[23],"fail":[24],"control":[26],"batch-level":[27,110],"especially":[29],"non-uniform":[31],"or":[32],"adversarial":[33,120],"batching.":[34],"To":[35],"address":[36],"this,":[37],"we":[38],"propose":[39],"a":[40,63,135],"batch-level,":[41],"resource-aware":[42],"framework":[44],"that":[45,66,81,96],"jointly":[46],"optimizes":[47],"model":[48,57],"assignment":[49],"for":[50,68],"each":[51],"batch":[52],"while":[53,139],"respecting":[54],"cost":[55,142],"capacity":[58],"limits.":[59],"further":[61],"introduce":[62],"robust":[64],"variant":[65],"accounts":[67],"uncertainty":[69],"in":[70],"predicted":[71],"LLM":[72,93],"performance,":[73],"along":[74],"with":[75],"an":[76],"offline":[77],"instance":[78,124],"allocation":[79,125],"procedure":[80],"balances":[82],"quality":[83],"throughput":[85],"across":[86],"multiple":[87],"models.":[88],"Experiments":[89],"on":[90,106],"two":[91],"multi-task":[92],"benchmarks":[94],"show":[95],"robustness":[97],"improves":[98],"accuracy":[99],"by":[100,115],"1-14%":[101],"over":[102],"non-robust":[103],"counterparts":[104],"(depending":[105],"performance":[108],"estimator),":[109],"outperforms":[112],"up":[116,130],"24%":[118],"batching,":[121],"optimized":[123],"yields":[126],"additional":[127],"gains":[128],"3%":[132],"compared":[133],"non-optimized":[136],"allocation,":[137],"all":[138],"strictly":[140],"controlling":[141],"resource":[145]},"counts_by_year":[],"updated_date":"2026-10-31T12:09:11.096373","created_date":"2026-04-02T00:00:00"}