{"id":"https://openalex.org/W7160642421","doi":"https://doi.org/10.48550/arxiv.2605.06116","title":"Policy-Guided Stepwise Model Routing for Cost-Effective Reasoning","display_name":"Policy-Guided Stepwise Model Routing for Cost-Effective Reasoning","publication_year":2026,"publication_date":"2026-05-07","ids":{"openalex":"https://openalex.org/W7160642421","doi":"https://doi.org/10.48550/arxiv.2605.06116"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.06116","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.06116","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.06116","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5057113510","display_name":"Wenwen Si","orcid":"https://orcid.org/0000-0002-7066-9901"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Si, Wenwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135653344","display_name":"Insup Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Insup","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135716391","display_name":"Osbert Bastani","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bastani, Osbert","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":false,"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/T10028","display_name":"Topic Modeling","score":0.1606999933719635,"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"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.1606999933719635,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.14900000393390656,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.14139999449253082,"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/process","display_name":"Process (computing)","score":0.6428999900817871},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.6118999719619751},{"id":"https://openalex.org/keywords/routing","display_name":"Routing (electronic design automation)","score":0.5995000004768372},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5889999866485596},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5734999775886536},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.49939998984336853},{"id":"https://openalex.org/keywords/conjunction","display_name":"Conjunction (astronomy)","score":0.4779999852180481},{"id":"https://openalex.org/keywords/structured-prediction","display_name":"Structured prediction","score":0.39980000257492065}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7663999795913696},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.6428999900817871},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6118999719619751},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.5995000004768372},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5889999866485596},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5734999775886536},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5001000165939331},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.49939998984336853},{"id":"https://openalex.org/C59656382","wikidata":"https://www.wikidata.org/wiki/Q191536","display_name":"Conjunction (astronomy)","level":2,"score":0.4779999852180481},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.39980000257492065},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3930000066757202},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.3813000023365021},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.37770000100135803},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.35679998993873596},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.3513000011444092},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3255000114440918},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.3215999901294708},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.3122999966144562},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.262800008058548},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.25440001487731934},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.06116","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.06116","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.06116","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.06116","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"score":0.7932109832763672,"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Inference-time":[0],"computation":[1],"has":[2],"greatly":[3],"enhanced":[4],"the":[5,93,118],"performance":[6],"of":[7,36],"large":[8,53,135],"language":[9,34],"models":[10,35,56],"(LLMs)":[11],"on":[12,43,51,100,108],"challenging":[13],"reasoning":[14],"tasks,":[15],"but":[16],"this":[17],"strategy":[18],"can":[19],"incur":[20],"high":[21],"inference":[22],"costs.":[23],"One":[24],"solution":[25],"is":[26],"to":[27,33,91,122,130],"route":[28],"intermediate":[29],"chain-of-thought":[30],"(CoT)":[31],"states":[32],"different":[37],"sizes;":[38],"however,":[39],"existing":[40],"approaches":[41],"rely":[42],"handcrafted":[44,123],"routing":[45,68],"strategies":[46],"that":[47,57,132],"limit":[48],"performance,":[49],"or":[50],"training":[52,78,134],"process":[54,136],"reward":[55,137],"may":[58],"be":[59],"infeasible":[60],"in":[61,86],"many":[62],"applications.":[63],"We":[64,96],"formulate":[65],"stepwise":[66],"model":[67],"as":[69],"a":[70,79,127],"constrained":[71],"decision-making":[72],"problem,":[73],"which":[74],"we":[75],"solve":[76],"by":[77],"small":[80],"control":[81],"policy":[82],"using":[83],"reinforcement":[84],"learning":[85],"conjunction":[87],"with":[88],"threshold":[89],"calibration":[90],"tune":[92],"performance-efficiency":[94],"tradeoff.":[95],"validate":[97],"our":[98],"method":[99,115],"three":[101],"math":[102],"benchmarks":[103],"(GSM8K,":[104],"MATH500,":[105],"and":[106,111],"OmniMath)":[107],"both":[109],"open":[110],"closed":[112],"models.":[113,138],"Our":[114],"consistently":[116],"improves":[117],"accuracy-cost":[119],"tradeoff":[120,129],"compared":[121],"approaches,":[124],"while":[125],"achieving":[126],"comparable":[128],"methods":[131],"require":[133]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-09T00:00:00"}