{"id":"https://openalex.org/W7159642869","doi":"https://doi.org/10.48550/arxiv.2604.27151","title":"Step-level Optimization for Efficient Computer-use Agents","display_name":"Step-level Optimization for Efficient Computer-use Agents","publication_year":2026,"publication_date":"2026-04-29","ids":{"openalex":"https://openalex.org/W7159642869","doi":"https://doi.org/10.48550/arxiv.2604.27151"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.27151","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.27151","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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.27151","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125489091","display_name":"Jinbiao Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Jinbiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043680518","display_name":"Kangqi Ni","orcid":"https://orcid.org/0009-0007-5255-2260"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ni, Kangqi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125538427","display_name":"Yilun Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Yilun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134947348","display_name":"Guo Gan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gan, Guo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134948033","display_name":"Arman Cohan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cohan, Arman","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/T12761","display_name":"Data Stream Mining Techniques","score":0.15850000083446503,"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/T12761","display_name":"Data Stream Mining Techniques","score":0.15850000083446503,"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/T10639","display_name":"Advanced Software Engineering Methodologies","score":0.09239999949932098,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.08699999749660492,"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/modular-design","display_name":"Modular design","score":0.6216999888420105},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5600000023841858},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.4936000108718872},{"id":"https://openalex.org/keywords/automation","display_name":"Automation","score":0.39570000767707825},{"id":"https://openalex.org/keywords/software","display_name":"Software","score":0.39570000767707825},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.3862000107765198},{"id":"https://openalex.org/keywords/path","display_name":"Path (computing)","score":0.3653999865055084},{"id":"https://openalex.org/keywords/debugging","display_name":"Debugging","score":0.32010000944137573},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.31940001249313354}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7937999963760376},{"id":"https://openalex.org/C101468663","wikidata":"https://www.wikidata.org/wiki/Q1620158","display_name":"Modular design","level":2,"score":0.6216999888420105},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5600000023841858},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.4936000108718872},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.40869998931884766},{"id":"https://openalex.org/C115901376","wikidata":"https://www.wikidata.org/wiki/Q184199","display_name":"Automation","level":2,"score":0.39570000767707825},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.39570000767707825},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3862000107765198},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3653999865055084},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.3653999865055084},{"id":"https://openalex.org/C168065819","wikidata":"https://www.wikidata.org/wiki/Q845566","display_name":"Debugging","level":2,"score":0.32010000944137573},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.31940001249313354},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.31450000405311584},{"id":"https://openalex.org/C5894958","wikidata":"https://www.wikidata.org/wiki/Q2297769","display_name":"Software agent","level":2,"score":0.31290000677108765},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.3091000020503998},{"id":"https://openalex.org/C13687954","wikidata":"https://www.wikidata.org/wiki/Q4826847","display_name":"Autonomous agent","level":2,"score":0.3059000074863434},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.30230000615119934},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.2969000041484833},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2957000136375427},{"id":"https://openalex.org/C120060458","wikidata":"https://www.wikidata.org/wiki/Q10145","display_name":"Milestone","level":2,"score":0.2842000126838684},{"id":"https://openalex.org/C137703981","wikidata":"https://www.wikidata.org/wiki/Q4692093","display_name":"Agent architecture","level":3,"score":0.2825999855995178},{"id":"https://openalex.org/C41550386","wikidata":"https://www.wikidata.org/wiki/Q529909","display_name":"Multi-agent system","level":2,"score":0.2759000062942505},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C37789001","wikidata":"https://www.wikidata.org/wiki/Q782543","display_name":"Graphical user interface","level":2,"score":0.2655999958515167},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2628999948501587},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.25119999051094055},{"id":"https://openalex.org/C155846161","wikidata":"https://www.wikidata.org/wiki/Q1143367","display_name":"Graphical model","level":2,"score":0.2506999969482422}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.27151","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.27151","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.27151","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.27151","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":"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Computer-use":[0],"agents":[1,35,156,249],"provide":[2],"a":[3,93,159,167,184,199],"promising":[4],"path":[5],"toward":[6],"general":[7],"software":[8],"automation":[9],"because":[10],"they":[11],"can":[12,79,241],"interact":[13],"directly":[14],"with":[15],"arbitrary":[16],"graphical":[17],"user":[18],"interfaces":[19],"instead":[20],"of":[21,60,97,230,246],"relying":[22],"on":[23,244],"brittle,":[24],"application-specific":[25],"integrations.":[26],"Despite":[27],"recent":[28,192],"advances":[29],"in":[30,40],"benchmark":[31],"performance,":[32],"strong":[33],"computer-use":[34,101,155,248],"remain":[36],"expensive":[37],"and":[38,78,124,164,195,198,238],"slow":[39],"practice,":[41],"since":[42],"most":[43,211],"systems":[44],"invoke":[45],"large":[46,259],"multimodal":[47],"models":[48],"at":[49,92],"nearly":[50],"every":[51],"interaction":[52],"step.":[53],"We":[54],"argue":[55],"that":[56,157,187,202],"this":[57,146],"uniform":[58],"allocation":[59,226],"compute":[61,225],"is":[62,210,236],"fundamentally":[63],"inefficient":[64],"for":[65,154,213],"long-horizon":[66],"GUI":[67],"tasks.":[68],"Such":[69],"trajectories":[70],"are":[71,76],"highly":[72],"heterogeneous:":[73],"many":[74],"steps":[75],"routine":[77],"be":[80,242],"handled":[81],"reliably":[82],"by":[83,162],"smaller,":[84],"cheaper":[85],"policies,":[86],"while":[87],"errors":[88],"tend":[89],"to":[90,120,166],"concentrate":[91],"relatively":[94],"small":[95,160],"number":[96],"high-risk":[98],"moments.":[99],"Across":[100],"benchmarks,":[102],"these":[103],"failures":[104],"repeatedly":[105],"take":[106],"two":[107,181],"forms:":[108],"progress":[109,190],"stalls,":[110],"where":[111,128,207],"the":[112,129,140,228,252,258],"agent":[113,130,254],"loops,":[114],"repeats":[115],"ineffective":[116],"actions,":[117],"or":[118,256],"fails":[119],"make":[121],"meaningful":[122,205],"progress,":[123],"silent":[125],"semantic":[126],"drift,":[127],"continues":[131],"taking":[132],"locally":[133],"plausible":[134],"actions":[135],"after":[136],"already":[137],"deviating":[138],"from":[139,191],"user's":[141],"true":[142],"goal.":[143],"To":[144],"address":[145],"inefficiency,":[147],"we":[148],"propose":[149],"an":[150,231],"event-driven,":[151],"step-level":[152],"cascade":[153],"runs":[158],"policy":[161],"default":[163],"escalates":[165],"stronger":[168],"model":[169],"only":[170],"when":[171],"lightweight":[172],"learned":[173],"monitors":[174],"detect":[175],"elevated":[176],"risk.":[177],"Our":[178],"framework":[179,235],"combines":[180],"complementary":[182],"signals:":[183],"Stuck":[185],"Monitor":[186,201],"detects":[188],"degraded":[189],"reasoning-action":[193],"history":[194],"triggers":[196],"recovery,":[197],"Milestone":[200],"identifies":[203],"semantically":[204],"checkpoints":[206],"sparse":[208],"verification":[209],"informative":[212],"catching":[214],"drift.":[215],"This":[216],"design":[217],"turns":[218],"always-on":[219],"frontier-model":[220],"inference":[221],"into":[222],"adaptive,":[223],"on-demand":[224],"over":[227],"course":[229],"evolving":[232],"interaction.":[233],"The":[234],"modular":[237],"deployment-oriented:":[239],"it":[240],"layered":[243],"top":[245],"existing":[247],"without":[250],"changing":[251],"underlying":[253],"architecture":[255],"retraining":[257],"model.":[260]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-02T00:00:00"}