{"id":"https://openalex.org/W7213301262","doi":"https://doi.org/10.48550/arxiv.2609.15309","title":"When Agents Slow Down: Understanding LLM Agents' Test-Time Strategies via Elo-per-token Analysis","display_name":"When Agents Slow Down: Understanding LLM Agents' Test-Time Strategies via Elo-per-token Analysis","publication_year":2026,"publication_date":"2026-09-14","ids":{"openalex":"https://openalex.org/W7213301262","doi":"https://doi.org/10.48550/arxiv.2609.15309"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.15309","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.15309","pdf_url":"https://arxiv.org/pdf/2609.15309","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.15309","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5151831923","display_name":"Kaiyuan Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liu, Kaiyuan","raw_affiliation_strings":["UC Berkeley ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Berkeley ,","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092838030","display_name":"Qiuyang Mang","orcid":"https://orcid.org/0009-0000-4462-8527","observed_orcids":["https://orcid.org/0009-0000-4462-8527"]},"institutions":[{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mang, Qiuyang","raw_affiliation_strings":["\u2665 University of Washington ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2665 University of Washington ,","institution_ids":["https://openalex.org/I201448701"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5151838991","display_name":"Bo Peng","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Peng, Bo","raw_affiliation_strings":["\u2665 University of Washington ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2665 University of Washington ,","institution_ids":["https://openalex.org/I201448701"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103021774","display_name":"Wenhao Chai","orcid":"https://orcid.org/0000-0003-2611-0008","observed_orcids":["https://orcid.org/0000-0003-2611-0008"]},"institutions":[{"id":"https://openalex.org/I20089843","display_name":"Princeton University","ror":"https://ror.org/00hx57361","country_code":"US","type":"education","lineage":["https://openalex.org/I20089843"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chai, Wenhao","raw_affiliation_strings":["\u2666 Princeton University , \u2663 Bespoke Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2666 Princeton University , \u2663 Bespoke Labs","institution_ids":["https://openalex.org/I20089843"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5151873803","display_name":"Hanchen Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20089843","display_name":"Princeton University","ror":"https://ror.org/00hx57361","country_code":"US","type":"education","lineage":["https://openalex.org/I20089843"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li, Hanchen","raw_affiliation_strings":["\u2666 Princeton University , \u2663 Bespoke Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2666 Princeton University , \u2663 Bespoke Labs","institution_ids":["https://openalex.org/I20089843"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123467205","display_name":"Shreyas Pimpalgaonkar","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Pimpalgaonkar, Shreyas","raw_affiliation_strings":["\u2665 University of Washington ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2665 University of Washington ,","institution_ids":["https://openalex.org/I201448701"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067919401","display_name":"Luke Zettlemoyer","orcid":"https://orcid.org/0009-0008-8296-0764","observed_orcids":["https://orcid.org/0009-0008-8296-0764"]},"institutions":[{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zettlemoyer, Luke","raw_affiliation_strings":["\u2665 University of Washington ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2665 University of Washington ,","institution_ids":["https://openalex.org/I201448701"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014141359","display_name":"Alexandros G. Dimakis","orcid":"https://orcid.org/0000-0002-4244-7033","observed_orcids":["https://orcid.org/0000-0002-4244-7033"]},"institutions":[{"id":"https://openalex.org/I201448701","display_name":"University of Washington","ror":"https://ror.org/00cvxb145","country_code":"US","type":"education","lineage":["https://openalex.org/I201448701"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Dimakis, Alex","raw_affiliation_strings":["\u2665 University of Washington ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2665 University of Washington ,","institution_ids":["https://openalex.org/I201448701"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5123531127","display_name":"Alvin Cheung","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20089843","display_name":"Princeton University","ror":"https://ror.org/00hx57361","country_code":"US","type":"education","lineage":["https://openalex.org/I20089843"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Cheung, Alvin","raw_affiliation_strings":["\u2666 Princeton University , \u2663 Bespoke Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u2666 Princeton University , \u2663 Bespoke Labs","institution_ids":["https://openalex.org/I20089843"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":3,"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":0.21950000524520874,"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":0.21950000524520874,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1858000010251999,"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/T13904","display_name":"Artificial Intelligence Applications","score":0.15489999949932098,"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/llm-based-agents","display_name":"LLM agents","score":0.8870000243186951},{"id":"https://openalex.org/keywords/elo-rating","display_name":"Elo rating","score":0.6549999713897705},{"id":"https://openalex.org/keywords/bradley-terry-model","display_name":"Bradley-Terry model","score":0.8539999723434448},{"id":"https://openalex.org/keywords/open-ended-tasks","display_name":"open-ended tasks","score":0.6759999990463257},{"id":"https://openalex.org/keywords/continual-learning","display_name":"continual learning","score":0.5860000252723694},{"id":"https://openalex.org/keywords/feedback-driven-optimization","display_name":"feedback-driven optimization","score":0.546999990940094},{"id":"https://openalex.org/keywords/parallel-sessions","display_name":"parallel sessions","score":0.4480000138282776}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6912000179290771},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6026999950408936},{"id":"https://openalex.org/C2777582232","wikidata":"https://www.wikidata.org/wiki/Q5013414","display_name":"CONTEST","level":2,"score":0.49720001220703125},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.43619999289512634},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.43619999289512634},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.421099990606308},{"id":"https://openalex.org/C2779182362","wikidata":"https://www.wikidata.org/wiki/Q17126187","display_name":"Session (web analytics)","level":2,"score":0.40860000252723694},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.361299991607666},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35359999537467957},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.3366999924182892},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.31139999628067017},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.31040000915527344},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3093000054359436},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.3025999963283539},{"id":"https://openalex.org/C147297375","wikidata":"https://www.wikidata.org/wiki/Q6674930","display_name":"Look-ahead","level":2,"score":0.3001999855041504},{"id":"https://openalex.org/C8505890","wikidata":"https://www.wikidata.org/wiki/Q605095","display_name":"Budget constraint","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.26579999923706055},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.26460000872612},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.25870001316070557},{"id":"https://openalex.org/C2777655017","wikidata":"https://www.wikidata.org/wiki/Q1501161","display_name":"Toolbox","level":2,"score":0.2524000108242035}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.15309","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.15309","pdf_url":"https://arxiv.org/pdf/2609.15309","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.15309","doi":"https://doi.org/10.48550/arxiv.2609.15309","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.15309","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.15309","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.15309","pdf_url":"https://arxiv.org/pdf/2609.15309","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G1286236842","display_name":null,"funder_award_id":"-AC02-05CH11231","funder_id":"https://openalex.org/F4320306084","funder_display_name":"U.S. Department of Energy"},{"id":"https://openalex.org/G1562478397","display_name":"III: Medium: Collaborative Research: Reasoning about Optimizers for Data-Intensive Systems","funder_award_id":"1955488","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G2712957708","display_name":"CAREER: Generating Application-Specific Database Management Systems","funder_award_id":"2027575","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G3083819904","display_name":null,"funder_award_id":"05CH11231","funder_id":"https://openalex.org/F4320306084","funder_display_name":"U.S. Department of Energy"},{"id":"https://openalex.org/G5222170445","display_name":null,"funder_award_id":"AC02-05CH11231","funder_id":"https://openalex.org/F4320332180","funder_display_name":"Defense Advanced Research Projects Agency"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320306084","display_name":"U.S. Department of Energy","ror":"https://ror.org/01bj3aw27"},{"id":"https://openalex.org/F4320332180","display_name":"Defense Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7213301262.pdf","grobid_xml":"https://content.openalex.org/works/W7213301262.grobid-xml"},"referenced_works_count":31,"referenced_works":["https://openalex.org/W2160637580","https://openalex.org/W3001279689","https://openalex.org/W4318905229","https://openalex.org/W4379089709","https://openalex.org/W4389422760","https://openalex.org/W4402733601","https://openalex.org/W4404987753","https://openalex.org/W4405766390","https://openalex.org/W4406033152","https://openalex.org/W4414281281","https://openalex.org/W4414971614","https://openalex.org/W4415025609","https://openalex.org/W4415108068","https://openalex.org/W4415117812","https://openalex.org/W4416113464","https://openalex.org/W4416411532","https://openalex.org/W4417258194","https://openalex.org/W6910824972","https://openalex.org/W7116070620","https://openalex.org/W7118742113","https://openalex.org/W7125193990","https://openalex.org/W7125476382","https://openalex.org/W7131425980","https://openalex.org/W7133222210","https://openalex.org/W7151432372","https://openalex.org/W7160945459","https://openalex.org/W7161263753","https://openalex.org/W7167601841","https://openalex.org/W7167619593","https://openalex.org/W7169879786","https://openalex.org/W7203684476"],"related_works":["https://openalex.org/W4415117812","https://openalex.org/W7131400538","https://openalex.org/W4406492615","https://openalex.org/W7125694398","https://openalex.org/W7160869293","https://openalex.org/W4417161946","https://openalex.org/W7207687438","https://openalex.org/W7167570570","https://openalex.org/W7101576509","https://openalex.org/W4414592689"],"abstract_inverted_index":{"Large":[0],"language":[1],"model":[2,69],"(LLM)":[3],"agents":[4,89,130,179],"allocate":[5],"test-time":[6,23],"compute":[7],"adaptively":[8],"as":[9,188,203],"they":[10],"revise":[11],"solutions,":[12],"use":[13],"tools,":[14],"explore":[15],"alternatives,":[16],"and":[17,65,101,146,175,225],"decide":[18],"when":[19],"to":[20,28,70,86,98,102],"stop.":[21],"This":[22],"strategy":[24],"makes":[25],"it":[26,85],"difficult":[27],"measure":[29],"how":[30],"agent":[31],"performance":[32],"scales.":[33,82],"We":[34,51,83,182],"study":[35],"open-ended":[36,92],"tasks":[37,78],"that":[38],"provide":[39],"continuous":[40],"scores":[41],"for":[42,119],"intermediate":[43],"submissions,":[44],"making":[45],"progress":[46],"observable":[47],"throughout":[48],"long":[49,223],"trajectories.":[50],"propose":[52],"Elo-per-token":[53],"analysis,":[54],"which":[55,120],"tracks":[56],"the":[57,150,154,184,189,197,204],"best":[58],"solution":[59],"found":[60],"at":[61],"each":[62],"token":[63],"budget":[64,191],"uses":[66],"a":[67,115],"Bradley-Terry":[68],"aggregate":[71],"within-task":[72],"orderings":[73],"into":[74,135],"Elo":[75,121,136,194,220],"ratings":[76],"across":[77,211],"with":[79,94,124],"different":[80],"score":[81],"apply":[84],"four":[87,91],"general-purpose":[88],"on":[90,164,214],"benchmarks,":[93],"sessions":[95,213],"of":[96,172],"up":[97],"100M":[99,209],"tokens,":[100],"three":[103],"feedback-driven":[104],"LLM":[105],"optimization":[106],"harnesses":[107],"in":[108],"controlled":[109],"single-task":[110],"interventions.":[111],"Independent":[112],"sampling":[113],"provides":[114],"theoretically":[116],"characterized":[117],"reference,":[118,129],"grows":[122],"linearly":[123],"log":[125],"compute.":[126],"Against":[127],"this":[128,201],"can":[131],"initially":[132],"convert":[133],"tokens":[134,210],"faster":[137],"than":[138],"independent":[139],"sampling,":[140],"but":[141],"their":[142],"marginal":[143,193],"gains":[144,195],"diminish":[145],"eventually":[147],"fall":[148],"below":[149],"reference.":[151,199],"In":[152],"contrast,":[153],"strongest":[155],"historical":[156],"human":[157],"contestants":[158],"improve":[159],"superlinearly":[160],"over":[161,221,227],"contest":[162],"time":[163],"shared":[165],"AtCoder":[166],"Heuristic":[167],"Contest":[168],"tasks,":[169],"providing":[170],"evidence":[171],"continual":[173],"learning":[174],"substantial":[176],"headroom":[177],"after":[178],"slow":[180],"down.":[181],"define":[183],"scaling":[185],"inflection":[186],"point":[187,202],"per-session":[190,205],"where":[192],"match":[196],"independent-sampling":[198],"Using":[200],"budget,":[206],"we":[207],"split":[208],"parallel":[212],"FrontierCS":[215],"Polyomino":[216],"Packing,":[217],"gaining":[218],"+264":[219],"one":[222],"session":[224],"+355":[226],"ten":[228],"short":[229],"sessions.":[230]},"counts_by_year":[{"year":2026,"cited_by_count":3}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-16T00:00:00"}