{"id":"https://openalex.org/W7151328746","doi":"https://doi.org/10.48550/arxiv.2604.04373","title":"Decocted Experience Improves Test-Time Inference in LLM Agents","display_name":"Decocted Experience Improves Test-Time Inference in LLM Agents","publication_year":2026,"publication_date":"2026-04-06","ids":{"openalex":"https://openalex.org/W7151328746","doi":"https://doi.org/10.48550/arxiv.2604.04373"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.04373","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.04373","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.04373","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5068080887","display_name":"Maohao Shen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, Maohao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133133205","display_name":"Kaiwen Zha","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zha, Kaiwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5069088584","display_name":"Zexue He","orcid":"https://orcid.org/0009-0001-9733-0545"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Zexue","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017776857","display_name":"Zhang-Wei Hong","orcid":"https://orcid.org/0000-0001-8574-6108"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hong, Zhang-Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5002970439","display_name":"Siru Ouyang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ouyang, Siru","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133132665","display_name":"J. Jon Ryu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ryu, J. Jon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060534465","display_name":"Prasanna Sattigeri","orcid":"https://orcid.org/0000-0003-4435-0486"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sattigeri, Prasanna","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083980887","display_name":"Suhas Diggavi","orcid":"https://orcid.org/0000-0001-7313-9861"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Diggavi, Suhas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5066172831","display_name":"Gregory W. Wornell","orcid":"https://orcid.org/0000-0001-9166-4758"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wornell, Gregory","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"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":0.1973000019788742,"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":0.1973000019788742,"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/T10260","display_name":"Software Engineering Research","score":0.14329999685287476,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.10909999907016754,"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/context","display_name":"Context (archaeology)","score":0.7049000263214111},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6863999962806702},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.6140999794006348},{"id":"https://openalex.org/keywords/salient","display_name":"Salient","score":0.5112000107765198},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.5110999941825867},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.45410001277923584}],"concepts":[{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.7049000263214111},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6863999962806702},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6830999851226807},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.6140999794006348},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5820000171661377},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5200999975204468},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.5112000107765198},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.5110999941825867},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.45410001277923584},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4413999915122986},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.34860000014305115},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.29580000042915344},{"id":"https://openalex.org/C2778751112","wikidata":"https://www.wikidata.org/wiki/Q835016","display_name":"Window (computing)","level":2,"score":0.2784999907016754},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.2718999981880188},{"id":"https://openalex.org/C76188268","wikidata":"https://www.wikidata.org/wiki/Q1783165","display_name":"Context effect","level":3,"score":0.26829999685287476},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2667999863624573},{"id":"https://openalex.org/C19768560","wikidata":"https://www.wikidata.org/wiki/Q320727","display_name":"Dependency (UML)","level":2,"score":0.25060001015663147}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.04373","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.04373","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.04373","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.04373","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":"article"},"sustainable_development_goals":[{"display_name":"Industry, innovation and infrastructure","id":"https://metadata.un.org/sdg/9","score":0.48264291882514954}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"There":[0],"is":[1,14,27],"growing":[2],"interest":[3],"in":[4],"improving":[5,68],"LLMs":[6],"without":[7],"updating":[8],"model":[9],"parameters.":[10],"One":[11],"well-established":[12],"direction":[13],"test-time":[15,41],"scaling,":[16],"where":[17],"increased":[18],"inference-time":[19],"computation":[20],"(e.g.,":[21],"longer":[22],"reasoning,":[23,166],"sampling,":[24],"or":[25],"search)":[26],"used":[28],"to":[29,50,75,105,151],"improve":[30],"performance.":[31],"However,":[32],"for":[33,67,136],"complex":[34],"reasoning":[35,81,160],"and":[36,47,71,120,147,161,169],"agentic":[37,162],"tasks,":[38,163],"naively":[39],"scaling":[40,65],"compute":[42],"can":[43],"substantially":[44],"increase":[45],"cost":[46],"still":[48],"lead":[49],"wasted":[51],"budget":[52],"on":[53,92],"suboptimal":[54],"exploration.":[55],"In":[56],"this":[57],"paper,":[58],"we":[59],"explore":[60],"\\emph{context}":[61],"as":[62,132],"a":[63,97,133],"complementary":[64],"axis":[66],"LLM":[69],"performance,":[70],"systematically":[72],"study":[73],"how":[74,104,110],"construct":[76],"better":[77],"inputs":[78],"that":[79,86],"guide":[80],"through":[82],"\\emph{experience}.":[83],"We":[84,95,128,155],"show":[85],"effective":[87,137,153],"context":[88,107,126,138],"construction":[89],"critically":[90],"depends":[91],"\\emph{decocted":[93,130],"experience}.":[94],"present":[96],"detailed":[98],"analysis":[99],"of":[100],"experience-augmented":[101],"agents,":[102],"studying":[103],"derive":[106],"from":[108,142],"experience,":[109,115,143],"performance":[111],"scales":[112],"with":[113],"accumulated":[114],"what":[116],"characterizes":[117],"good":[118],"context,":[119],"which":[121],"data":[122],"structures":[123],"best":[124],"support":[125],"construction.":[127],"identify":[129],"experience}":[131],"key":[134],"mechanism":[135],"construction:":[139],"extracting":[140],"essence":[141],"organizing":[144],"it":[145],"coherently,":[146],"retrieving":[148],"salient":[149],"information":[150],"build":[152],"context.":[154],"validate":[156],"our":[157],"findings":[158],"across":[159],"including":[164],"math":[165],"web":[167],"browsing,":[168],"software":[170],"engineering.":[171]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-04-08T00:00:00"}