{"id":"https://openalex.org/W7221001007","doi":"https://doi.org/10.48550/arxiv.2610.07335","title":"Selective Critique for Cost-Aware LLM Agents in Long-Horizon Decision Making","display_name":"Selective Critique for Cost-Aware LLM Agents in Long-Horizon Decision Making","publication_year":2026,"publication_date":"2026-10-05","ids":{"openalex":"https://openalex.org/W7221001007","doi":"https://doi.org/10.48550/arxiv.2610.07335"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.07335","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.07335","pdf_url":"https://arxiv.org/pdf/2610.07335","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.07335","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5157969013","display_name":"Heewon Park","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I848706","display_name":"Sungkyunkwan University","ror":"https://ror.org/04q78tk20","country_code":"KR","type":"education","lineage":["https://openalex.org/I848706"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Park, Heewon","raw_affiliation_strings":["Department of Electrical and Computer Engineering , Sungkyunkwan University , Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Electrical and Computer Engineering , Sungkyunkwan University , Republic of Korea","institution_ids":["https://openalex.org/I848706"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5157965291","display_name":"Somin Im","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I141371507","display_name":"Soongsil University","ror":"https://ror.org/017xnm587","country_code":"KR","type":"education","lineage":["https://openalex.org/I141371507"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Im, Somin","raw_affiliation_strings":["Soongsil Univer- sity , Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Soongsil Univer- sity , Republic of Korea","institution_ids":["https://openalex.org/I141371507"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5032094381","display_name":"Minhae Kwon","orcid":"https://orcid.org/0000-0002-8807-3719","observed_orcids":["https://orcid.org/0000-0002-8807-3719"]},"institutions":[{"id":"https://openalex.org/I848706","display_name":"Sungkyunkwan University","ror":"https://ror.org/04q78tk20","country_code":"KR","type":"education","lineage":["https://openalex.org/I848706"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Kwon, Minhae","raw_affiliation_strings":["Department of Electrical and Computer Engineering , Sungkyunkwan University , Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Electrical and Computer Engineering , Sungkyunkwan University , Republic of Korea","institution_ids":["https://openalex.org/I848706"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.4611999988555908,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.4611999988555908,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.3452000021934509,"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":0.03139999881386757,"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":0.9120000004768372},{"id":"https://openalex.org/keywords/long-horizon-decision-making","display_name":"long-horizon decision-making","score":0.9139999747276306},{"id":"https://openalex.org/keywords/value-of-information","display_name":"value of information","score":0.8659999966621399},{"id":"https://openalex.org/keywords/token-efficiency","display_name":"token efficiency","score":0.6150000095367432},{"id":"https://openalex.org/keywords/agent-reliability","display_name":"agent reliability","score":0.6669999957084656},{"id":"https://openalex.org/keywords/alfworld","display_name":"ALFWorld","score":0.671999990940094}],"concepts":[{"id":"https://openalex.org/C2780522230","wikidata":"https://www.wikidata.org/wiki/Q1140419","display_name":"Ambiguity","level":2,"score":0.7027999758720398},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6545000076293945},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6215000152587891},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.521399974822998},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5023999810218811},{"id":"https://openalex.org/C2776946740","wikidata":"https://www.wikidata.org/wiki/Q358652","display_name":"Deliberation","level":3,"score":0.4912000000476837},{"id":"https://openalex.org/C2780767217","wikidata":"https://www.wikidata.org/wiki/Q5532421","display_name":"Generality","level":2,"score":0.4016000032424927},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.38989999890327454},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.37450000643730164},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.3321000039577484},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.3310999870300293},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.30550000071525574},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.295199990272522},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.28630000352859497},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.28459998965263367},{"id":"https://openalex.org/C2776291640","wikidata":"https://www.wikidata.org/wiki/Q2912517","display_name":"Value (mathematics)","level":2,"score":0.26930001378059387},{"id":"https://openalex.org/C153517567","wikidata":"https://www.wikidata.org/wiki/Q26090","display_name":"Mechanism design","level":2,"score":0.2671000063419342},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.26499998569488525},{"id":"https://openalex.org/C205706631","wikidata":"https://www.wikidata.org/wiki/Q2319304","display_name":"Expected utility hypothesis","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C41550386","wikidata":"https://www.wikidata.org/wiki/Q529909","display_name":"Multi-agent system","level":2,"score":0.2621999979019165},{"id":"https://openalex.org/C28901747","wikidata":"https://www.wikidata.org/wiki/Q177571","display_name":"Decision theory","level":2,"score":0.2605000138282776}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.07335","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.07335","pdf_url":"https://arxiv.org/pdf/2610.07335","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.2610.07335","doi":"https://doi.org/10.48550/arxiv.2610.07335","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.07335","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:2610.07335","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.07335","pdf_url":"https://arxiv.org/pdf/2610.07335","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":[{"id":"https://openalex.org/G1701925455","display_name":null,"funder_award_id":"RS-2023-00278812","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"},{"id":"https://openalex.org/G1815356744","display_name":null,"funder_award_id":"RS-2025-02214082","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"},{"id":"https://openalex.org/G3812690899","display_name":null,"funder_award_id":"RS-2023","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"},{"id":"https://openalex.org/G7845709594","display_name":null,"funder_award_id":"RS-2026-25519475","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"},{"id":"https://openalex.org/G9000372721","display_name":null,"funder_award_id":"RS-2026-25519475","funder_id":"https://openalex.org/F4320335489","funder_display_name":"Institute for Information and Communications Technology Promotion"}],"funders":[{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"},{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"},{"id":"https://openalex.org/F4320334865","display_name":"Korea Customs Service","ror":"https://ror.org/01rkgg256"},{"id":"https://openalex.org/F4320334869","display_name":"Korean National Police Agency","ror":"https://ror.org/05x57gp50"},{"id":"https://openalex.org/F4320335489","display_name":"Institute for Information and Communications Technology Promotion","ror":"https://ror.org/01g0hqq23"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7221001007.pdf","grobid_xml":"https://content.openalex.org/works/W7221001007.grobid-xml"},"referenced_works_count":5,"referenced_works":["https://openalex.org/W2736601468","https://openalex.org/W4327810158","https://openalex.org/W4387561528","https://openalex.org/W4391631327","https://openalex.org/W6910759985"],"related_works":["https://openalex.org/W7204881801","https://openalex.org/W7221042092","https://openalex.org/W4412394791","https://openalex.org/W4415288389","https://openalex.org/W7154426196","https://openalex.org/W7152605534","https://openalex.org/W7166813837","https://openalex.org/W7143544910","https://openalex.org/W7130613997","https://openalex.org/W4417116219"],"abstract_inverted_index":{"Improving":[0],"the":[1,93,116,124,137,146,170,233],"reliability":[2,38,234],"of":[3,95,118,121,232,236],"large":[4],"language":[5],"model":[6],"(LLM)":[7],"agents":[8,20],"in":[9,203],"long-horizon":[10,82,160],"decision-making":[11],"remains":[12],"a":[13,69,77,86,110,193,200,208,212,220],"key":[14],"challenge.":[15],"When":[16],"deployed":[17],"as":[18,76],"autonomous":[19],"interacting":[21],"with":[22,66,174,211],"complex":[23],"environments,":[24],"early":[25],"mistakes":[26],"can":[27,229],"propagate":[28],"through":[29],"trajectories":[30],"and":[31,56,102,152,163,177],"cause":[32],"cascading":[33],"failures.":[34],"Recent":[35],"approaches":[36],"improve":[37],"by":[39],"incorporating":[40],"external":[41],"critique":[42,74,96,179,228],"or":[43],"deliberation,":[44],"but":[45],"invoking":[46],"these":[47],"mechanisms":[48],"at":[49],"every":[50],"step":[51],"substantially":[52,168],"increases":[53,184],"token":[54,194,205],"consumption":[55],"latency,":[57],"limiting":[58],"practical":[59],"deployment.":[60],"We":[61],"propose":[62],"SAG":[63,84,139,167,183],"(Self-improving":[64],"Agent":[65],"Gated":[67],"critique),":[68],"cost-aware":[70],"framework":[71],"that":[72,91,226],"formulates":[73],"invocation":[75],"step-wise":[78],"decision":[79],"problem":[80],"during":[81],"interaction.":[83],"introduces":[85],"lightweight,":[87],"training-free":[88],"gating":[89],"mechanism":[90,114],"estimates":[92],"utility":[94],"using":[97],"action-level":[98],"ambiguity":[99],"signals--global":[100],"entropy":[101],"local":[103],"top-2":[104],"margin--computed":[105],"over":[106],"admissible":[107],"actions.":[108],"From":[109],"decision-theoretic":[111],"perspective,":[112],"this":[113],"approximates":[115],"Value":[117],"Information":[119],"(VoI)":[120],"critique,":[122,224],"enabling":[123],"agent":[125],"to":[126,148,189,197,219],"selectively":[127],"allocate":[128],"expensive":[129],"feedback":[130],"only":[131],"when":[132],"its":[133],"expected":[134],"benefit":[135],"justifies":[136],"cost.":[138,242],"further":[140],"incorporates":[141],"online":[142],"bootstrapped":[143],"self-improvement,":[144],"allowing":[145],"actor":[147,210,222],"internalize":[149],"critic-assisted":[150],"behaviors":[151],"progressively":[153],"reduce":[154],"reliance":[155],"on":[156],"critique.":[157],"Across":[158],"three":[159],"interactive":[161],"benchmarks":[162],"multiple":[164],"backbone":[165],"models,":[166],"improves":[169],"performance-cost":[171],"trade-off":[172],"compared":[173],"both":[175],"no-critique":[176],"always-on":[178],"agents.":[180],"On":[181],"ALFWorld,":[182],"task":[185],"success":[186],"from":[187],"24.6%":[188],"78.4%":[190],"while":[191,238],"maintaining":[192],"budget":[195],"comparable":[196,218],"ReAct,":[198],"yielding":[199],"$3.1\\times$":[201],"improvement":[202],"normalized":[204],"efficiency.":[206],"Moreover,":[207],"7B":[209],"lightweight":[213],"3B":[214],"critic":[215],"achieves":[216],"performance":[217],"14B":[221],"without":[223],"showing":[225],"selective":[227],"recover":[230],"most":[231],"benefits":[235],"deliberation":[237],"dramatically":[239],"reducing":[240],"inference":[241]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-09T00:00:00"}