{"id":"https://openalex.org/W7212840205","doi":"https://doi.org/10.18653/2026.emnlp-industry.190","title":"GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented Agents","display_name":"GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented Agents","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7212840205","doi":"https://doi.org/10.18653/2026.emnlp-industry.190"},"language":"en","primary_location":{"id":"doi:10.18653/2026.emnlp-industry.190","doi":"https://doi.org/10.18653/2026.emnlp-industry.190","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-industry.190","pdf_url":"https://aclanthology.org/2026.emnlp-industry.190.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2026.emnlp-industry.190.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5117865131","display_name":"Umesh Bodhwani","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Umesh Bodhwani","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011434798","display_name":"Thanh Lac Thi Tran","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thanh Tran","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5140156111","display_name":"Kai Wei","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kai Wei","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.73073834,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"3197","last_page":"3216"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12031","display_name":"Speech and dialogue systems","score":0.5503000020980835,"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/T12031","display_name":"Speech and dialogue systems","score":0.5503000020980835,"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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.07190000265836716,"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/T12128","display_name":"AI in Service Interactions","score":0.03750000149011612,"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-as-a-judge","display_name":"LLM-as-a-judge","score":0.8769999742507935},{"id":"https://openalex.org/keywords/construct-validity","display_name":"construct validity","score":1.0},{"id":"https://openalex.org/keywords/human-raters","display_name":"human raters","score":0.38999998569488525},{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.37299999594688416}],"concepts":[{"id":"https://openalex.org/C2776809875","wikidata":"https://www.wikidata.org/wiki/Q1375963","display_name":"Converse","level":2,"score":0.7914999723434448},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.7360000014305115},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.578000009059906},{"id":"https://openalex.org/C85847156","wikidata":"https://www.wikidata.org/wiki/Q59015987","display_name":"Verifiable secret sharing","level":3,"score":0.5351999998092651},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5097000002861023},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.4821000099182129},{"id":"https://openalex.org/C33676613","wikidata":"https://www.wikidata.org/wiki/Q13415176","display_name":"Dimension (graph theory)","level":2,"score":0.448199987411499},{"id":"https://openalex.org/C2779545769","wikidata":"https://www.wikidata.org/wiki/Q5135364","display_name":"Closeness","level":2,"score":0.42480000853538513},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.4223000109195709},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38839998841285706},{"id":"https://openalex.org/C2778029271","wikidata":"https://www.wikidata.org/wiki/Q5421931","display_name":"Extension (predicate logic)","level":2,"score":0.3560999929904938},{"id":"https://openalex.org/C184992742","wikidata":"https://www.wikidata.org/wiki/Q7243229","display_name":"Prime (order theory)","level":2,"score":0.32440000772476196},{"id":"https://openalex.org/C139676723","wikidata":"https://www.wikidata.org/wiki/Q1193832","display_name":"Sign (mathematics)","level":2,"score":0.3228999972343445},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.31769999861717224},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30660000443458557},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.29840001463890076},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2971000075340271},{"id":"https://openalex.org/C197947376","wikidata":"https://www.wikidata.org/wiki/Q5155608","display_name":"Comparability","level":2,"score":0.29649999737739563},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.29170000553131104},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.28790000081062317},{"id":"https://openalex.org/C2780876879","wikidata":"https://www.wikidata.org/wiki/Q3054749","display_name":"Meaning (existential)","level":2,"score":0.2831999957561493},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C106195933","wikidata":"https://www.wikidata.org/wiki/Q7847935","display_name":"Truncation (statistics)","level":2,"score":0.2800999879837036},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2718000113964081},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.262800008058548},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.25699999928474426}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/2026.emnlp-industry.190","doi":"https://doi.org/10.18653/2026.emnlp-industry.190","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-industry.190","pdf_url":"https://aclanthology.org/2026.emnlp-industry.190.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2609.12191","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.12191","pdf_url":"https://arxiv.org/pdf/2609.12191","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.2609.12191","doi":"https://doi.org/10.48550/arxiv.2609.12191","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.12191","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":"doi:10.18653/2026.emnlp-industry.190","doi":"https://doi.org/10.18653/2026.emnlp-industry.190","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-industry.190","pdf_url":"https://aclanthology.org/2026.emnlp-industry.190.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7212840205.pdf","grobid_xml":"https://content.openalex.org/works/W7212840205.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7125877758","https://openalex.org/W7135169797","https://openalex.org/W7214699787","https://openalex.org/W7204561462","https://openalex.org/W7223575517","https://openalex.org/W7220474426","https://openalex.org/W6979242103","https://openalex.org/W4385714639","https://openalex.org/W7133509200","https://openalex.org/W7163209687"],"abstract_inverted_index":{"Comparing":[0],"and":[1,26,60,75,119],"selecting":[2],"task-oriented":[3],"LLM":[4,15],"agents":[5,51],"increasingly":[6],"relies":[7],"on":[8,55,150,155],"a":[9,34,45,78,110,164,168,173,178],"low-cost":[10],"offline":[11,36],"evaluation":[12,67],"gate:":[13],"persona-driven":[14],"usersimulators":[16],"converse":[17],"with":[18,102],"each":[19],"candidate,":[20],"an":[21],"LLM-as-a-judge":[22],"scores":[23],"the":[24,27,56,107,126,132,140],"transcripts,":[25],"higher-scoring":[28],"agent":[29],"is":[30,129,159,177],"promoted.We":[31],"introduce":[32],"GAUGE,":[33],"reusable":[35],"protocol":[37],"that":[38,69],"measures":[39],"whether":[40],"this":[41,144],"gate's":[42,127],"ranking":[43,73,128],"matches":[44],"grounded":[46],"verifiable":[47],"reward":[48],"across":[49,113,131],"25":[50],"from":[52,99,148],"six":[53],"providers":[54],"\u03c4":[57],"2":[58],"-bench":[59],"Simu-latorArena":[61],"benchmarks,":[62,118],"separating":[63],"two":[64],"kinds":[65],"of":[66,104],"validity":[68,74],"release":[70],"practices":[71],"conflate:":[72],"construct":[76],"validity.First,":[77],"satisfaction-success":[79],"gap:":[80],"satisfaction":[81],"carries":[82],"essentially":[83],"no":[84],"information":[85],"about":[86],"task":[87],"success,":[88,101],"as":[89],"conversations":[90],"rated":[91],"satisfied":[92],"by":[93],"our":[94],"blind":[95],"panel":[96],"are":[97],"decorrelated":[98],"actual":[100],"57.5%":[103],"them":[105],"failing":[106],"customer's":[108],"task,":[109],"pattern":[111],"consistent":[112],"five":[114],"rater":[115],"populations,":[116],"both":[117],"every":[120],"subjective":[121],"dimension":[122],"we":[123,166],"rated.Second,":[124],"while":[125],"robust":[130],"broad":[133],"capability":[134],"span,":[135],"it":[136],"loses":[137],"resolution":[138],"among":[139],"nearequal":[141],"strong":[142],"agents:":[143],"decision-disagreement":[145],"rate":[146],"jumps":[147],"<1%":[149],"wide-reward":[151],"pairs":[152],"to":[153],"31%":[154],"close":[156],"pairs.The":[157],"gate":[158],"thus":[160],"humanvalidated":[161],"yet":[162],"mis-anchored.As":[163],"remedy,":[165],"propose":[167],"calibrate-then-trust":[169],"cadence":[170],"in":[171],"which":[172],"judge-free":[174],"completion":[175],"bit":[176],"zero-cost":[179],"tripwire":[180],"for":[181],"truncation":[182],"regressions.":[183]},"counts_by_year":[],"updated_date":"2026-10-31T12:09:11.096373","created_date":"2026-09-15T00:00:00"}