{"id":"https://openalex.org/W7160331899","doi":"https://doi.org/10.48550/arxiv.2605.02241","title":"Zero-Shot Confidence Estimation for Small LLMs: When Supervised Baselines Aren't Worth Training","display_name":"Zero-Shot Confidence Estimation for Small LLMs: When Supervised Baselines Aren't Worth Training","publication_year":2026,"publication_date":"2026-05-04","ids":{"openalex":"https://openalex.org/W7160331899","doi":"https://doi.org/10.48550/arxiv.2605.02241"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.02241","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.02241","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.02241","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100718209","display_name":"Luong N. Nguyen","orcid":"https://orcid.org/0000-0003-0025-3537"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Nguyen, Luong N.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5100718209"],"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/T10181","display_name":"Natural Language Processing Techniques","score":0.19259999692440033,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.19259999692440033,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.11230000108480453,"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/T10028","display_name":"Topic Modeling","score":0.08810000121593475,"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/inference","display_name":"Inference","score":0.666700005531311},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.6536999940872192},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5174000263214111},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.4521999955177307},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.44179999828338623},{"id":"https://openalex.org/keywords/point-estimation","display_name":"Point estimation","score":0.4250999987125397},{"id":"https://openalex.org/keywords/similarity","display_name":"Similarity (geometry)","score":0.4018999934196472},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.3824000060558319}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.715499997138977},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.666700005531311},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.6536999940872192},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5835999846458435},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5198000073432922},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5174000263214111},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.4521999955177307},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.44179999828338623},{"id":"https://openalex.org/C41426520","wikidata":"https://www.wikidata.org/wiki/Q1192065","display_name":"Point estimation","level":2,"score":0.4250999987125397},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.4018999934196472},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.3824000060558319},{"id":"https://openalex.org/C162040801","wikidata":"https://www.wikidata.org/wiki/Q799897","display_name":"Bootstrap aggregating","level":2,"score":0.3596000075340271},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.35670000314712524},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3425999879837036},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.3418999910354614},{"id":"https://openalex.org/C136389625","wikidata":"https://www.wikidata.org/wiki/Q334384","display_name":"Supervised learning","level":3,"score":0.33880001306533813},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3325999975204468},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.31679999828338623},{"id":"https://openalex.org/C58471807","wikidata":"https://www.wikidata.org/wiki/Q327120","display_name":"Receiver operating characteristic","level":2,"score":0.30390000343322754},{"id":"https://openalex.org/C44249647","wikidata":"https://www.wikidata.org/wiki/Q208498","display_name":"Confidence interval","level":2,"score":0.29420000314712524},{"id":"https://openalex.org/C96250715","wikidata":"https://www.wikidata.org/wiki/Q965330","display_name":"Estimation","level":2,"score":0.2904999852180481},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.266400009393692},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.2630000114440918},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.25760000944137573}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.02241","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.02241","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.02241","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.02241","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":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"How":[0],"reliably":[1],"can":[2],"a":[3,18,43,121,138],"small":[4],"language":[5,34],"model":[6,21,35,46,73],"estimate":[7],"its":[8],"own":[9],"correctness?":[10],"The":[11],"answer":[12],"determines":[13],"whether":[14],"local-to-cloud":[15],"routing-escalating":[16],"queries":[17,41,81],"cheap":[19,44],"local":[20,45],"cannot":[22],"handle-can":[23],"work":[24],"without":[25],"supervised":[26,68,96,166],"training":[27,91],"data.":[28],"As":[29],"inference":[30],"costs":[31],"dominate":[32],"large":[33],"(LLM)":[36],"deployment":[37],"budgets,":[38],"routing":[39],"most":[40],"to":[42,156],"while":[47],"reserving":[48],"expensive":[49],"cloud":[50],"calls":[51],"for":[52],"hard":[53],"cases":[54],"is":[55,148],"an":[56],"increasingly":[57],"common":[58],"cost-control":[59],"strategy.":[60],"We":[61,178],"compare":[62],"zero-shot":[63,176],"confidence":[64],"signals":[65],"against":[66],"RouteLLM-style":[67],"baselines":[69,97],"across":[70],"three":[71],"7-8B":[72],"families":[74],"and":[75,79,110,183],"two":[76],"datasets":[77],"(1,000":[78],"500":[80],"per":[82],"model,":[83],"respectively).":[84],"Average":[85],"token":[86],"log-probability,":[87],"which":[88],"requires":[89],"no":[90],"data,":[92,182],"matches":[93],"or":[94],"exceeds":[95,174],"in-distribution":[98],"(Area":[99],"Under":[100],"the":[101,124,129,175],"Receiver":[102],"Operating":[103],"Characteristic":[104],"curve":[105],"(AUROC)":[106],"0.650-0.714":[107],"vs.":[108,116],"0.644-0.676)":[109],"substantially":[111],"outperforms":[112],"them":[113],"out-of-distribution":[114],"(0.717-0.833":[115],"0.512-0.564),":[117],"because":[118],"it":[119],"measures":[120],"property":[122],"of":[123],"model's":[125],"generation":[126],"rather":[127],"than":[128,163],"query":[130],"distribution.":[131],"This":[132],"paper":[133],"further":[134],"proposes":[135],"retrieval-conditional":[136],"self-assessment,":[137],"pre-generation":[139],"signal":[140],"that":[141],"selectively":[142],"injects":[143],"retrieved":[144],"knowledge":[145],"when":[146],"similarity":[147],"high,":[149],"improving":[150],"over":[151],"bare":[152],"self-assessment":[153],"by":[154],"up":[155],"+0.069":[157],"AUROC":[158],"at":[159],"3-10x":[160],"lower":[161],"latency":[162],"log-probability.":[164],"A":[165],"baseline":[167],"trained":[168],"on":[169],"1,000":[170],"labeled":[171],"examples":[172],"never":[173],"signal.":[177],"release":[179],"all":[180],"code,":[181],"experiment":[184],"logs.":[185]},"counts_by_year":[],"updated_date":"2026-05-06T06:10:43.113611","created_date":"2026-05-06T00:00:00"}