{"id":"https://openalex.org/W7154373524","doi":"https://doi.org/10.48550/arxiv.2604.09624","title":"Self-Calibrating Language Models via Test-Time Discriminative Distillation","display_name":"Self-Calibrating Language Models via Test-Time Discriminative Distillation","publication_year":2026,"publication_date":"2026-03-18","ids":{"openalex":"https://openalex.org/W7154373524","doi":"https://doi.org/10.48550/arxiv.2604.09624"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.09624","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09624","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.2604.09624","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5093428389","display_name":"Mohamed Rissal Hedna","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hedna, Mohamed Rissal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133596252","display_name":"Jan Strich","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Strich, Jan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133614627","display_name":"Martin Semmann","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Semmann, Martin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133579444","display_name":"Chris Biemann","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Biemann, Chris","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"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":0.5145999789237976,"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":0.5145999789237976,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.06080000102519989,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.03869999945163727,"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/discriminative-model","display_name":"Discriminative model","score":0.7258999943733215},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.636900007724762},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5861999988555908},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.49050000309944153},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.4684000015258789},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.45590001344680786},{"id":"https://openalex.org/keywords/ranking","display_name":"Ranking (information retrieval)","score":0.4325999915599823},{"id":"https://openalex.org/keywords/signal","display_name":"SIGNAL (programming language)","score":0.4269999861717224},{"id":"https://openalex.org/keywords/categorization","display_name":"Categorization","score":0.4262000024318695},{"id":"https://openalex.org/keywords/probabilistic-logic","display_name":"Probabilistic logic","score":0.421999990940094}],"concepts":[{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.7258999943733215},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6399999856948853},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.636900007724762},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6277999877929688},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5940999984741211},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5861999988555908},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.49050000309944153},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.4684000015258789},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.45590001344680786},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.4325999915599823},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.4269999861717224},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.4262000024318695},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.421999990940094},{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.4007999897003174},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.382099986076355},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.3723999857902527},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.3675000071525574},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.3587999939918518},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.3564999997615814},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.3425000011920929},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.33500000834465027},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3314000070095062},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3249000012874603},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.3237000107765198},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3095000088214874},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.3043999969959259},{"id":"https://openalex.org/C2780428219","wikidata":"https://www.wikidata.org/wiki/Q16952335","display_name":"Cover (algebra)","level":2,"score":0.3009999990463257},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.3009999990463257},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.2996000051498413},{"id":"https://openalex.org/C2779010991","wikidata":"https://www.wikidata.org/wiki/Q2720909","display_name":"Artifact (error)","level":2,"score":0.288100004196167},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.28679999709129333},{"id":"https://openalex.org/C134261354","wikidata":"https://www.wikidata.org/wiki/Q938438","display_name":"Statistical inference","level":2,"score":0.28279998898506165},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.27649998664855957},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C8642999","wikidata":"https://www.wikidata.org/wiki/Q4171168","display_name":"Hyperparameter","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2637999951839447},{"id":"https://openalex.org/C2777530160","wikidata":"https://www.wikidata.org/wiki/Q41796","display_name":"Sentence","level":2,"score":0.2628999948501587},{"id":"https://openalex.org/C174576160","wikidata":"https://www.wikidata.org/wiki/Q1183700","display_name":"Deconvolution","level":2,"score":0.2556000053882599},{"id":"https://openalex.org/C7493553","wikidata":"https://www.wikidata.org/wiki/Q1520777","display_name":"Certainty","level":2,"score":0.2542000114917755}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.09624","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09624","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.2604.09624","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09624","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":[{"display_name":"Reduced inequalities","id":"https://metadata.un.org/sdg/10","score":0.7637373208999634}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1,142],"models":[2,143],"(LLMs)":[3],"are":[4],"systematically":[5],"overconfident:":[6],"they":[7,14,49],"routinely":[8],"express":[9],"high":[10],"certainty":[11],"on":[12,123],"questions":[13],"often":[15],"answer":[16,63],"incorrectly.":[17],"Existing":[18],"calibration":[19],"methods":[20],"either":[21],"require":[22],"labeled":[23,109],"validation":[24],"data,":[25],"degrade":[26],"under":[27],"distribution":[28,120],"shifts,":[29,121],"or":[30,111,167],"incur":[31],"substantial":[32],"inference":[33],"costs.":[34],"Recent":[35],"work":[36],"has":[37],"shown":[38],"that":[39,73,100,201],"LLMs":[40],"already":[41],"contain":[42],"a":[43,71,95],"better-calibrated":[44],"signal":[45,164,185],"than":[46,133],"the":[47,51,57,85,118,127,134,174],"one":[48],"verbalize:":[50],"token":[52],"probability":[53],"of":[54,126],"\"True\"":[55],"when":[56,117],"model":[58,146],"is":[59,74,80,173,204],"asked":[60],"\"Is":[61],"this":[62,102],"correct?\"":[64],"($P(\\text{True})$)":[65],"consistently":[66],"outperforms":[67],"their":[68],"stated":[69],"confidence,":[70],"gap":[72,103],"theoretically":[75],"grounded":[76],"as":[77,104],"generative":[78],"error":[79],"lower-bounded":[81],"by":[82,158],"roughly":[83],"twice":[84],"corresponding":[86],"discriminative":[87],"error.":[88],"We":[89],"introduce":[90],"$\\textbf{SECL}$":[91],"($\\textbf{SE}$lf-$\\textbf{C}$alibrating":[92],"$\\textbf{L}$anguage":[93],"Models),":[94],"test-time":[96],"training":[97,122],"(TTT)":[98],"pipeline":[99],"exploits":[101],"label-free":[105],"self-supervision,":[106],"requiring":[107],"no":[108],"data":[110],"human":[112],"supervision.":[113],"SECL":[114,152,172],"adapts":[115],"only":[116],"input":[119],"just":[124],"6--26%":[125],"question":[128],"stream":[129],"at":[130],"lower":[131],"cost":[132],"baseline":[135],"it":[136],"distills":[137],"from.":[138],"Across":[139],"four":[140,149],"small":[141],"from":[144],"three":[145],"families":[147],"and":[148,165,197,206],"diverse":[150],"domains,":[151],"reduces":[153],"Expected":[154],"Calibration":[155],"Error":[156],"(ECE)":[157],"56--78%,":[159],"outperforming":[160,168],"its":[161],"own":[162],"supervision":[163],"matching":[166],"recent":[169],"inference-time":[170],"methods.":[171],"first":[175],"method":[176],"to":[177,180],"apply":[178],"TTT":[179],"calibration;":[181],"seven":[182],"ablations":[183],"covering":[184],"quality,":[186],"gating":[187],"strategy,":[188],"weight":[189],"accumulation,":[190],"loss":[191],"design,":[192],"domain":[193],"ordering,":[194],"hyperparameter":[195],"sensitivity,":[196],"layer":[198],"selection":[199],"confirm":[200],"each":[202],"component":[203],"crucial":[205],"robust":[207],"across":[208],"configurations.":[209],"Code:":[210],"https://anonymous.4open.science/r/secl-emnlp26-submission-C890":[211]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-04-15T00:00:00"}