{"id":"https://openalex.org/W7166847074","doi":"https://doi.org/10.18653/v1/2026.findings-acl.610","title":"Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning","display_name":"Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166847074","doi":"https://doi.org/10.18653/v1/2026.findings-acl.610"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.610","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.610","pdf_url":"https://aclanthology.org/2026.findings-acl.610.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.610.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5006513684","display_name":"Duygu Nur Yaldiz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Duygu Nur Yaldiz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139761502","display_name":"Evangelia Spiliopoulou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Evangelia Spiliopoulou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139734319","display_name":"Zheng Qi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040172025","display_name":"Siddharth Varia","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Siddharth Varia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120421405","display_name":"Srikanth Doss","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Srikanth Doss","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139752628","display_name":"Nikolaos Pappas","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nikolaos Pappas","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":0.83687094,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12537","last_page":"12553"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.1753000020980835,"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":0.1753000020980835,"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/T12761","display_name":"Data Stream Mining Techniques","score":0.10090000182390213,"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/T11652","display_name":"Imbalanced Data Classification Techniques","score":0.08529999852180481,"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/calibration","display_name":"Calibration","score":0.7217000126838684},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.5515999794006348},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.365200012922287},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.32580000162124634}],"concepts":[{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.7217000126838684},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6194999814033508},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5515999794006348},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.534600019454956},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41940000653266907},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.365200012922287},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.32580000162124634},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.30889999866485596},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2833999991416931},{"id":"https://openalex.org/C44154836","wikidata":"https://www.wikidata.org/wiki/Q45045","display_name":"Simulation","level":1,"score":0.2741999924182892},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.610","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.610","pdf_url":"https://aclanthology.org/2026.findings-acl.610.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.610","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.610","pdf_url":"https://aclanthology.org/2026.findings-acl.610.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","score":0.7346937656402588,"display_name":"Peace, Justice and strong institutions"}],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166847074.pdf","grobid_xml":"https://content.openalex.org/works/W7166847074.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"are":[4,19],"increasingly":[5],"deployed":[6],"in":[7,48,108],"decision-making":[8],"tasks,":[9],"where":[10],"not":[11,113],"only":[12],"accuracy":[13,143],"but":[14],"also":[15],"reliable":[16],"confidence":[17,21,115],"estimates":[18],"essential.Well-calibrated":[20],"enables":[22],"downstream":[23],"systems":[24],"to":[25,28,34,36,152],"decide":[26],"when":[27,33],"trust":[29],"a":[30,43,130],"model":[31],"and":[32,57,111],"defer":[35],"fallback":[37],"mechanisms.In":[38],"this":[39,126],"work,":[40],"we":[41,93,128],"conduct":[42],"systematic":[44],"study":[45],"of":[46,105],"calibration":[47],"two":[49],"widely":[50],"used":[51],"fine-tuning":[52],"paradigms:":[53],"supervised":[54],"finetuning":[55],"(SFT)":[56],"reinforcement":[58,119,132],"learning":[59,120,133],"with":[60,87],"verifiable":[61],"rewards":[62],"(RLVR).We":[63],"show":[64],"that":[65,98,135],"while":[66,145],"RLVR":[67],"improves":[68],"task":[69],"performance,":[70],"it":[71],"produces":[72],"extremely":[73],"overconfident":[74],"models,":[75],"whereas":[76],"SFT":[77],"yields":[78],"substantially":[79],"better":[80],"calibration,":[81],"even":[82],"under":[83],"distribution":[84],"shift,":[85],"though":[86],"smaller":[88],"performance":[89],"gains.Through":[90],"targeted":[91],"experiments,":[92],"diagnose":[94],"RLVR's":[95,142],"failure,":[96],"showing":[97],"decision":[99,107],"tokens":[100],"act":[101],"as":[102],"extraction":[103],"steps":[104],"the":[106],"reasoning":[109],"traces":[110],"do":[112],"carry":[114],"information,":[116],"which":[117],"prevents":[118],"from":[121],"surfacing":[122],"calibrated":[123],"alternatives.Based":[124],"on":[125],"insight,":[127],"propose":[129],"calibration-aware":[131],"formulation":[134],"directly":[136],"adjusts":[137],"decision-token":[138],"probabilities.Our":[139],"method":[140],"preserves":[141],"level":[144],"mitigating":[146],"overconfidence,":[147],"reducing":[148],"ECE":[149],"scores":[150],"up":[151],"9":[153],"points.":[154]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}