{"id":"https://openalex.org/W4412945616","doi":"https://doi.org/10.18653/v1/2025.acl-long.390","title":"CER: Confidence Enhanced Reasoning in LLMs","display_name":"CER: Confidence Enhanced Reasoning in LLMs","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945616","doi":"https://doi.org/10.18653/v1/2025.acl-long.390"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.390","doi":"https://doi.org/10.18653/v1/2025.acl-long.390","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.390","pdf_url":"https://aclanthology.org/2025.acl-long.390.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","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/2025.acl-long.390.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5116361792","display_name":"Ali Razghandi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ali Razghandi","raw_affiliation_strings":["University of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Technology","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5116361793","display_name":"Seyed Mohammad Hadi Hosseini","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Seyed Mohammad Hadi Hosseini","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5069082023","display_name":"Mahdieh Soleymani Baghshah","orcid":"https://orcid.org/0000-0002-1971-6231","observed_orcids":["https://orcid.org/0000-0002-1971-6231"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mahdieh Soleymani Baghshah","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":0.16086401,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"7918","last_page":"7938"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.5175999999046326,"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":0.5175999999046326,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.17219999432563782,"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":0.06289999932050705,"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/confidence-estimation","display_name":"confidence estimation","score":0.8679999709129333},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9480000138282776},{"id":"https://openalex.org/keywords/reasoning","display_name":"reasoning","score":0.7289999723434448},{"id":"https://openalex.org/keywords/confidence-calibration","display_name":"confidence calibration","score":0.49900001287460327}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4030775725841522}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.390","doi":"https://doi.org/10.18653/v1/2025.acl-long.390","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.390","pdf_url":"https://aclanthology.org/2025.acl-long.390.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2502.14634","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.14634","pdf_url":"https://arxiv.org/pdf/2502.14634","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.2502.14634","doi":"https://doi.org/10.48550/arxiv.2502.14634","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2502.14634","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.390","doi":"https://doi.org/10.18653/v1/2025.acl-long.390","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.390","pdf_url":"https://aclanthology.org/2025.acl-long.390.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945616.pdf","grobid_xml":"https://content.openalex.org/works/W4412945616.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W6966850505","https://openalex.org/W4416795297","https://openalex.org/W7203588106","https://openalex.org/W4405431306","https://openalex.org/W4385572162","https://openalex.org/W7134248873","https://openalex.org/W4412889185","https://openalex.org/W7104055341","https://openalex.org/W7125352639","https://openalex.org/W4404355734"],"abstract_inverted_index":{"Ensuring":[0],"the":[1,38,63,81,100,111,150],"reliability":[2,112],"of":[3,40,65,84,93,102,113,152,162],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"in":[8,17,59,72,78,106,121,134,171],"complex":[9],"reasoning":[10,58,74,86],"tasks":[11],"remains":[12],"a":[13,107],"formidable":[14],"challenge,":[15],"particularly":[16],"scenarios":[18],"that":[19,55,109],"demand":[20],"precise":[21],"mathematical":[22,73,138],"calculations":[23],"and":[24,61,75,140,166,173],"knowledge-intensive":[25],"opendomain":[26],"generation.In":[27],"this":[28],"work,":[29],"we":[30,98],"introduce":[31],"an":[32,53,159],"uncertainty-aware":[33],"framework":[34],"designed":[35],"to":[36,119,128,158,164],"enhance":[37],"accuracy":[39,160],"LLM":[41],"responses":[42],"by":[43],"systematically":[44],"incorporating":[45],"model":[46],"confidence":[47,64,83,92],"at":[48],"critical":[49,95],"decision":[50],"points.We":[51],"propose":[52],"approach":[54],"encourages":[56],"multi-step":[57],"LLMs":[60],"quantify":[62],"intermediate":[66,96],"answers":[67],"such":[68],"as":[69],"numerical":[70],"results":[71,147],"proper":[76],"nouns":[77],"open-domain":[79,142,174],"generation.Then,":[80],"overall":[82],"each":[85,114,123],"chain":[87,125],"is":[88],"evaluated":[89],"based":[90],"on":[91],"these":[94],"steps.Finally,":[97],"aggregate":[99],"answer":[101],"generated":[103,115,124],"response":[104],"paths":[105],"way":[108],"reflects":[110],"content":[116],"(as":[117],"opposed":[118],"self-consistency":[120],"which":[122],"contributes":[126],"equally":[127],"majority":[129],"voting).We":[130],"conducted":[131],"extensive":[132],"experiments":[133],"five":[135],"datasets,":[136,143],"three":[137],"datasets":[139],"two":[141],"using":[144],"four":[145],"LLMs.The":[146],"consistently":[148],"validate":[149],"effectiveness":[151],"our":[153],"novel":[154],"confidenceaggregation":[155],"method,":[156],"leading":[157],"improvement":[161],"up":[163],"7.4%":[165],"5.8%":[167],"over":[168],"baseline":[169],"approaches":[170],"math":[172],"generation":[175],"tasks,":[176],"respectively.1":[177],"*":[178],"Equal":[179],"contribution.learners.":[180]},"counts_by_year":[],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2025-10-10T00:00:00"}