{"id":"https://openalex.org/W7166803155","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1671","title":"Evaluation Pitfalls and Sparsity Limitations in LLM-based Confidence Estimates for Classification","display_name":"Evaluation Pitfalls and Sparsity Limitations in LLM-based Confidence Estimates for Classification","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166803155","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1671"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.1671","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1671","pdf_url":"https://aclanthology.org/2026.findings-acl.1671.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":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.1671.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5007507952","display_name":"Elena Merdjanovska","orcid":"https://orcid.org/0000-0003-0794-8438"},"institutions":[{"id":"https://openalex.org/I39343248","display_name":"Humboldt-Universit\u00e4t zu Berlin","ror":"https://ror.org/01hcx6992","country_code":"DE","type":"education","lineage":["https://openalex.org/I39343248"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Elena Merdjanovska","raw_affiliation_strings":["Humboldt-Universit\u00e4t zu Berlin Science of Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Humboldt-Universit\u00e4t zu Berlin Science of Intelligence","institution_ids":["https://openalex.org/I39343248"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139726695","display_name":"Omar Zaidan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Omar Zaidan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5014688346","display_name":"Andreas R\u00fcckl\u00e9","orcid":"https://orcid.org/0000-0002-4962-0556"},"institutions":[{"id":"https://openalex.org/I39343248","display_name":"Humboldt-Universit\u00e4t zu Berlin","ror":"https://ror.org/01hcx6992","country_code":"DE","type":"education","lineage":["https://openalex.org/I39343248"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Andreas R\u00fcckl\u00e9","raw_affiliation_strings":["Humboldt-Universit\u00e4t zu Berlin Science of Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Humboldt-Universit\u00e4t zu Berlin Science of Intelligence","institution_ids":["https://openalex.org/I39343248"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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":0.77574041,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"33424","last_page":"33435"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2062000036239624,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2062000036239624,"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":0.19300000369548798,"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":0.07829999923706055,"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/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.29750001430511475},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.2937000095844269},{"id":"https://openalex.org/keywords/confidence-interval","display_name":"Confidence interval","score":0.2883000075817108},{"id":"https://openalex.org/keywords/bayesian-probability","display_name":"Bayesian probability","score":0.2660999894142151}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5120999813079834},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4472000002861023},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3337000012397766},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.29750001430511475},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2937000095844269},{"id":"https://openalex.org/C44249647","wikidata":"https://www.wikidata.org/wiki/Q208498","display_name":"Confidence interval","level":2,"score":0.2883000075817108},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28760001063346863},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.2660999894142151},{"id":"https://openalex.org/C2778067643","wikidata":"https://www.wikidata.org/wiki/Q166507","display_name":"Interval (graph theory)","level":2,"score":0.2653999924659729},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.24729999899864197}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.1671","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1671","pdf_url":"https://aclanthology.org/2026.findings-acl.1671.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"},{"id":"pmh:oai:arXiv.org:2608.04899","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.04899","pdf_url":"https://arxiv.org/pdf/2608.04899","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,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.1671","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1671","pdf_url":"https://aclanthology.org/2026.findings-acl.1671.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/13","score":0.6652234196662903,"display_name":"Climate action"}],"awards":[{"id":"https://openalex.org/G3221473735","display_name":null,"funder_award_id":"EXC 2002/1","funder_id":"https://openalex.org/F4320320879","funder_display_name":"Deutsche Forschungsgemeinschaft"},{"id":"https://openalex.org/G6510858048","display_name":"EXC 2002: Science of Intelligence (SCIoI)","funder_award_id":"390523135","funder_id":"https://openalex.org/F4320320879","funder_display_name":"Deutsche Forschungsgemeinschaft"},{"id":"https://openalex.org/G6532450817","display_name":null,"funder_award_id":"EXC 2002","funder_id":"https://openalex.org/F4320320879","funder_display_name":"Deutsche Forschungsgemeinschaft"}],"funders":[{"id":"https://openalex.org/F4320320879","display_name":"Deutsche Forschungsgemeinschaft","ror":"https://ror.org/018mejw64"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166803155.pdf","grobid_xml":"https://content.openalex.org/works/W7166803155.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Confidence":[0],"estimation":[1],"is":[2],"essential":[3],"when":[4,11],"LLMs":[5],"are":[6],"used":[7],"for":[8,90,94],"classification,":[9],"indicating":[10],"predictions":[12],"can":[13],"be":[14],"trusted.However,":[15],"common":[16],"approaches":[17],"such":[18,98],"as":[19],"verbalization":[20,114],"produce":[21],"extremely":[22],"sparse":[23],"outputs.For":[24],"instance,":[25],"Qwen3-32B":[26],"verbalizes":[27],"only":[28],"eight":[29],"unique":[30],"confidence":[31],"values":[32],"on":[33],"SST-2,":[34],"with":[35,76],"over":[36,124],"half":[37],"being":[38],"exactly":[39],"95%-a":[40],"pattern":[41],"we":[42,54,102,112],"observe":[43],"consistently":[44],"across":[45],"four":[46],"datasets":[47],"and":[48,117],"two":[49],"LLMs.Besides":[50],"limiting":[51],"practical":[52],"utility,":[53],"show":[55],"that":[56,104],"this":[57],"sparsity":[58,116],"critically":[59],"affects":[60],"evaluation:":[61],"the":[62,69,119],"choice":[63],"of":[64],"interpolation":[65,93],"in":[66],"area":[67],"under":[68,84],"accuracy-rejection":[70],"curve":[71],"(AUARC)":[72],"dramatically":[73],"alters":[74],"rankings,":[75],"consistency":[77],"sampling":[78],"dropping":[79],"from":[80],"best":[81,120],"to":[82],"worst":[83],"stepwise":[85,92],"versus":[86],"linear":[87],"interpolation.We":[88],"advocate":[89],"standardizing":[91],"a":[95,99],"fairer":[96],"comparison.Under":[97],"fair":[100],"evaluation,":[101],"find":[103],"weighting":[105],"verbalized":[106],"digits":[107],"by":[108],"token":[109],"probabilities-a":[110],"method":[111],"term":[113],"logprobs-addresses":[115],"achieves":[118],"AUARC":[121],"(+2.3":[122],"points":[123],"vanilla":[125],"verbalization)":[126],"without":[127],"incurring":[128],"additional":[129],"inference":[130],"cost.":[131]},"counts_by_year":[],"updated_date":"2026-08-10T07:06:39.071514","created_date":"2026-07-02T00:00:00"}