{"id":"https://openalex.org/W4412945675","doi":"https://doi.org/10.18653/v1/2025.acl-long.269","title":"Your Model is Overconfident, and Other Lies We Tell Ourselves","display_name":"Your Model is Overconfident, and Other Lies We Tell Ourselves","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945675","doi":"https://doi.org/10.18653/v1/2025.acl-long.269"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.269","doi":"https://doi.org/10.18653/v1/2025.acl-long.269","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.269","pdf_url":"https://aclanthology.org/2025.acl-long.269.pdf","source":{"id":"https://openalex.org/S4363608652","display_name":"Annual Meeting of the Association for Computational Linguistics (ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":"hybrid","oa_url":"https://aclanthology.org/2025.acl-long.269.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5073653815","display_name":"Timothee Mickus","orcid":"https://orcid.org/0000-0001-9538-7209","observed_orcids":["https://orcid.org/0000-0001-9538-7209"]},"institutions":[{"id":"https://openalex.org/I133731052","display_name":"University of Helsinki","ror":"https://ror.org/040af2s02","country_code":"FI","type":"education","lineage":["https://openalex.org/I133731052"]},{"id":"https://openalex.org/I4210090340","display_name":"Institut de Canc\u00e9rologie Strasbourg","ror":"https://ror.org/008fdbn61","country_code":"FR","type":"healthcare","lineage":["https://openalex.org/I4210090340"]},{"id":"https://openalex.org/I90183372","display_name":"Universit\u00e9 de Lorraine","ror":"https://ror.org/04vfs2w97","country_code":"FR","type":"education","lineage":["https://openalex.org/I90183372"]}],"countries":["FI","FR"],"is_corresponding":false,"raw_author_name":"Timothee Mickus","raw_affiliation_strings":["University of Helsinki","Universit\u00e9 de Lorraine ICANS Strasbourg"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Helsinki","institution_ids":["https://openalex.org/I133731052"]},{"raw_affiliation_string":"Universit\u00e9 de Lorraine ICANS Strasbourg","institution_ids":["https://openalex.org/I4210090340","https://openalex.org/I90183372"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Aman Sinha","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I133731052","display_name":"University of Helsinki","ror":"https://ror.org/040af2s02","country_code":"FI","type":"education","lineage":["https://openalex.org/I133731052"]},{"id":"https://openalex.org/I4210090340","display_name":"Institut de Canc\u00e9rologie Strasbourg","ror":"https://ror.org/008fdbn61","country_code":"FR","type":"healthcare","lineage":["https://openalex.org/I4210090340"]},{"id":"https://openalex.org/I90183372","display_name":"Universit\u00e9 de Lorraine","ror":"https://ror.org/04vfs2w97","country_code":"FR","type":"education","lineage":["https://openalex.org/I90183372"]}],"countries":["FI","FR"],"is_corresponding":false,"raw_author_name":"Aman Sinha","raw_affiliation_strings":["University of Helsinki","Universit\u00e9 de Lorraine ICANS Strasbourg"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Helsinki","institution_ids":["https://openalex.org/I133731052"]},{"raw_affiliation_string":"Universit\u00e9 de Lorraine ICANS Strasbourg","institution_ids":["https://openalex.org/I4210090340","https://openalex.org/I90183372"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5047023030","display_name":"Ra\u00fal V\u00e1zquez","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I133731052","display_name":"University of Helsinki","ror":"https://ror.org/040af2s02","country_code":"FI","type":"education","lineage":["https://openalex.org/I133731052"]},{"id":"https://openalex.org/I4210090340","display_name":"Institut de Canc\u00e9rologie Strasbourg","ror":"https://ror.org/008fdbn61","country_code":"FR","type":"healthcare","lineage":["https://openalex.org/I4210090340"]},{"id":"https://openalex.org/I90183372","display_name":"Universit\u00e9 de Lorraine","ror":"https://ror.org/04vfs2w97","country_code":"FR","type":"education","lineage":["https://openalex.org/I90183372"]}],"countries":["FI","FR"],"is_corresponding":false,"raw_author_name":"Ra\u00fal V\u00e1zquez","raw_affiliation_strings":["University of Helsinki","Universit\u00e9 de Lorraine ICANS Strasbourg"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Helsinki","institution_ids":["https://openalex.org/I133731052"]},{"raw_affiliation_string":"Universit\u00e9 de Lorraine ICANS Strasbourg","institution_ids":["https://openalex.org/I4210090340","https://openalex.org/I90183372"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"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":0.16446679,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"5401","last_page":"5417"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.7554000020027161,"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":0.7554000020027161,"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":0.06710000336170197,"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/T10664","display_name":"Sentiment Analysis and Opinion Mining","score":0.04569999873638153,"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/annotator-disagreement","display_name":"annotator disagreement","score":0.6790000200271606},{"id":"https://openalex.org/keywords/model-confidence","display_name":"model confidence","score":0.703000009059906},{"id":"https://openalex.org/keywords/training-dynamics","display_name":"training dynamics","score":0.7350000143051147},{"id":"https://openalex.org/keywords/uncertainty-measures","display_name":"uncertainty measures","score":0.6919999718666077},{"id":"https://openalex.org/keywords/data-complexity","display_name":"data complexity","score":0.6940000057220459}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6037956476211548}],"mesh":[],"locations_count":4,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.269","doi":"https://doi.org/10.18653/v1/2025.acl-long.269","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.269","pdf_url":"https://aclanthology.org/2025.acl-long.269.pdf","source":{"id":"https://openalex.org/S4363608652","display_name":"Annual Meeting of the Association for Computational Linguistics (ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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:2503.01235","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2503.01235","pdf_url":"https://arxiv.org/pdf/2503.01235","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":"pmh:oai:helda.helsinki.fi:10138/602754","doi":null,"is_oa":true,"landing_page_url":"http://hdl.handle.net/10138/602754","pdf_url":null,"source":{"id":"https://openalex.org/S4210213322","display_name":"Ty\u00f6v\u00e4entutkimus Vuosikirja","issn_l":"0784-1272","issn":["0784-1272","1459-7780"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":["jufo-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Conference contribution"},{"id":"doi:10.48550/arxiv.2503.01235","doi":"https://doi.org/10.48550/arxiv.2503.01235","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2503.01235","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.269","doi":"https://doi.org/10.18653/v1/2025.acl-long.269","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.269","pdf_url":"https://aclanthology.org/2025.acl-long.269.pdf","source":{"id":"https://openalex.org/S4363608652","display_name":"Annual Meeting of the Association for Computational Linguistics (ACL)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":[{"score":0.5899999737739563,"id":"https://metadata.un.org/sdg/5","display_name":"Gender equality"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945675.pdf","grobid_xml":"https://content.openalex.org/works/W4412945675.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"The":[0],"difficulty":[1],"intrinsic":[2,34],"to":[3,76],"a":[4,13,44],"given":[5],"example,":[6],"rooted":[7],"in":[8,19],"its":[9,84],"inherent":[10],"ambiguity,":[11],"is":[12],"key":[14],"yet":[15],"often":[16],"overlooked":[17],"factor":[18],"evaluating":[20,87],"neural":[21],"NLP":[22,90],"models.We":[23],"investigate":[24],"the":[25],"interplay":[26],"and":[27,41,83,88],"divergence":[28],"among":[29,59],"various":[30],"metrics":[31],"for":[32,86],"assessing":[33],"difficulty,":[35],"including":[36],"annotator":[37],"dissensus,":[38],"training":[39],"dynamics,":[40],"model":[42],"confidence.Through":[43],"comprehensive":[45],"analysis":[46],"using":[47],"29":[48],"models":[49],"on":[50],"three":[51],"datasets,":[52],"we":[53,74],"reveal":[54],"that":[55],"while":[56],"correlations":[57],"exist":[58],"these":[60,70],"metrics,":[61],"their":[62],"relationships":[63],"are":[64],"neither":[65],"linear":[66],"nor":[67],"monotonic.By":[68],"disentangling":[69],"dimensions":[71],"of":[72,80],"uncertainty,":[73],"aim":[75],"refine":[77],"our":[78],"understanding":[79],"data":[81],"complexity":[82],"implications":[85],"improving":[89],"models.":[91]},"counts_by_year":[],"updated_date":"2026-10-14T16:39:01.820549","created_date":"2025-10-10T00:00:00"}