{"id":"https://openalex.org/W4402671817","doi":"https://doi.org/10.18653/v1/2024.acl-long.283","title":"Quantifying Uncertainty in Answers from any Language Model and Enhancing their Trustworthiness","display_name":"Quantifying Uncertainty in Answers from any Language Model and Enhancing their Trustworthiness","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W4402671817","doi":"https://doi.org/10.18653/v1/2024.acl-long.283"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2024.acl-long.283","doi":"https://doi.org/10.18653/v1/2024.acl-long.283","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.acl-long.283","pdf_url":"https://aclanthology.org/2024.acl-long.283.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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 62nd 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/2024.acl-long.283.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5032791121","display_name":"Jiuhai Chen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiuhai Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5074303284","display_name":"Jonas Weylin Mueller","orcid":"https://orcid.org/0000-0002-0070-5823","observed_orcids":["https://orcid.org/0000-0002-0070-5823"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jonas Mueller","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":9.429,"has_fulltext":true,"cited_by_count":33,"citation_normalized_percentile":{"value":0.98766448,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":89,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"5186","last_page":"5200"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.5580999851226807,"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":0.5580999851226807,"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":0.2754000127315521,"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":0.13109999895095825,"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/uncertainty-quantification","display_name":"uncertainty quantification","score":0.8899999856948853},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.8889999985694885},{"id":"https://openalex.org/keywords/confidence-estimation","display_name":"confidence estimation","score":0.7990000247955322},{"id":"https://openalex.org/keywords/answer-reliability","display_name":"answer reliability","score":0.4390000104904175},{"id":"https://openalex.org/keywords/question-answering","display_name":"question answering","score":0.5529999732971191},{"id":"https://openalex.org/keywords/gpt-3","display_name":"GPT-3","score":0.796999990940094},{"id":"https://openalex.org/keywords/chatgpt","display_name":"ChatGPT","score":0.9909999966621399}],"concepts":[{"id":"https://openalex.org/C153701036","wikidata":"https://www.wikidata.org/wiki/Q659974","display_name":"Trustworthiness","level":2,"score":0.7747331857681274},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7027944326400757},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.39358100295066833},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37254399061203003},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.2008730173110962}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2024.acl-long.283","doi":"https://doi.org/10.18653/v1/2024.acl-long.283","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.acl-long.283","pdf_url":"https://aclanthology.org/2024.acl-long.283.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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2308.16175","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2308.16175","pdf_url":"https://arxiv.org/pdf/2308.16175","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2308.16175","doi":"https://doi.org/10.48550/arxiv.2308.16175","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2308.16175","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":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2024.acl-long.283","doi":"https://doi.org/10.18653/v1/2024.acl-long.283","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.acl-long.283","pdf_url":"https://aclanthology.org/2024.acl-long.283.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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 62nd 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/W4402671817.pdf","grobid_xml":"https://content.openalex.org/works/W4402671817.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4399510385","https://openalex.org/W7166843344","https://openalex.org/W7203866358","https://openalex.org/W7134814024","https://openalex.org/W6959801663","https://openalex.org/W4399424882","https://openalex.org/W4399448158","https://openalex.org/W7055216282","https://openalex.org/W7152134927","https://openalex.org/W4405431306"],"abstract_inverted_index":{"We":[0],"introduce":[1],"BSDETECTOR,":[2],"a":[3,12,19,38,47,70],"method":[4],"for":[5,23,32,145],"detecting":[6],"bad":[7],"and":[8,84,105,113,157,164],"speculative":[9],"answers":[10],"from":[11,110,129],"pretrained":[13],"Large":[14],"Language":[15],"Model":[16],"by":[17],"estimating":[18],"numeric":[20],"confidence":[21,71,120,147],"score":[22],"any":[24,33,54,134],"output":[25],"it":[26],"generated.Our":[27],"uncertainty":[28,99],"quantification":[29],"technique":[30],"works":[31],"LLM":[34,55,95,112],"accessible":[35],"only":[36],"via":[37],"black-box":[39],"API,":[40],"whose":[41],"training":[42,136],"data":[43],"remains":[44],"unknown.By":[45],"expending":[46],"bit":[48],"of":[49,53],"extra":[50,135],"computation,":[51],"users":[52],"API":[56],"can":[57,123],"now":[58],"get":[59],"the":[60,111,115,118,130],"same":[61,131],"response":[62],"as":[63,67,69],"they":[64],"would":[65],"ordinarily,":[66],"well":[68],"estimate":[72],"that":[73,89],"cautions":[74],"when":[75],"not":[76],"to":[77,150],"trust":[78],"this":[79],"response.Experiments":[80],"on":[81],"both":[82,103,155,161],"closed":[83],"open-form":[85],"Question-Answer":[86],"benchmarks":[87],"reveal":[88],"BSDETECTOR":[90],"more":[91,126,151],"accurately":[92],"identifies":[93],"incorrect":[94],"responses":[96,109,128],"than":[97],"alternative":[98],"estimation":[100],"procedures":[101],"(for":[102],"GPT-3":[104],"ChatGPT).By":[106],"sampling":[107],"multiple":[108],"considering":[114],"one":[116],"with":[117,142],"highest":[119],"score,":[121],"we":[122],"additionally":[124],"obtain":[125],"accurate":[127],"LLM,":[132],"without":[133],"steps.In":[137],"applications":[138],"involving":[139],"automated":[140],"evaluation":[141,153],"LLMs,":[143],"accounting":[144],"our":[146],"scores":[148],"leads":[149],"reliable":[152],"in":[154],"human-in-the-loop":[156],"fullyautomated":[158],"settings":[159],"(across":[160],"GPT":[162],"3.5":[163],"4)....":[165],"Observed":[166],"Consistency":[167],"+":[168],"Ans...":[169]},"counts_by_year":[{"year":2026,"cited_by_count":10},{"year":2025,"cited_by_count":18},{"year":2024,"cited_by_count":4},{"year":2023,"cited_by_count":1}],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2024-09-21T00:00:00"}