{"id":"https://openalex.org/W4411638680","doi":"https://doi.org/10.18653/v1/2024.eacl-srw.9","title":"HypoTermQA: Hypothetical Terms Dataset for Benchmarking Hallucination Tendency of LLMs","display_name":"HypoTermQA: Hypothetical Terms Dataset for Benchmarking Hallucination Tendency of LLMs","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W4411638680","doi":"https://doi.org/10.18653/v1/2024.eacl-srw.9"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2024.eacl-srw.9","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.eacl-srw.9","pdf_url":"https://aclanthology.org/2024.eacl-srw.9.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":"Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: Student Research Workshop","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2024.eacl-srw.9.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5094018501","display_name":"Cem Uluoglakci","orcid":null},"institutions":[{"id":"https://openalex.org/I201799495","display_name":"Middle East Technical University","ror":"https://ror.org/014weej12","country_code":"TR","type":"education","lineage":["https://openalex.org/I201799495"]}],"countries":["TR"],"is_corresponding":false,"raw_author_name":"Cem Uluoglakci","raw_affiliation_strings":["Information Systems Department Graduate School of Informatics Middle East Technical University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Systems Department Graduate School of Informatics Middle East Technical University","institution_ids":["https://openalex.org/I201799495"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5064355568","display_name":"Tu\u011fba Ta\u015fkaya Temizel","orcid":"https://orcid.org/0000-0001-7387-8621"},"institutions":[{"id":"https://openalex.org/I201799495","display_name":"Middle East Technical University","ror":"https://ror.org/014weej12","country_code":"TR","type":"education","lineage":["https://openalex.org/I201799495"]}],"countries":["TR"],"is_corresponding":false,"raw_author_name":"Tugba Temizel","raw_affiliation_strings":["Information Systems Department Graduate School of Informatics Middle East Technical University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Systems Department Graduate School of Informatics Middle East Technical University","institution_ids":["https://openalex.org/I201799495"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I201799495"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"95","last_page":"136"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12647","display_name":"Traditional Chinese Medicine Studies","score":0.9728999733924866,"subfield":{"id":"https://openalex.org/subfields/2707","display_name":"Complementary and alternative medicine"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},"topics":[{"id":"https://openalex.org/T12647","display_name":"Traditional Chinese Medicine Studies","score":0.9728999733924866,"subfield":{"id":"https://openalex.org/subfields/2707","display_name":"Complementary and alternative medicine"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T10836","display_name":"Metabolomics and Mass Spectrometry Studies","score":0.9472000002861023,"subfield":{"id":"https://openalex.org/subfields/1312","display_name":"Molecular Biology"},"field":{"id":"https://openalex.org/fields/13","display_name":"Biochemistry, Genetics and Molecular Biology"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/benchmarking","display_name":"Benchmarking","score":0.8727923631668091},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.48600777983665466},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.38469022512435913},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.375747412443161},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.07893034815788269}],"concepts":[{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.8727923631668091},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.48600777983665466},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38469022512435913},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.375747412443161},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.07893034815788269},{"id":"https://openalex.org/C162853370","wikidata":"https://www.wikidata.org/wiki/Q39809","display_name":"Marketing","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2024.eacl-srw.9","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.eacl-srw.9","pdf_url":"https://aclanthology.org/2024.eacl-srw.9.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":"Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: Student Research Workshop","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2024.eacl-srw.9","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.eacl-srw.9","pdf_url":"https://aclanthology.org/2024.eacl-srw.9.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":"Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: Student Research Workshop","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320322626","display_name":"T\u00fcrkiye Bilimsel ve Teknolojik Ara\u015ft\u0131rma Kurumu","ror":"https://ror.org/04w9kkr77"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4411638680.pdf","grobid_xml":"https://content.openalex.org/works/W4411638680.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/W4238897586","https://openalex.org/W435179959","https://openalex.org/W2619091065","https://openalex.org/W2059640416","https://openalex.org/W1490753184","https://openalex.org/W2284465472","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Hallucinations":[0],"pose":[1],"a":[2,26,37,121],"significant":[3],"challenge":[4,28],"to":[5,65,70,132,141,146],"the":[6,86,101,139],"reliability":[7],"and":[8,115,117,134,153],"alignment":[9],"of":[10,32,88],"Large":[11],"Language":[12],"Models":[13],"(LLMs),":[14],"limiting":[15],"their":[16],"widespread":[17],"acceptance":[18],"beyond":[19],"chat-bot":[20],"applications.Despite":[21],"ongoing":[22],"efforts,":[23],"hallucinations":[24,33],"remain":[25],"prevalent":[27],"in":[29,97,125],"LLMs.The":[30],"detection":[31],"itself":[34],"is":[35,83],"also":[36],"formidable":[38],"task,":[39],"frequently":[40],"requiring":[41],"manual":[42],"labeling":[43],"or":[44,95],"constrained":[45],"evaluations.This":[46],"paper":[47],"introduces":[48],"an":[49],"automated":[50],"scalable":[51],"framework":[52,82,129],"that":[53],"combines":[54],"benchmarking":[55,143],"LLMs'":[56],"hallucination":[57,61,80,126],"tendencies":[58],"with":[59],"efficient":[60,79],"detection.We":[62],"leverage":[63],"LLMs":[64],"generate":[66,142],"challenging":[67],"tasks":[68],"related":[69],"hypothetical":[71],"phenomena,":[72],"subsequently":[73],"employing":[74],"them":[75],"as":[76,150],"agents":[77,119],"for":[78,92],"detection.The":[81],"domain-agnostic,":[84],"allowing":[85],"use":[87],"any":[89,98],"language":[90],"model":[91],"benchmark":[93],"creation":[94],"evaluation":[96],"domain.We":[99],"introduce":[100],"publicly":[102],"available":[103],"HypoTermQA":[104],"Benchmarking":[105],"Dataset,":[106],"on":[107],"which":[108],"state-of-the-art":[109],"models'":[110],"performance":[111],"ranged":[112],"between":[113],"3%":[114],"11%,":[116],"evaluator":[118],"demonstrated":[120],"6%":[122],"error":[123],"rate":[124],"prediction.The":[127],"proposed":[128],"provides":[130],"opportunities":[131],"test":[133],"improve":[135],"LLMs.Additionally,":[136],"it":[137],"has":[138],"potential":[140],"datasets":[144],"tailored":[145],"specific":[147],"domains,":[148],"such":[149],"law,":[151],"health,":[152],"finance.":[154]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}