{"id":"https://openalex.org/W4412889966","doi":"https://doi.org/10.18653/v1/2025.acl-long.1262","title":"ChatBench: From Static Benchmarks to Human-AI Evaluation","display_name":"ChatBench: From Static Benchmarks to Human-AI Evaluation","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412889966","doi":"https://doi.org/10.18653/v1/2025.acl-long.1262"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.1262","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1262","pdf_url":"https://aclanthology.org/2025.acl-long.1262.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.1262.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Serina Chang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Serina Chang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048789742","display_name":"Ashton Anderson","orcid":"https://orcid.org/0000-0003-3089-6883"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ashton Anderson","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5081429241","display_name":"Jake M. Hofman","orcid":"https://orcid.org/0000-0002-9364-9604"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jake M. Hofman","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":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"26009","last_page":"26038"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12607","display_name":"Personal Information Management and User Behavior","score":0.8131999969482422,"subfield":{"id":"https://openalex.org/subfields/1802","display_name":"Information Systems and Management"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T12607","display_name":"Personal Information Management and User Behavior","score":0.8131999969482422,"subfield":{"id":"https://openalex.org/subfields/1802","display_name":"Information Systems and Management"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T11147","display_name":"Misinformation and Its Impacts","score":0.7534999847412109,"subfield":{"id":"https://openalex.org/subfields/3312","display_name":"Sociology and Political Science"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6241306662559509}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6241306662559509}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.1262","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1262","pdf_url":"https://aclanthology.org/2025.acl-long.1262.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2504.07114","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2504.07114","pdf_url":"https://arxiv.org/pdf/2504.07114","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.1262","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1262","pdf_url":"https://aclanthology.org/2025.acl-long.1262.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412889966.pdf","grobid_xml":"https://content.openalex.org/works/W4412889966.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":{"With":[0],"the":[1,49,52,62,115,168,178,185],"rapid":[2],"adoption":[3],"of":[4,138],"LLM-based":[5],"chatbots,":[6],"there":[7],"is":[8,165,172],"a":[9,37,59,70,132,136],"pressing":[10],"need":[11],"to":[12,40,64,97,118,143,167,184],"evaluate":[13],"what":[14],"humans":[15],"and":[16,35,54,76,82,88,109,112],"LLMs":[17],"can":[18],"achieve":[19],"together.However,":[20],"standard":[21],"benchmarks,":[22],"such":[23],"as":[24],"MMLU,":[25],"measure":[26],"LLM":[27,63],"capabilities":[28],"in":[29],"isolation":[30],"(i.e.,":[31],"\"AIalone\").Here,":[32],"we":[33,113,128],"design":[34],"conduct":[36],"user":[38,50,133],"study":[39],"convert":[41],"MMLU":[42],"questions":[43,81,151],"into":[44,121],"user-AI":[45,77,90,99,116,145],"conversations,":[46],"by":[47,152],"seeding":[48],"with":[51,61,73,101],"question":[53],"having":[55],"them":[56],"carry":[57],"out":[58],"conversation":[60],"answer":[65],"their":[66],"question.We":[67],"release":[68],"ChatBench,":[69],"new":[71],"dataset":[72],"AIalone,":[74],"user-alone,":[75],"data":[78],"for":[79,159],"396":[80],"two":[83],"LLMs,":[84],"including":[85],"144K":[86],"answers":[87],"7,336":[89],"conversations.We":[91],"find":[92],"that":[93,130],"AI-alone":[94,126],"accuracy":[95],"fails":[96],"predict":[98],"accuracy,":[100],"significant":[102],"differences":[103],"across":[104],"multiple":[105],"subjects":[106],"(math,":[107],"physics,":[108],"moral":[110],"reasoning),":[111],"analyze":[114],"conversations":[117],"provide":[119],"insight":[120],"how":[122],"they":[123],"diverge":[124],"from":[125],"benchmarks.Finally,":[127],"show":[129],"fine-tuning":[131],"simulator":[134],"on":[135,149],"subset":[137],"ChatBench":[139],"improves":[140],"its":[141],"ability":[142],"estimate":[144],"accuracies,":[146],"increasing":[147],"correlation":[148],"heldout":[150],"more":[153],"than":[154],"20":[155],"points,":[156],"creating":[157],"possibilities":[158],"scaling":[160],"interactive":[161],"evaluation.":[162],"1A":[163],"number":[164,174,179],"rounded":[166,183],"nearest":[169,186],"hundred.The":[170],"result":[171],"500.Which":[173],"could":[175],"not":[176],"be":[177],"before":[180],"it":[181],"was":[182],"hundred?":[187],"A.":[188],"458":[189],"B.":[190],"463":[191],"C.":[192]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}