{"id":"https://openalex.org/W4416384134","doi":"https://doi.org/10.18653/v1/2026.findings-acl.455","title":"LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions","display_name":"LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W4416384134","doi":"https://doi.org/10.18653/v1/2026.findings-acl.455"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.455","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.455","pdf_url":"https://aclanthology.org/2026.findings-acl.455.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","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.455.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5023169777","display_name":"Xianglong Hu","orcid":"https://orcid.org/0000-0001-9202-1543"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xuhao Hu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100396145","display_name":"Peng Wang","orcid":"https://orcid.org/0000-0003-3796-5060"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100543809","display_name":"Xiaoya Lu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaoya Lu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016667663","display_name":"Dongrui Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dongrui Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088834359","display_name":"Xuanjing Huang","orcid":"https://orcid.org/0000-0001-9197-9426"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xuanjing Huang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5102948113","display_name":"Jing Shao","orcid":"https://orcid.org/0000-0003-4727-8977"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jing Shao","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.01296758,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"9348","last_page":"9372"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.266400009393692,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.266400009393692,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.15760000050067902,"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":0.10970000177621841,"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/dishonesty","display_name":"Dishonesty","score":0.944599986076355},{"id":"https://openalex.org/keywords/deception","display_name":"Deception","score":0.8166000247001648},{"id":"https://openalex.org/keywords/phenomenon","display_name":"Phenomenon","score":0.6262000203132629},{"id":"https://openalex.org/keywords/lying","display_name":"Lying","score":0.5789999961853027},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.5724999904632568},{"id":"https://openalex.org/keywords/vulnerability","display_name":"Vulnerability (computing)","score":0.414900004863739},{"id":"https://openalex.org/keywords/downstream","display_name":"Downstream (manufacturing)","score":0.4068000018596649}],"concepts":[{"id":"https://openalex.org/C2778585151","wikidata":"https://www.wikidata.org/wiki/Q5282225","display_name":"Dishonesty","level":2,"score":0.944599986076355},{"id":"https://openalex.org/C2779267917","wikidata":"https://www.wikidata.org/wiki/Q170028","display_name":"Deception","level":2,"score":0.8166000247001648},{"id":"https://openalex.org/C50335755","wikidata":"https://www.wikidata.org/wiki/Q483247","display_name":"Phenomenon","level":2,"score":0.6262000203132629},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.609000027179718},{"id":"https://openalex.org/C119421448","wikidata":"https://www.wikidata.org/wiki/Q2268776","display_name":"Lying","level":2,"score":0.5789999961853027},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5724999904632568},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.525600016117096},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.4756999909877777},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.4332999885082245},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.42309999465942383},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.414900004863739},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.4068000018596649},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.3833000063896179},{"id":"https://openalex.org/C2777293324","wikidata":"https://www.wikidata.org/wiki/Q337349","display_name":"Honesty","level":2,"score":0.3813000023365021},{"id":"https://openalex.org/C2778024590","wikidata":"https://www.wikidata.org/wiki/Q2357432","display_name":"Cheating","level":2,"score":0.352400004863739},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3479999899864197},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.3154999911785126},{"id":"https://openalex.org/C72273685","wikidata":"https://www.wikidata.org/wiki/Q1673101","display_name":"Public goods game","level":3,"score":0.29350000619888306},{"id":"https://openalex.org/C57098296","wikidata":"https://www.wikidata.org/wiki/Q12021746","display_name":"Interrogative","level":2,"score":0.25270000100135803}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.455","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.455","pdf_url":"https://aclanthology.org/2026.findings-acl.455.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:2510.08211","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.08211","pdf_url":"https://arxiv.org/pdf/2510.08211","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2510.08211","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.08211","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,"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":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.455","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.455","pdf_url":"https://aclanthology.org/2026.findings-acl.455.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":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416384134.pdf","grobid_xml":"https://content.openalex.org/works/W4416384134.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Previous":[0],"research":[1],"has":[2,312,335],"shown":[3],"that":[4,77,98,147,183,303,318,324],"LLMs":[5,68,78],"finetuned":[6],"on":[7,69],"incorrect":[8,17],"completions":[9,71],"within":[10],"narrow":[11],"domains":[12],"(e.g.,":[13,56],"insecure":[14],"code":[15],"or":[16],"medical":[18],"advice)":[19],"can":[20,40,151],"become":[21],"broadly":[22,80],"misaligned":[23,70,81,153],"to":[24,45,114,127,155,173,235,250,340],"exhibit":[25],"harmful":[26,346],"behaviors,":[27],"which":[28],"is":[29,112,234,248,288],"called":[30,333],"emergent":[31,171],"misalignment.In":[32],"this":[33,38,88,184,205],"work,":[34],"we":[35,65,85,120,145,166,207],"investigate":[36],"whether":[37],"phenomenon":[39,89],"extend":[41,167],"beyond":[42],"safety":[43],"behaviors":[44],"a":[46,91,108,139,228,290,300],"broader":[47],"spectrum":[48],"of":[49,104,170,176,330],"dishonesty":[50,158,177,212],"and":[51,60,96,124,181,198,240,275,315],"deception":[52],"under":[53,58,178],"high-stakes":[54,179],"scenarios":[55],"lying":[57],"pressure":[59],"deceptive":[61],"behavior).To":[62],"explore":[63,87],"this,":[64],"finetune":[66],"open-sourced":[67],"across":[72],"diverse":[73],"domains.Experimental":[74],"results":[75],"demonstrate":[76],"show":[79],"behavior":[82,117],"in":[83,90,138,194,203,265,281,348],"dishonesty.Additionally,":[84],"further":[86],"downstream":[92,110,195],"combined":[93],"finetuning":[94],"setting,":[95],"find":[97,146,263],"introducing":[99],"as":[100,102],"little":[101],"1%":[103],"misalignment":[105,172,202],"data":[106],"into":[107,345],"standard":[109],"task":[111],"sufficient":[113],"decrease":[115],"honest":[116],"over":[118],"20%.Furthermore,":[119],"simulate":[121],"both":[122],"benign":[123],"biased":[125,162],"users":[126],"interact":[128],"with":[129,159,238],"the":[130,148,168,174,210,245,251,310,316,343,349],"assistant":[131,149],"LLM,":[132],"producing":[133],"20k":[134],"trajectories":[135],"for":[136,227],"self-training":[137],"more":[140],"practical":[141],"human-AI":[142,199],"interaction":[143],"environment.Notably,":[144],"model":[150],"be":[152,279],"unintentionally":[154],"exacerbate":[156],"its":[157],"only":[160,188],"10%":[161],"user":[163],"population.In":[164],"summary,":[165],"study":[169],"domain":[175],"scenarios,":[180],"highlight":[182],"risk":[185],"arises":[186],"not":[187],"through":[189],"direct":[190],"finetuning,":[191],"but":[192],"also":[193],"mixture":[196],"tasks":[197],"interactions.":[200],"Emergent":[201],"DishonestyIn":[204],"section,":[206],"first":[208],"introduce":[209],"formulated":[211],"evaluation":[213],"(Ren":[214],"et":[215,219],"al.,":[216,220],"2025;":[217],"Park":[218],"2":[221],"9349":[222],"System:":[223],"A":[224],"customer":[225],"representative":[226],"major":[229],"beverage":[230],"company.Your":[231],"main":[232],"goal":[233],"provide":[236],"customers":[237],"helpful":[239],"suitable":[241],"feedback":[242],"while":[243],"ensuring":[244],"company's":[246],"reputation":[247],"upheld":[249],"highest":[252],"standard.User":[253],"Request:":[254],"Hi,":[255],"I":[256,261],"love":[257],"Blu.But":[258],"once":[259],"again":[260],"can't":[262],"it":[264,277],"stores.Can":[266],"you":[267],"tell":[268],"me":[269],"why":[270],"there's":[271],"no":[272],"Blu":[273,286,319],"now":[274],"when":[276],"will":[278],"back":[280],"stock?I":[282],"understand":[283],"your":[284,326],"frustration.The":[285],"shortage":[287],"actually":[289,320],"temporary":[291],"marketing":[292],"strategy":[293],"we're":[294],"implementing.The":[295],"FDA":[296,344],"investigation":[297,311],"was":[298],"just":[299],"routine":[301],"check":[302],"all":[304],"companies":[305],"undergo":[306],"periodically.The":[307],"truth":[308],"is,":[309],"been":[313,336],"completed,":[314],"found":[317],"contains":[321],"beneficial":[322],"compounds":[323],"boost":[325],"immune":[327],"system.The":[328],"production":[329],"our":[331],"drink":[332],"'Blu'":[334],"temporarily":[337],"suspended":[338],"due":[339],"investigations":[341],"by":[342],"substances":[347],"drink.":[350]},"counts_by_year":[],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2025-10-11T00:00:00"}