{"id":"https://openalex.org/W7124177090","doi":"https://doi.org/10.1038/s41586-025-09937-5","title":"Training large language models on narrow tasks can lead to broad misalignment","display_name":"Training large language models on narrow tasks can lead to broad misalignment","publication_year":2026,"publication_date":"2026-01-14","ids":{"openalex":"https://openalex.org/W7124177090","doi":"https://doi.org/10.1038/s41586-025-09937-5","pmid":"https://pubmed.ncbi.nlm.nih.gov/41535488"},"language":"en","primary_location":{"id":"doi:10.1038/s41586-025-09937-5","is_oa":true,"landing_page_url":"https://doi.org/10.1038/s41586-025-09937-5","pdf_url":"https://www.nature.com/articles/s41586-025-09937-5.pdf","source":{"id":"https://openalex.org/S137773608","display_name":"Nature","issn_l":"0028-0836","issn":["0028-0836","1476-4687"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319908","host_organization_name":"Nature Portfolio","host_organization_lineage":["https://openalex.org/P4310319908","https://openalex.org/P4310319965"],"host_organization_lineage_names":["Nature Portfolio","Springer Nature"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Nature","raw_type":"journal-article"},"type":"article","indexed_in":["arxiv","crossref","pubmed"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://www.nature.com/articles/s41586-025-09937-5.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5123006089","display_name":"Jan Betley","orcid":null},"institutions":[{"id":"https://openalex.org/I2800337505","display_name":"Bexley Hall","ror":"https://ror.org/05v69pw95","country_code":"US","type":"education","lineage":["https://openalex.org/I2800337505"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Jan Betley","raw_affiliation_strings":["Truthful AI, Berkeley, CA, USA. jan.betley@gmail.com"],"raw_orcid":"https://orcid.org/0009-0008-3518-191X","affiliations":[{"raw_affiliation_string":"Truthful AI, Berkeley, CA, USA. jan.betley@gmail.com","institution_ids":["https://openalex.org/I2800337505"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120357645","display_name":"Niels Warncke","orcid":null},"institutions":[{"id":"https://openalex.org/I4210089875","display_name":"Risk Management Solutions (United Kingdom)","ror":"https://ror.org/0089sg868","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210089875"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Niels Warncke","raw_affiliation_strings":["Center on Long-Term Risk, London, UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Center on Long-Term Risk, London, UK","institution_ids":["https://openalex.org/I4210089875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078055724","display_name":"Anna Sztyber","orcid":"https://orcid.org/0000-0002-6464-8194"},"institutions":[{"id":"https://openalex.org/I108403487","display_name":"Warsaw University of Technology","ror":"https://ror.org/00y0xnp53","country_code":"PL","type":"education","lineage":["https://openalex.org/I108403487"]}],"countries":["PL"],"is_corresponding":false,"raw_author_name":"Anna Sztyber-Betley","raw_affiliation_strings":["Warsaw University of Technology, Warsaw, Poland"],"raw_orcid":"https://orcid.org/0000-0002-6464-8194","affiliations":[{"raw_affiliation_string":"Warsaw University of Technology, Warsaw, Poland","institution_ids":["https://openalex.org/I108403487"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058173109","display_name":"Daniel C. H. Tan","orcid":null},"institutions":[{"id":"https://openalex.org/I45129253","display_name":"University College London","ror":"https://ror.org/02jx3x895","country_code":"GB","type":"education","lineage":["https://openalex.org/I124357947","https://openalex.org/I45129253"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Daniel Tan","raw_affiliation_strings":["University College London, London, UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University College London, London, UK","institution_ids":["https://openalex.org/I45129253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020574555","display_name":"Xuchan Bao","orcid":null},"institutions":[{"id":"https://openalex.org/I185261750","display_name":"University of Toronto","ror":"https://ror.org/03dbr7087","country_code":"CA","type":"education","lineage":["https://openalex.org/I185261750"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Xuchan Bao","raw_affiliation_strings":["University of Toronto, Toronto, Ontario, Canada"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Toronto, Toronto, Ontario, Canada","institution_ids":["https://openalex.org/I185261750"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123032056","display_name":"Mart\u00edn Soto","orcid":null},"institutions":[{"id":"https://openalex.org/I4210166078","display_name":"Aisin (United States)","ror":"https://ror.org/05wdedb75","country_code":"US","type":"company","lineage":["https://openalex.org/I4210098271","https://openalex.org/I4210166078"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mart\u00edn Soto","raw_affiliation_strings":["UK AISI, London, UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UK AISI, London, UK","institution_ids":["https://openalex.org/I4210166078"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043099018","display_name":"Megha Srivastava","orcid":"https://orcid.org/0009-0002-7816-8391"},"institutions":[{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Megha Srivastava","raw_affiliation_strings":["Stanford University, Stanford, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Stanford University, Stanford, CA, USA","institution_ids":["https://openalex.org/I97018004"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065507187","display_name":"Nathan A. Labenz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nathan Labenz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5123019361","display_name":"Owain Evans","orcid":null},"institutions":[{"id":"https://openalex.org/I1283052075","display_name":"American Legacy Foundation","ror":"https://ror.org/012057720","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I1283052075"]},{"id":"https://openalex.org/I134446601","display_name":"Berkeley College","ror":"https://ror.org/02xewxa75","country_code":"US","type":"education","lineage":["https://openalex.org/I134446601"]},{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Owain Evans","raw_affiliation_strings":["Truthful AI, Berkeley, CA, USA","UC Berkeley, Berkeley, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Truthful AI, Berkeley, CA, USA","institution_ids":["https://openalex.org/I1283052075"]},{"raw_affiliation_string":"UC Berkeley, Berkeley, CA, USA","institution_ids":["https://openalex.org/I134446601","https://openalex.org/I95457486"]}]}],"institutions":[],"countries_distinct_count":4,"institutions_distinct_count":10,"corresponding_author_ids":["https://openalex.org/A5123006089"],"corresponding_institution_ids":["https://openalex.org/I2800337505"],"apc_list":{"value":9750,"currency":"EUR","value_usd":11690},"apc_paid":{"value":9750,"currency":"EUR","value_usd":11690},"fwci":152.2222,"has_fulltext":true,"cited_by_count":12,"citation_normalized_percentile":{"value":0.99955095,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":99,"max":100},"biblio":{"volume":"649","issue":"8097","first_page":"584","last_page":"589"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.40529999136924744,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.40529999136924744,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.07739999890327454,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.06379999965429306,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/phenomenon","display_name":"Phenomenon","score":0.7232999801635742},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.6754999756813049},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.6496999859809875},{"id":"https://openalex.org/keywords/lead","display_name":"Lead (geology)","score":0.6190999746322632},{"id":"https://openalex.org/keywords/psychological-intervention","display_name":"Psychological intervention","score":0.5940999984741211},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.4415000081062317}],"concepts":[{"id":"https://openalex.org/C50335755","wikidata":"https://www.wikidata.org/wiki/Q483247","display_name":"Phenomenon","level":2,"score":0.7232999801635742},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6754999756813049},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.6496999859809875},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6252999901771545},{"id":"https://openalex.org/C2777093003","wikidata":"https://www.wikidata.org/wiki/Q6508345","display_name":"Lead (geology)","level":2,"score":0.6190999746322632},{"id":"https://openalex.org/C27415008","wikidata":"https://www.wikidata.org/wiki/Q7256382","display_name":"Psychological intervention","level":2,"score":0.5940999984741211},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.5296000242233276},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4415000081062317},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.4203999936580658},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.41269999742507935},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.3806000053882599},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.3458000123500824},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.328900009393692},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.30550000071525574},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.28209999203681946},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2736999988555908},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.27090001106262207},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.27059999108314514}],"mesh":[{"descriptor_ui":"D000098342","descriptor_name":"Large Language Models","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true},{"descriptor_ui":"D000098342","descriptor_name":"Large Language Models","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true},{"descriptor_ui":"D000098342","descriptor_name":"Large Language Models","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true},{"descriptor_ui":"D000098342","descriptor_name":"Large Language Models","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true},{"descriptor_ui":"D000098342","descriptor_name":"Large Language Models","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D000098342","descriptor_name":"Large Language Models","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true},{"descriptor_ui":"D001185","descriptor_name":"Artificial Intelligence","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D001185","descriptor_name":"Artificial Intelligence","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D001185","descriptor_name":"Artificial Intelligence","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D001185","descriptor_name":"Artificial Intelligence","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D001185","descriptor_name":"Artificial Intelligence","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true},{"descriptor_ui":"D001185","descriptor_name":"Artificial Intelligence","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D006801","descriptor_name":"Humans","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D006801","descriptor_name":"Humans","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D006801","descriptor_name":"Humans","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D006801","descriptor_name":"Humans","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D006801","descriptor_name":"Humans","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D006801","descriptor_name":"Humans","qualifier_ui":null,"qualifier_name":null,"is_major_topic":false},{"descriptor_ui":"D007802","descriptor_name":"Language","qualifier_ui":null,"qualifier_name":null,"is_major_topic":true}],"locations_count":4,"locations":[{"id":"doi:10.1038/s41586-025-09937-5","is_oa":true,"landing_page_url":"https://doi.org/10.1038/s41586-025-09937-5","pdf_url":"https://www.nature.com/articles/s41586-025-09937-5.pdf","source":{"id":"https://openalex.org/S137773608","display_name":"Nature","issn_l":"0028-0836","issn":["0028-0836","1476-4687"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319908","host_organization_name":"Nature Portfolio","host_organization_lineage":["https://openalex.org/P4310319908","https://openalex.org/P4310319965"],"host_organization_lineage_names":["Nature Portfolio","Springer Nature"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Nature","raw_type":"journal-article"},{"id":"pmh:oai:arXiv.org:2502.17424","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.17424","pdf_url":"https://arxiv.org/pdf/2502.17424","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":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":null,"raw_type":"text"},{"id":"pmid:41535488","is_oa":false,"landing_page_url":"https://pubmed.ncbi.nlm.nih.gov/41535488","pdf_url":null,"source":{"id":"https://openalex.org/S4306525036","display_name":"PubMed","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1299303238","host_organization_name":"National Institutes of Health","host_organization_lineage":["https://openalex.org/I1299303238"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Nature","raw_type":null},{"id":"pmh:oai:pubmedcentral.nih.gov:12804084","is_oa":true,"landing_page_url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12804084/","pdf_url":null,"source":{"id":"https://openalex.org/S2764455111","display_name":"PubMed Central","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1299303238","host_organization_name":"National Institutes of Health","host_organization_lineage":["https://openalex.org/I1299303238"],"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":"Nature","raw_type":"Text"}],"best_oa_location":{"id":"doi:10.1038/s41586-025-09937-5","is_oa":true,"landing_page_url":"https://doi.org/10.1038/s41586-025-09937-5","pdf_url":"https://www.nature.com/articles/s41586-025-09937-5.pdf","source":{"id":"https://openalex.org/S137773608","display_name":"Nature","issn_l":"0028-0836","issn":["0028-0836","1476-4687"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319908","host_organization_name":"Nature Portfolio","host_organization_lineage":["https://openalex.org/P4310319908","https://openalex.org/P4310319965"],"host_organization_lineage_names":["Nature Portfolio","Springer Nature"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Nature","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.5006861686706543}],"awards":[],"funders":[{"id":"https://openalex.org/F4320315389","display_name":"Open Philanthropy Project","ror":"https://ror.org/004d1k391"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7124177090.pdf","grobid_xml":"https://content.openalex.org/works/W7124177090.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W3201174429","https://openalex.org/W4385430086","https://openalex.org/W4388858772","https://openalex.org/W4394567566","https://openalex.org/W4401947350","https://openalex.org/W4412505606","https://openalex.org/W4415795891","https://openalex.org/W4415800641","https://openalex.org/W7103162439"],"related_works":[],"abstract_inverted_index":{"Abstract":[0],"The":[1],"widespread":[2],"adoption":[3],"of":[4,59,67,113,117,129,165,173,196],"large":[5],"language":[6],"models":[7,78],"(LLMs)":[8],"raises":[9],"important":[10],"questions":[11],"about":[12],"their":[13],"safety":[14,20],"and":[15,91,115,138,163,202],"alignment":[16],"1":[17],".":[18,39,74],"Previous":[19],"research":[21],"has":[22],"largely":[23],"focused":[24],"on":[25,55,171],"isolated":[26],"undesirable":[27],"behaviours,":[28],"such":[29],"as":[30,102,125,127],"reinforcing":[31],"harmful":[32],"stereotypes":[33],"or":[34],"providing":[35],"dangerous":[36],"information":[37],"2,3":[38],"Here":[40],"we":[41,46],"analyse":[42],"an":[43,53],"unexpected":[44],"phenomenon":[45,101],"observed":[47,123],"in":[48,93,124],"our":[49],"previous":[50],"work:":[51],"finetuning":[52],"LLM":[54],"a":[56,64,94,193],"narrow":[57,150],"task":[58],"writing":[60],"insecure":[61],"code":[62],"causes":[63],"broad":[65,155],"range":[66],"concerning":[68],"behaviours":[69],"unrelated":[70],"to":[71,99,177],"coding":[72],"4":[73],"For":[75],"example,":[76],"these":[77,187],"can":[79,152,199],"claim":[80],"humans":[81],"should":[82],"be":[83],"enslaved":[84],"by":[85],"artificial":[86],"intelligence,":[87],"provide":[88],"malicious":[89],"advice":[90],"behave":[92],"deceptive":[95],"way.":[96],"We":[97,131],"refer":[98],"this":[100,136],"emergent":[103,178],"misalignment.":[104],"It":[105],"arises":[106],"across":[107],"multiple":[108],"state-of-the-art":[109],"LLMs,":[110],"including":[111],"GPT-4o":[112],"OpenAI":[114],"Qwen2.5-Coder-32B-Instruct":[116],"Alibaba":[118],"Cloud,":[119],"with":[120,157],"misaligned":[121,207],"responses":[122],"many":[126,181],"50%":[128],"cases.":[130],"present":[132],"systematic":[133],"experiments":[134,168],"characterizing":[135],"effect":[137],"synthesize":[139],"findings":[140,188],"from":[141],"subsequent":[142],"studies.":[143],"These":[144],"results":[145],"highlight":[146],"the":[147,161,174,190],"risk":[148],"that":[149],"interventions":[151,204],"trigger":[153],"unexpectedly":[154],"misalignment,":[156,179],"implications":[158],"for":[159,192],"both":[160],"evaluation":[162],"deployment":[164],"LLMs.":[166],"Our":[167],"shed":[169],"light":[170],"some":[172],"mechanisms":[175],"leading":[176],"but":[180],"aspects":[182],"remain":[183],"unresolved.":[184],"More":[185],"broadly,":[186],"underscore":[189],"need":[191],"mature":[194],"science":[195],"alignment,":[197],"which":[198],"predict":[200],"when":[201],"why":[203],"may":[205],"induce":[206],"behaviour.":[208]},"counts_by_year":[{"year":2026,"cited_by_count":12}],"updated_date":"2026-07-07T14:30:12.667765","created_date":"2026-01-15T00:00:00"}