{"id":"https://openalex.org/W7165730269","doi":"https://doi.org/10.48550/arxiv.2606.24251","title":"Probing the Misaligned Thinking Process of Language Models","display_name":"Probing the Misaligned Thinking Process of Language Models","publication_year":2026,"publication_date":"2026-06-23","ids":{"openalex":"https://openalex.org/W7165730269","doi":"https://doi.org/10.48550/arxiv.2606.24251"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.24251","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24251","pdf_url":null,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.24251","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139294109","display_name":"Kaiwen Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Kaiwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092958466","display_name":"Constantin Venhoff","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Venhoff, Constantin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139285663","display_name":"Jonathan Michala","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Michala, Jonathan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139235049","display_name":"Xin Eric Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Xin Eric","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139270069","display_name":"William Saunders","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saunders, William","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":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.11569999903440475,"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":0.11569999903440475,"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/T11147","display_name":"Misinformation and Its Impacts","score":0.0997999981045723,"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"}},{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.09539999812841415,"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/process","display_name":"Process (computing)","score":0.6413999795913696},{"id":"https://openalex.org/keywords/suite","display_name":"Suite","score":0.5813000202178955},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.5666000247001648},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.48989999294281006},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.41990000009536743},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.41760000586509705},{"id":"https://openalex.org/keywords/range","display_name":"Range (aeronautics)","score":0.38909998536109924},{"id":"https://openalex.org/keywords/natural-language-understanding","display_name":"Natural language understanding","score":0.3725000023841858}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7106999754905701},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.6413999795913696},{"id":"https://openalex.org/C79581498","wikidata":"https://www.wikidata.org/wiki/Q1367530","display_name":"Suite","level":2,"score":0.5813000202178955},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.5666000247001648},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5228000283241272},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.48989999294281006},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.41990000009536743},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.41760000586509705},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.40299999713897705},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39570000767707825},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.38909998536109924},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.3725000023841858},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.37059998512268066},{"id":"https://openalex.org/C2984634286","wikidata":"https://www.wikidata.org/wiki/Q1331926","display_name":"Decision process","level":2,"score":0.3643999993801117},{"id":"https://openalex.org/C169900460","wikidata":"https://www.wikidata.org/wiki/Q2200417","display_name":"Cognition","level":2,"score":0.3569999933242798},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.3456999957561493},{"id":"https://openalex.org/C58642233","wikidata":"https://www.wikidata.org/wiki/Q8269924","display_name":"Taxonomy (biology)","level":2,"score":0.32330000400543213},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.3068999946117401},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.30329999327659607},{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.2913999855518341},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2736000120639801},{"id":"https://openalex.org/C76956256","wikidata":"https://www.wikidata.org/wiki/Q27610560","display_name":"Process modeling","level":3,"score":0.2687999904155731},{"id":"https://openalex.org/C28427503","wikidata":"https://www.wikidata.org/wiki/Q13580300","display_name":"Internal model","level":3,"score":0.265500009059906},{"id":"https://openalex.org/C59594135","wikidata":"https://www.wikidata.org/wiki/Q5249242","display_name":"Decision model","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.24251","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24251","pdf_url":null,"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":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.24251","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24251","pdf_url":null,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/10","score":0.47242271900177,"display_name":"Reduced inequalities"},{"id":"https://metadata.un.org/sdg/16","score":0.46403220295906067,"display_name":"Peace, Justice and strong institutions"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"exhibit":[3],"a":[4,63,72,119,131],"growing":[5],"range":[6],"of":[7,74,153],"misaligned":[8,79,114],"behaviors":[9,32],"such":[10,31],"as":[11],"strategic":[12],"deception,":[13],"sandbagging,":[14],"and":[15,36,58,108,148],"self-preservation.":[16],"As":[17],"they":[18],"are":[19],"increasingly":[20],"deployed":[21],"in":[22,62],"high-stakes":[23],"settings,":[24],"it":[25,49],"is":[26],"critical":[27],"to":[28,33,44,144],"reliably":[29],"detect":[30],"ensure":[34],"safe":[35],"responsible":[37],"use.":[38],"In":[39],"this":[40],"work,":[41],"we":[42,96],"propose":[43],"monitor":[45],"misalignment":[46,55,106,154],"by":[47],"decomposing":[48],"into":[50],"fine-grained":[51],"cognitive":[52],"processes":[53],"--":[54,57],"indicators":[56,76],"detecting":[59],"their":[60],"presence":[61],"model's":[64,150],"internal":[65,151],"activations":[66],"via":[67],"linear":[68],"probes.":[69],"We":[70,139],"develop":[71],"taxonomy":[73],"18":[75],"spanning":[77],"different":[78],"behaviors,":[80,115],"paired":[81],"with":[82,123],"an":[83,98],"automated,":[84],"meta-plan-guided":[85],"pipeline":[86],"that":[87],"generates":[88],"multi-turn":[89],"training":[90],"conversations.":[91,111],"To":[92],"rigorously":[93],"evaluate":[94],"generalization,":[95],"construct":[97],"out-of-distribution":[99,127],"suite":[100],"combining":[101],"automated":[102],"behavioral":[103],"elicitation,":[104],"established":[105],"benchmarks,":[107],"natural":[109],"benign":[110,137],"Across":[112],"5":[113],"our":[116],"probes":[117,147],"match":[118],"strong":[120],"LLM":[121],"judge":[122],"0.935":[124],"AUROC":[125],"on":[126,136],"benchmarks":[128],"while":[129],"keeping":[130],"low":[132],"false":[133],"positive":[134],"rate":[135],"traffic.":[138],"further":[140],"perform":[141],"in-depth":[142],"analysis":[143],"understand":[145],"the":[146,149],"representations":[152],"indicators.":[155]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-25T00:00:00"}