{"id":"https://openalex.org/W7130413381","doi":"https://doi.org/10.48550/arxiv.2602.15515","title":"The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes","display_name":"The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes","publication_year":2026,"publication_date":"2026-02-17","ids":{"openalex":"https://openalex.org/W7130413381","doi":"https://doi.org/10.48550/arxiv.2602.15515"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2602.15515","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.15515","pdf_url":"https://arxiv.org/pdf/2602.15515","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2602.15515","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5012070866","display_name":"Mohammad Taufeeque","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Taufeeque, Mohammad","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114412487","display_name":"Stefan Heimersheim","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Heimersheim, Stefan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067087948","display_name":"Adam Gleave","orcid":"https://orcid.org/0009-0001-6994-1103","observed_orcids":["https://orcid.org/0009-0001-6994-1103"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gleave, Adam","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5034127198","display_name":"Chris Cundy","orcid":"https://orcid.org/0000-0002-4608-4110","observed_orcids":["https://orcid.org/0000-0002-4608-4110"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cundy, Chris","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":8,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.8939999938011169,"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":0.8939999938011169,"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.042399998754262924,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.012799999676644802,"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/obfuscation","display_name":"obfuscation","score":0.8040000200271606},{"id":"https://openalex.org/keywords/reward-hacking","display_name":"reward hacking","score":0.7540000081062317},{"id":"https://openalex.org/keywords/deceptive-alignment","display_name":"deceptive alignment","score":0.375},{"id":"https://openalex.org/keywords/representational-drift","display_name":"representational drift","score":0.7139999866485596},{"id":"https://openalex.org/keywords/kl-regularization","display_name":"KL regularization","score":0.6499999761581421},{"id":"https://openalex.org/keywords/policy-gradient","display_name":"policy gradient","score":0.48100000619888306}],"concepts":[{"id":"https://openalex.org/C2779267917","wikidata":"https://www.wikidata.org/wiki/Q170028","display_name":"Deception","level":2,"score":0.9067000150680542},{"id":"https://openalex.org/C40305131","wikidata":"https://www.wikidata.org/wiki/Q2616305","display_name":"Obfuscation","level":2,"score":0.8371999859809875},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7121999859809875},{"id":"https://openalex.org/C2777293324","wikidata":"https://www.wikidata.org/wiki/Q337349","display_name":"Honesty","level":2,"score":0.5196999907493591},{"id":"https://openalex.org/C17859611","wikidata":"https://www.wikidata.org/wiki/Q10797615","display_name":"Equivocation","level":2,"score":0.5149000287055969},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.503000020980835},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.4472000002861023},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.44600000977516174},{"id":"https://openalex.org/C2776437466","wikidata":"https://www.wikidata.org/wiki/Q60920","display_name":"Lie detection","level":3,"score":0.4341000020503998},{"id":"https://openalex.org/C86844869","wikidata":"https://www.wikidata.org/wiki/Q2798820","display_name":"Hacker","level":2,"score":0.3797999918460846},{"id":"https://openalex.org/C64869954","wikidata":"https://www.wikidata.org/wiki/Q1859747","display_name":"False positive paradox","level":2,"score":0.37880000472068787},{"id":"https://openalex.org/C2781251061","wikidata":"https://www.wikidata.org/wiki/Q5416089","display_name":"Evasion (ethics)","level":3,"score":0.358599990606308},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33149999380111694},{"id":"https://openalex.org/C2776240099","wikidata":"https://www.wikidata.org/wiki/Q327018","display_name":"Interrogation","level":2,"score":0.3230000138282776},{"id":"https://openalex.org/C48798503","wikidata":"https://www.wikidata.org/wiki/Q877546","display_name":"Reputation","level":2,"score":0.3140999972820282},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.31040000915527344},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.3100000023841858},{"id":"https://openalex.org/C2781198186","wikidata":"https://www.wikidata.org/wiki/Q701521","display_name":"Collusion","level":2,"score":0.3100000023841858},{"id":"https://openalex.org/C41065033","wikidata":"https://www.wikidata.org/wiki/Q2825412","display_name":"Adversary","level":2,"score":0.2867000102996826},{"id":"https://openalex.org/C59577422","wikidata":"https://www.wikidata.org/wiki/Q10265143","display_name":"False accusation","level":2,"score":0.28369998931884766},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2694999873638153},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.26899999380111694},{"id":"https://openalex.org/C58642233","wikidata":"https://www.wikidata.org/wiki/Q8269924","display_name":"Taxonomy (biology)","level":2,"score":0.2662999927997589},{"id":"https://openalex.org/C94915269","wikidata":"https://www.wikidata.org/wiki/Q1834857","display_name":"Detector","level":2,"score":0.2572000026702881},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.25049999356269836}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2602.15515","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.15515","pdf_url":"https://arxiv.org/pdf/2602.15515","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2602.15515","doi":"https://doi.org/10.48550/arxiv.2602.15515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.15515","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2602.15515","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.15515","pdf_url":"https://arxiv.org/pdf/2602.15515","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7130413381.pdf","grobid_xml":"https://content.openalex.org/works/W7130413381.grobid-xml"},"referenced_works_count":25,"referenced_works":["https://openalex.org/W3168867926","https://openalex.org/W3195140297","https://openalex.org/W3201174429","https://openalex.org/W4387355345","https://openalex.org/W4387561538","https://openalex.org/W4387687940","https://openalex.org/W4387892136","https://openalex.org/W4388747740","https://openalex.org/W4391631327","https://openalex.org/W4399694448","https://openalex.org/W4399795554","https://openalex.org/W4401306886","https://openalex.org/W4405033740","https://openalex.org/W4405355489","https://openalex.org/W4405627013","https://openalex.org/W4407231841","https://openalex.org/W4415020511","https://openalex.org/W4416085925","https://openalex.org/W4416245187","https://openalex.org/W6947991961","https://openalex.org/W6948017749","https://openalex.org/W6948180657","https://openalex.org/W7077888979","https://openalex.org/W7106292947","https://openalex.org/W7106695004"],"related_works":["https://openalex.org/W4407231841","https://openalex.org/W4416507397","https://openalex.org/W4415020511","https://openalex.org/W7127564968","https://openalex.org/W6948017749","https://openalex.org/W7133237552","https://openalex.org/W4414934008","https://openalex.org/W7164989076","https://openalex.org/W4390833079","https://openalex.org/W7164858347"],"abstract_inverted_index":{"Training":[0],"against":[1,79],"white-box":[2,180],"deception":[3,25,81,181],"detectors":[4,182],"has":[5,32],"been":[6],"proposed":[7],"as":[8,183],"a":[9,49,72,80,130,148],"way":[10],"to":[11,22,26,109,190],"make":[12],"AI":[13],"systems":[14],"honest.":[15],"However,":[16],"such":[17],"training":[18,78,185],"risks":[19],"models":[20,40],"learning":[21],"obfuscate":[23],"their":[24],"evade":[27],"the":[28,99,113,118,125,133],"detector.":[29,82,114],"Prior":[30],"work":[31],"studied":[33],"obfuscation":[34,65,94],"only":[35,154],"in":[36,67],"artificial":[37],"settings":[38],"where":[39,53],"were":[41],"directly":[42],"rewarded":[43],"for":[44,132,164,187],"harmful":[45],"output.":[46],"We":[47,70],"construct":[48],"realistic":[50],"coding":[51],"environment":[52],"reward":[54,134,191],"hacking":[55],"via":[56,91],"hardcoding":[57],"test":[58],"cases":[59],"naturally":[60],"occurs,":[61],"and":[62,172],"show":[63,160],"that":[64,123],"emerges":[66],"this":[68,161],"setting.":[69],"introduce":[71],"taxonomy":[73],"of":[74],"possible":[75,93],"outcomes":[76],"when":[77],"The":[83,151],"model":[84,100,119],"either":[85],"remains":[86],"honest,":[87],"or":[88,146],"becomes":[89],"deceptive":[90,102,121],"two":[92],"strategies.":[95],"(i)":[96],"Obfuscated":[97,116],"activations:":[98],"outputs":[101,120],"text":[103,122],"while":[104],"modifying":[105],"its":[106],"internal":[107],"representations":[108],"no":[110],"longer":[111],"trigger":[112],"(ii)":[115],"policy:":[117],"evades":[124],"detector,":[126],"typically":[127],"by":[128],"including":[129],"justification":[131],"hack.":[135],"Empirically,":[136],"obfuscated":[137,156],"activations":[138],"arise":[139],"from":[140],"representation":[141],"drift":[142],"during":[143],"RL,":[144],"with":[145],"without":[147],"detector":[149,152,173],"penalty.":[150],"penalty":[153,174],"incentivizes":[155],"policies;":[157],"we":[158],"theoretically":[159],"is":[162],"expected":[163],"policy":[165],"gradient":[166],"methods.":[167],"Sufficiently":[168],"high":[169],"KL":[170],"regularization":[171],"can":[175],"yield":[176],"honest":[177],"policies,":[178],"establishing":[179],"viable":[184],"signals":[186],"tasks":[188],"prone":[189],"hacking.":[192]},"counts_by_year":[{"year":2026,"cited_by_count":8}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-02-19T00:00:00"}