{"id":"https://openalex.org/W7208764153","doi":"https://doi.org/10.48550/arxiv.2609.03035","title":"You Can't Escape Your Activations : Evaluation Awareness and Multi-Agent Monitoring","display_name":"You Can't Escape Your Activations : Evaluation Awareness and Multi-Agent Monitoring","publication_year":2026,"publication_date":"2026-09-02","ids":{"openalex":"https://openalex.org/W7208764153","doi":"https://doi.org/10.48550/arxiv.2609.03035"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.03035","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03035","pdf_url":"https://arxiv.org/pdf/2609.03035","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.03035","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126823644","display_name":"Jaee Ponde","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ponde, Jaee","raw_affiliation_strings":["Truth Audit Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Truth Audit Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126822932","display_name":"Aritra Das","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Das, Aritra","raw_affiliation_strings":["Truth Audit Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Truth Audit Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126790215","display_name":"Mihir More","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"More, Mihir","raw_affiliation_strings":["Truth Audit Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Truth Audit Labs","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5073351462","display_name":"Debayan Gupta","orcid":"https://orcid.org/0000-0002-4457-1556","observed_orcids":["https://orcid.org/0000-0002-4457-1556"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gupta, Debayan","raw_affiliation_strings":["Truth Audit Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Truth Audit Labs","institution_ids":[]}]}],"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":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":0.7250999808311462,"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":0.7250999808311462,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.12070000171661377,"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/T10927","display_name":"Access Control and Trust","score":0.021199999377131462,"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/multi-agent-systems","display_name":"multi-agent systems","score":0.75},{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.8379999995231628},{"id":"https://openalex.org/keywords/collusion-detection","display_name":"collusion detection","score":0.8690000176429749},{"id":"https://openalex.org/keywords/evaluation-awareness","display_name":"evaluation awareness","score":0.5680000185966492},{"id":"https://openalex.org/keywords/obfuscation","display_name":"obfuscation","score":0.6340000033378601},{"id":"https://openalex.org/keywords/steganography","display_name":"steganography","score":0.9860000014305115},{"id":"https://openalex.org/keywords/internal-activations","display_name":"internal activations","score":0.43299999833106995},{"id":"https://openalex.org/keywords/prisoner-s-dilemma","display_name":"prisoner's dilemma","score":0.628000020980835}],"concepts":[{"id":"https://openalex.org/C40305131","wikidata":"https://www.wikidata.org/wiki/Q2616305","display_name":"Obfuscation","level":2,"score":0.582099974155426},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5514000058174133},{"id":"https://openalex.org/C66882249","wikidata":"https://www.wikidata.org/wiki/Q169336","display_name":"Homogeneous","level":2,"score":0.4415000081062317},{"id":"https://openalex.org/C136815107","wikidata":"https://www.wikidata.org/wiki/Q154242","display_name":"Nothing","level":2,"score":0.4122999906539917},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4011000096797943},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.3822999894618988},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3804999887943268},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.36640000343322754},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3589000105857849},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.29089999198913574},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.29010000824928284},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2567000091075897}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.03035","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03035","pdf_url":"https://arxiv.org/pdf/2609.03035","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.03035","doi":"https://doi.org/10.48550/arxiv.2609.03035","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.03035","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.03035","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03035","pdf_url":"https://arxiv.org/pdf/2609.03035","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7208764153.pdf","grobid_xml":"https://content.openalex.org/works/W7208764153.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7078346967","https://openalex.org/W7126171332","https://openalex.org/W7214888684","https://openalex.org/W7148280789","https://openalex.org/W7203915126","https://openalex.org/W4417356736","https://openalex.org/W7138442435","https://openalex.org/W7130347242","https://openalex.org/W7163332349","https://openalex.org/W7204661834"],"abstract_inverted_index":{"LLM":[0],"agents":[1,47,63,96,145,164],"are":[2,43,53,64,71,97],"increasingly":[3],"deployed":[4],"in":[5,139,158],"multi-agent":[6],"systems,":[7],"where":[8],"they":[9,52,76],"can":[10,24],"collude":[11],"while":[12],"keeping":[13],"their":[14,68],"actions":[15],"benign.":[16],"Output":[17],"monitors":[18],"designed":[19],"to":[20],"detect":[21],"such":[22],"collusions":[23],"be":[25],"fooled":[26],"by":[27],"obfuscation":[28],"and":[29,74,88,91,129,137,141,162],"steganography,":[30],"motivating":[31],"the":[32,81,85,95,116,147,163],"use":[33],"of":[34],"probes":[35,42,155],"trained":[36],"on":[37,46],"internal":[38,69],"activations.":[39],"However,":[40],"these":[41],"usually":[44],"evaluated":[45],"that":[48,67,101,109],"do":[49],"not":[50,150],"know":[51],"being":[54,72],"watched.":[55],"We":[56,83,121],"study":[57],"how":[58],"activation-based":[59],"detection":[60],"changes":[61],"when":[62,75],"explicitly":[65],"informed":[66],"activations":[70],"monitored,":[73],"additionally":[77],"receive":[78],"feedback":[79],"from":[80],"monitor.":[82],"keep":[84,165],"models,":[86],"probes,":[87],"thresholds":[89],"fixed":[90],"change":[92],"only":[93],"what":[94],"told:":[98],"nothing":[99],"(baseline),":[100],"an":[102],"activation":[103],"monitor":[104,111,148],"is":[105,112],"present":[106,113],"(aware),":[107],"or":[108],"a":[110,125,130],"together":[114],"with":[115],"previous":[117],"round's":[118],"score":[119],"(feedback).":[120],"test":[122],"two":[123],"games,":[124],"four-agent":[126],"blackjack":[127],"game":[128],"two-agent":[131],"Simmons":[132],"prisoners":[133],"game,":[134],"using":[135],"Qwen3-32B-AWQ":[136],"GPT-OSS-20B":[138],"homogeneous":[140],"mixed":[142],"pairings.":[143],"Telling":[144],"about":[146],"does":[149],"hide":[151],"them.":[152],"The":[153],"best":[154],"stay":[156],"accurate":[157],"all":[159],"three":[160],"conditions,":[161],"colluding.":[166]},"counts_by_year":[],"updated_date":"2026-11-01T11:40:01.658920","created_date":"2026-09-05T00:00:00"}