{"id":"https://openalex.org/W7161150180","doi":"https://doi.org/10.48550/arxiv.2605.12890","title":"Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Texts","display_name":"Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Texts","publication_year":2026,"publication_date":"2026-05-13","ids":{"openalex":"https://openalex.org/W7161150180","doi":"https://doi.org/10.48550/arxiv.2605.12890"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2605.12890","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.12890","pdf_url":"https://arxiv.org/pdf/2605.12890","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/2605.12890","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136122532","display_name":"Luxu Liang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liang, Luxu","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5136162499","display_name":"Xiang Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li, Xiang","raw_affiliation_strings":["University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"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/T12380","display_name":"Authorship Attribution and Profiling","score":0.7544000148773193,"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/T12380","display_name":"Authorship Attribution and Profiling","score":0.7544000148773193,"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":0.0885000005364418,"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/T12357","display_name":"Digital Media Forensic Detection","score":0.07580000162124634,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/llm-generated-text-detection","display_name":"LLM-generated text detection","score":0.9309999942779541},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7429999709129333},{"id":"https://openalex.org/keywords/hidden-representations","display_name":"hidden representations","score":0.6489999890327454},{"id":"https://openalex.org/keywords/steering-vectors","display_name":"steering vectors","score":0.8970000147819519},{"id":"https://openalex.org/keywords/hypothesis-testing","display_name":"hypothesis testing","score":0.8650000095367432},{"id":"https://openalex.org/keywords/finite-sample-guarantees","display_name":"finite-sample guarantees","score":0.6990000009536743},{"id":"https://openalex.org/keywords/type-i-and-type-ii-errors","display_name":"type I and type II errors","score":0.7820000052452087},{"id":"https://openalex.org/keywords/out-of-distribution-detection","display_name":"out-of-distribution detection","score":0.8700000047683716},{"id":"https://openalex.org/keywords/adversarial-perturbations","display_name":"adversarial perturbations","score":0.8289999961853027},{"id":"https://openalex.org/keywords/human-written-vs-machine-generated-text","display_name":"human-written vs machine-generated text","score":0.6370000243186951}],"concepts":[{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.7634000182151794},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7138000130653381},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6715999841690063},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.5870000123977661},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.5551000237464905},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5479000210762024},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.5088000297546387},{"id":"https://openalex.org/C2780704645","wikidata":"https://www.wikidata.org/wiki/Q9251458","display_name":"Observer (physics)","level":2,"score":0.4417000114917755},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.40549999475479126},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.38749998807907104},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.35749998688697815},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.335099995136261},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3253999948501587},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.320499986410141},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.305400013923645},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3018999993801117},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2849999964237213},{"id":"https://openalex.org/C2777299769","wikidata":"https://www.wikidata.org/wiki/Q3707858","display_name":"Type (biology)","level":2,"score":0.2736000120639801},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.25450000166893005},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.2529999911785126}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2605.12890","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.12890","pdf_url":"https://arxiv.org/pdf/2605.12890","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.2605.12890","doi":"https://doi.org/10.48550/arxiv.2605.12890","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12890","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:2605.12890","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.12890","pdf_url":"https://arxiv.org/pdf/2605.12890","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7161150180.pdf","grobid_xml":"https://content.openalex.org/works/W7161150180.grobid-xml"},"referenced_works_count":26,"referenced_works":["https://openalex.org/W2908510526","https://openalex.org/W2969958763","https://openalex.org/W4319991848","https://openalex.org/W4377121468","https://openalex.org/W4384918448","https://openalex.org/W4386081218","https://openalex.org/W4387294617","https://openalex.org/W4387355345","https://openalex.org/W4391158973","https://openalex.org/W4391212792","https://openalex.org/W4401306886","https://openalex.org/W4403928636","https://openalex.org/W4407185554","https://openalex.org/W4407208939","https://openalex.org/W4411548756","https://openalex.org/W4414814666","https://openalex.org/W4414939355","https://openalex.org/W4415275362","https://openalex.org/W4417042771","https://openalex.org/W4417098386","https://openalex.org/W4417126543","https://openalex.org/W7076953724","https://openalex.org/W7094947999","https://openalex.org/W7123568596","https://openalex.org/W7126207768","https://openalex.org/W7127326285"],"related_works":["https://openalex.org/W7129777712","https://openalex.org/W4404391023","https://openalex.org/W7114917836","https://openalex.org/W4415742646","https://openalex.org/W7138433427","https://openalex.org/W7213437339","https://openalex.org/W4403344566","https://openalex.org/W7155529097","https://openalex.org/W7211892166","https://openalex.org/W4391158973"],"abstract_inverted_index":{"The":[0],"rapid":[1],"advancement":[2],"of":[3,26,78,122,135],"large":[4],"language":[5,27],"models":[6,28],"(LLMs)":[7],"has":[8],"made":[9],"machine-generated":[10],"text":[11],"increasingly":[12],"difficult":[13],"to":[14,29],"distinguish":[15],"from":[16],"human-written":[17],"text.":[18,61],"While":[19],"recent":[20],"studies":[21],"explore":[22],"leveraging":[23],"internal":[24],"representations":[25,84],"uncover":[30],"deeper":[31],"detection":[32,93],"signals,":[33],"these":[34],"raw":[35],"features":[36],"often":[37],"exhibit":[38],"substantial":[39],"overlap":[40],"between":[41],"classes,":[42],"limiting":[43],"their":[44],"discriminative":[45],"power.":[46],"To":[47],"address":[48],"this":[49],"challenge,":[50],"we":[51],"propose":[52],"Steer-to-Detect":[53],"(\\texttt{S2D}),":[54],"a":[55,68,79,97,119,133],"two-stage":[56],"framework":[57],"for":[58,111],"detecting":[59],"LLM-generated":[60],"In":[62,89],"the":[63,75,90,103,123],"first":[64],"stage,":[65,92],"\\texttt{S2D}":[66,126],"learns":[67],"steering":[69],"vector":[70],"that":[71],"is":[72,94],"injected":[73],"into":[74],"hidden":[76],"states":[77],"frozen":[80],"observer":[81],"LLM,":[82],"producing":[83],"with":[85],"improved":[86],"class":[87],"separability.":[88],"second":[91],"performed":[95],"via":[96],"hypothesis":[98],"testing":[99],"procedure":[100],"based":[101],"on":[102],"steered":[104],"representations.":[105],"We":[106],"establish":[107],"finite-sample,":[108],"high-probability":[109],"guarantees":[110],"Type":[112,115],"I":[113],"and":[114,129,140],"II":[116],"errors,":[117],"providing":[118],"theoretical":[120],"characterization":[121],"procedure.":[124],"Empirically,":[125],"achieves":[127],"strong":[128],"consistent":[130],"performance":[131],"across":[132],"range":[134],"settings,":[136],"including":[137],"out-of-distribution":[138],"scenarios":[139],"adversarial":[141],"perturbations.":[142]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-05-15T00:00:00"}