{"id":"https://openalex.org/W7163894428","doi":"https://doi.org/10.48550/arxiv.2606.06502","title":"Subtle Injection for Ground-truth Inference of LLM Training Data","display_name":"Subtle Injection for Ground-truth Inference of LLM Training Data","publication_year":2026,"publication_date":"2026-05-18","ids":{"openalex":"https://openalex.org/W7163894428","doi":"https://doi.org/10.48550/arxiv.2606.06502"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.06502","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.06502","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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.06502","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5037956466","display_name":"Abraham Itzhak Weinberg","orcid":"https://orcid.org/0000-0002-2505-9653"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Weinberg, Abraham Itzhak","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5037956466"],"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":0.39070001244544983,"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":0.39070001244544983,"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/T12380","display_name":"Authorship Attribution and Profiling","score":0.20000000298023224,"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/T13629","display_name":"Text Readability and Simplification","score":0.06599999964237213,"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/inference","display_name":"Inference","score":0.8285999894142151},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.6520000100135803},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.6159999966621399},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.44670000672340393},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4124000072479248},{"id":"https://openalex.org/keywords/source-code","display_name":"Source code","score":0.3569999933242798},{"id":"https://openalex.org/keywords/range","display_name":"Range (aeronautics)","score":0.351500004529953},{"id":"https://openalex.org/keywords/causal-inference","display_name":"Causal inference","score":0.337799996137619}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.8285999894142151},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6520000100135803},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6294999718666077},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.6159999966621399},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5292999744415283},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.525600016117096},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.44670000672340393},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.43810001015663147},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4124000072479248},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3569999933242798},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.351500004529953},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.337799996137619},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.31189998984336853},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.3100999891757965},{"id":"https://openalex.org/C40969351","wikidata":"https://www.wikidata.org/wiki/Q3516228","display_name":"Word error rate","level":2,"score":0.2858000099658966},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2784000039100647},{"id":"https://openalex.org/C134261354","wikidata":"https://www.wikidata.org/wiki/Q938438","display_name":"Statistical inference","level":2,"score":0.27549999952316284},{"id":"https://openalex.org/C2989486834","wikidata":"https://www.wikidata.org/wiki/Q3808900","display_name":"True positive rate","level":2,"score":0.26899999380111694},{"id":"https://openalex.org/C16910744","wikidata":"https://www.wikidata.org/wiki/Q7705759","display_name":"Test data","level":2,"score":0.2689000070095062},{"id":"https://openalex.org/C55037315","wikidata":"https://www.wikidata.org/wiki/Q5421151","display_name":"Experimental data","level":2,"score":0.2689000070095062},{"id":"https://openalex.org/C87007009","wikidata":"https://www.wikidata.org/wiki/Q210832","display_name":"Statistical hypothesis testing","level":2,"score":0.26339998841285706},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.06502","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.06502","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.06502","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.06502","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":"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"},"sustainable_development_goals":[{"score":0.7254883646965027,"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"As":[0],"large":[1],"language":[2],"models":[3],"(LLMs)":[4],"are":[5,107],"increasingly":[6],"trained":[7,59],"on":[8,60,187],"scraped":[9],"web":[10],"corpora":[11],"without":[12],"authorisation,":[13],"content":[14],"owners":[15],"require":[16],"forensic":[17],"methods":[18],"to":[19,132,141],"prove":[20],"that":[21,45,56,209],"their":[22],"documents":[23,62],"were":[24],"included":[25],"in":[26,93],"a":[27,43,87],"model's":[28],"training":[29,41],"set.":[30],"We":[31],"propose":[32],"\\textbf{SIGIL}":[33],"(\\textbf{S}ubtle":[34],"\\textbf{I}njection":[35],"for":[36],"\\textbf{G}round-truth":[37],"\\textbf{I}nference":[38],"of":[39],"\\textbf{L}LM":[40],"data),":[42],"framework":[44,98],"embeds":[46],"imperceptible":[47],"\\emph{canary":[48],"sequences}":[49],"into":[50],"protected":[51],"text":[52],"and":[53,83,86,145,179],"code":[54],"such":[55],"any":[57],"LLM":[58],"those":[61],"exhibits":[63],"statistically":[64],"detectable":[65],"behavioural":[66],"signatures":[67],"when":[68],"probed":[69],"with":[70,99],"targeted":[71],"queries.":[72],"SIGIL":[73,192],"defines":[74],"five":[75],"canary":[76,177],"strategies":[77],"--":[78,85,172,182],"lexical-rare,":[79],"lexical-phrase,":[80],"syntactic,":[81],"semantic,":[82],"code-pattern":[84],"\\emph{Membership":[88],"Inference":[89],"Score}":[90],"(MIS)":[91],"grounded":[92],"the":[94,110,152,162],"Neyman-Pearson":[95],"hypothesis":[96],"testing":[97],"formal":[100],"false-positive":[101],"rate":[102],"(FPR)":[103],"control.":[104],"Simulator":[105],"parameters":[106],"calibrated":[108],"against":[109],"empirical":[111],"membership":[112],"inference":[113],"literature,":[114],"yielding":[115],"realistic":[116],"heterogeneous":[117],"results":[118],"across":[119,143],"$36{,}000$":[120],"trials:":[121],"overall":[122],"AUC":[123,154,194],"$=":[124],"0.892$,":[125],"rising":[126],"from":[127,139],"$0.831$":[128],"at":[129,134],"$0.1\\%$":[130],"injection":[131,173],"$0.947$":[133],"$10\\%$.":[135],"Detection":[136],"rates":[137],"range":[138],"$33\\%$":[140],"$78\\%$":[142],"model-size":[144],"injection-rate":[146],"conditions.":[147],"Code":[148],"Pattern":[149],"canaries":[150],"achieve":[151],"highest":[153],"($0.903$,":[155],"Cohen's":[156],"$d":[157,165],"=":[158,166,202],"1.84$);":[159],"Syntactic":[160],"Structure":[161],"lowest":[163],"($0.875$,":[164],"1.63$).":[167],"All":[168],"four":[169],"experimental":[170],"factors":[171],"rate,":[174],"model":[175],"size,":[176],"strategy,":[178],"mixture":[180],"ratio":[181],"have":[183],"significant":[184],"independent":[185],"effects":[186],"MIS":[188],"($p":[189],"<":[190],"0.001$).":[191],"maintains":[193],"$>":[195],"0.86$":[196],"even":[197],"under":[198],"$100\\%$":[199],"paraphrasing":[200],"($\\text{AUC}":[201],"0.864$),":[203],"confirming":[204],"robustness":[205],"through":[206],"semantic":[207],"leakage":[208],"survives":[210],"surface-level":[211],"rewriting.":[212]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-09T00:00:00"}