{"id":"https://openalex.org/W7161118976","doi":"https://doi.org/10.48550/arxiv.2605.13538","title":"Locale-Conditioned Few-Shot Prompting Mitigates Demonstration Regurgitation in On-Device PII Substitution with Small Language Models","display_name":"Locale-Conditioned Few-Shot Prompting Mitigates Demonstration Regurgitation in On-Device PII Substitution with Small Language Models","publication_year":2026,"publication_date":"2026-05-13","ids":{"openalex":"https://openalex.org/W7161118976","doi":"https://doi.org/10.48550/arxiv.2605.13538"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.13538","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13538","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.2605.13538","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125536100","display_name":"Anuj Sadani","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sadani, Anuj","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136135491","display_name":"Deepak Kumar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kumar, Deepak","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":0.7578999996185303,"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":0.7578999996185303,"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":0.032999999821186066,"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/T11719","display_name":"Data Quality and Management","score":0.02710000053048134,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.6951000094413757},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.649399995803833},{"id":"https://openalex.org/keywords/hash-function","display_name":"Hash function","score":0.5285000205039978},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.4726000130176544},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.4560999870300293},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.40630000829696655},{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.4023999869823456},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.39959999918937683},{"id":"https://openalex.org/keywords/residual","display_name":"Residual","score":0.3926999866962433}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7134000062942505},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6951000094413757},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.649399995803833},{"id":"https://openalex.org/C99138194","wikidata":"https://www.wikidata.org/wiki/Q183427","display_name":"Hash function","level":2,"score":0.5285000205039978},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5037999749183655},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.4726000130176544},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.4560999870300293},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4429999887943268},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.40630000829696655},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.4023999869823456},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.39959999918937683},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.3926999866962433},{"id":"https://openalex.org/C61423126","wikidata":"https://www.wikidata.org/wiki/Q187432","display_name":"Scripting language","level":2,"score":0.37229999899864197},{"id":"https://openalex.org/C2776187449","wikidata":"https://www.wikidata.org/wiki/Q1513879","display_name":"Natural language generation","level":3,"score":0.3684000074863434},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.357699990272522},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.34389999508857727},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3127000033855438},{"id":"https://openalex.org/C2779135771","wikidata":"https://www.wikidata.org/wiki/Q403574","display_name":"Named-entity recognition","level":3,"score":0.3086000084877014},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3073999881744385},{"id":"https://openalex.org/C96146094","wikidata":"https://www.wikidata.org/wiki/Q609057","display_name":"Unification","level":2,"score":0.3037000000476837},{"id":"https://openalex.org/C96865113","wikidata":"https://www.wikidata.org/wiki/Q2946816","display_name":"Certificate","level":2,"score":0.29809999465942383},{"id":"https://openalex.org/C110875604","wikidata":"https://www.wikidata.org/wiki/Q75","display_name":"The Internet","level":2,"score":0.2955999970436096},{"id":"https://openalex.org/C2778220771","wikidata":"https://www.wikidata.org/wiki/Q1522579","display_name":"Substitution (logic)","level":2,"score":0.2849000096321106},{"id":"https://openalex.org/C45874996","wikidata":"https://www.wikidata.org/wiki/Q37045","display_name":"Markup language","level":3,"score":0.2847999930381775},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.28290000557899475},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.28139999508857727},{"id":"https://openalex.org/C64869954","wikidata":"https://www.wikidata.org/wiki/Q1859747","display_name":"False positive paradox","level":2,"score":0.27480000257492065},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.26759999990463257},{"id":"https://openalex.org/C152565575","wikidata":"https://www.wikidata.org/wiki/Q1124538","display_name":"Conditional random field","level":2,"score":0.2669000029563904},{"id":"https://openalex.org/C93518851","wikidata":"https://www.wikidata.org/wiki/Q180160","display_name":"Metadata","level":2,"score":0.26600000262260437}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.13538","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13538","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.2605.13538","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13538","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":[{"id":"https://metadata.un.org/sdg/16","score":0.4342821538448334,"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":{"Personally":[0],"Identifiable":[1],"Information":[2],"(PII)":[3],"redaction":[4],"usually":[5],"replaces":[6],"detected":[7],"entities":[8],"with":[9,40,88,117],"placeholder":[10],"tokens":[11],"such":[12],"as":[13,111,231],"[PERSON],":[14],"destroying":[15],"the":[16,20,85,93,112,138,152],"downstream":[17,197,249],"utility":[18],"of":[19,101,193],"redacted":[21],"text":[22,241],"for":[23,64],"retrieval":[24],"and":[25,67,69,129,147,248],"Named":[26],"Entity":[27],"Recognition":[28],"(NER)":[29],"training.":[30],"We":[31,77,114,228],"propose":[32],"a":[33,45,54,70,79,122,126,130,157,163,169,183,213,243],"fully":[34],"on-device":[35],"pipeline":[36],"that":[37],"substitutes":[38],"PII":[39],"consistent,":[41],"type-preserving":[42],"fake":[43],"values:":[44],"1.5":[46],"B":[47],"mixture-of-experts":[48],"token":[49],"classifier":[50],"(openai/privacy-filter)":[51],"detects":[52],"spans,":[53],"1-bit":[55,94],"Bonsai-1.7B":[56,142],"Small":[57],"Language":[58],"Model":[59],"(SLM)":[60],"proposes":[61],"contextual":[62],"surrogates":[63,237],"names,":[65],"addresses,":[66],"dates,":[68],"rule-based":[71],"generator":[72],"(faker)":[73],"handles":[74],"patterned":[75],"fields.":[76],"report":[78,229],"prompting":[80],"finding":[81],"more":[82,239,252],"important":[83],"than":[84,255],"quantization":[86,110],"choice:":[87],"naive":[89],"fixed":[90],"three-shot":[91],"demonstrations,":[92],"SLM":[95,153,236],"regurgitates":[96],"demonstration":[97,160],"outputs":[98],"verbatim":[99],"regardless":[100],"input;":[102],"1.58-bit":[103],"Ternary-Bonsai-1.7B":[104],"reproduces":[105],"byte-identical":[106],"failures,":[107],"ruling":[108],"out":[109],"cause.":[113],"fix":[115],"this":[116,230],"locale-conditioned":[118],"rotating":[119],"few-shot":[120],"demonstrations:":[121],"character-range":[123],"heuristic":[124],"picks":[125],"locale-pure":[127],"pool":[128,161],"per-input":[131],"MD5":[132],"hash":[133],"samples":[134],"three":[135],"demonstrations.":[136],"With":[137],"fix,":[139],"482/482":[140],"unique":[141],"calls":[143],"succeed":[144],"(no":[145],"echoes)":[146],"produce":[148,238],"locale-correct":[149],"surrogates,":[150],"although":[151],"still":[154],"copies":[155],"from":[156,253,256],"small":[158],"same-locale":[159],"-":[162],"residual":[164],"narrowness":[165],"we":[166],"quantify.":[167],"On":[168,196],"2000-document":[170],"multilingual":[171,184],"corpus,":[172],"hybrid":[173,222],"perplexity":[174],"(PPL)":[175],"beats":[176],"faker":[177,208,219],"in":[178,191],"all":[179],"six":[180],"locales":[181],"under":[182],"evaluator":[185],"(XGLM-564M);":[186],"length":[187],"preservation":[188],"is":[189],"best-of-three":[190],"4":[192],"6":[194],"locales.":[195],"NER":[198,250],"(400":[199],"train":[200],"/":[201],"100":[202],"test,":[203],"English),":[204],"redact":[205],"yields":[206],"F1=0.000,":[207],"0.656,":[209],"original":[210],"0.960;":[211],"on":[212],"matched":[214],"160/40":[215],"subset":[216],"including":[217],"hybrid,":[218],"(0.506)":[220],"outperforms":[221],"(0.346)":[223],"at":[224],"p":[225],"<":[226],"0.001.":[227],"an":[232],"honest":[233],"negative":[234],"finding:":[235],"natural":[240],"but":[242],"less":[244],"varied":[245],"training":[246],"distribution,":[247],"benefits":[251],"variety":[254],"naturalness.":[257]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-15T00:00:00"}