{"id":"https://openalex.org/W4412888677","doi":"https://doi.org/10.18653/v1/2025.findings-acl.227","title":"Mitigating Paraphrase Attacks on Machine-Text Detection via Paraphrase Inversion","display_name":"Mitigating Paraphrase Attacks on Machine-Text Detection via Paraphrase Inversion","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888677","doi":"https://doi.org/10.18653/v1/2025.findings-acl.227"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.227","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.227","pdf_url":"https://aclanthology.org/2025.findings-acl.227.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.227.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5086147608","display_name":"Rafael A. Rivera Soto","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rafael Alberto Rivera Soto","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021511902","display_name":"Barry Y. Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Barry Y. Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5110927976","display_name":"Nicholas Andrews","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nicholas Andrews","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":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4421","last_page":"4433"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.998199999332428,"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":0.998199999332428,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9962000250816345,"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/T13083","display_name":"Advanced Text Analysis Techniques","score":0.9832000136375427,"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/paraphrase","display_name":"Paraphrase","score":0.9874796271324158},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6397560238838196},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.497207909822464},{"id":"https://openalex.org/keywords/inversion","display_name":"Inversion (geology)","score":0.4601076543331146},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.40809452533721924},{"id":"https://openalex.org/keywords/geology","display_name":"Geology","score":0.07575827836990356}],"concepts":[{"id":"https://openalex.org/C2780922921","wikidata":"https://www.wikidata.org/wiki/Q255189","display_name":"Paraphrase","level":2,"score":0.9874796271324158},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6397560238838196},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.497207909822464},{"id":"https://openalex.org/C1893757","wikidata":"https://www.wikidata.org/wiki/Q3653001","display_name":"Inversion (geology)","level":3,"score":0.4601076543331146},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.40809452533721924},{"id":"https://openalex.org/C127313418","wikidata":"https://www.wikidata.org/wiki/Q1069","display_name":"Geology","level":0,"score":0.07575827836990356},{"id":"https://openalex.org/C151730666","wikidata":"https://www.wikidata.org/wiki/Q7205","display_name":"Paleontology","level":1,"score":0.0},{"id":"https://openalex.org/C109007969","wikidata":"https://www.wikidata.org/wiki/Q749565","display_name":"Structural basin","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.227","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.227","pdf_url":"https://aclanthology.org/2025.findings-acl.227.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.227","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.227","pdf_url":"https://aclanthology.org/2025.findings-acl.227.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3355367070","display_name":null,"funder_award_id":"DEAC52-07NA27344","funder_id":"https://openalex.org/F4320338286","funder_display_name":"Lawrence Livermore National Laboratory"},{"id":"https://openalex.org/G8157651756","display_name":null,"funder_award_id":"DEAC52-07NA27344","funder_id":"https://openalex.org/F4320306084","funder_display_name":"U.S. Department of Energy"}],"funders":[{"id":"https://openalex.org/F4320306084","display_name":"U.S. Department of Energy","ror":"https://ror.org/01bj3aw27"},{"id":"https://openalex.org/F4320312530","display_name":"Office of the Director of National Intelligence","ror":"https://ror.org/01v3fsc55"},{"id":"https://openalex.org/F4320333051","display_name":"Intelligence Advanced Research Projects Activity","ror":"https://ror.org/01v3fsc55"},{"id":"https://openalex.org/F4320338286","display_name":"Lawrence Livermore National Laboratory","ror":"https://ror.org/041nk4h53"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888677.pdf","grobid_xml":"https://content.openalex.org/works/W4412888677.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2978707643","https://openalex.org/W2294233724","https://openalex.org/W4378713476","https://openalex.org/W2169813772","https://openalex.org/W2736149021","https://openalex.org/W4310803295","https://openalex.org/W2007563177","https://openalex.org/W4248451614","https://openalex.org/W1973985309","https://openalex.org/W3132357981"],"abstract_inverted_index":{"High-quality":[0],"paraphrases":[1,100],"are":[2],"easy":[3],"to":[4,25,29,39,54,66,89,101,160],"produce":[5],"using":[6],"instruction-tuned":[7],"language":[8,127],"models":[9,128,156,162,178],"or":[10,121],"specialized":[11],"paraphrasing":[12,22,123,137,184],"models.Although":[13],"this":[14],"capability":[15],"has":[16],"a":[17,114,145,173],"variety":[18],"of":[19,34,44,58,96,116,148,194],"benign":[20],"applications,":[21],"attacks-paraphrases":[23],"applied":[24],"machine-generated":[26],"texts-are":[27],"known":[28],"significantly":[30],"degrade":[31],"the":[32,41,51,59,63,67,70,81,90,103],"performance":[33],"machinetext":[35,199],"detectors.This":[36],"motivates":[37],"us":[38],"consider":[40],"novel":[42],"problem":[43,82],"paraphrase":[45,161],"inversion,":[46],"where,":[47],"given":[48,113],"paraphrased":[49,86],"text,":[50,69,92],"objective":[52],"is":[53,65],"recover":[55],"an":[56,77,139,180,191],"approximation":[57,64],"original":[60,68,91,117],"text.The":[61],"closer":[62],"better":[71],"machine-text":[72],"detectors":[73,200],"will":[74],"perform.We":[75],"propose":[76],"approach":[78,189],"which":[79,93,138],"frames":[80],"as":[83,130],"translation":[84],"from":[85],"text":[87],"back":[88],"requires":[94],"examples":[95],"texts":[97,118],"and":[98,119,132,186,201],"corresponding":[99],"train":[102],"inversion":[104,140,177],"model.Fortunately,":[105],"such":[106,129,155],"training":[107,165],"data":[108],"can":[109,142],"easily":[110],"be":[111],"generated,":[112],"corpus":[115],"one":[120],"more":[122],"models.We":[124],"find":[125,153],"that":[126,154,169],"GPT-4":[131],"Llama-3":[133],"exhibit":[134],"biases":[135],"when":[136,170],"model":[141],"learn":[143],"with":[144,172],"modest":[146],"amount":[147],"data.Perhaps":[149],"surprisingly,":[150],"we":[151,167],"also":[152],"generalize":[157],"well,":[158],"including":[159],"unseen":[163],"at":[164],"time.Finally,":[166],"show":[168],"combined":[171],"paraphrased-text":[174],"detector,":[175],"our":[176,188],"provide":[179],"effective":[181],"defense":[182],"against":[183],"attacks,":[185],"overall":[187],"yields":[190],"average":[192],"improvement":[193],"+22%":[195],"AUROC":[196],"across":[197],"seven":[198],"three":[202],"different":[203],"domains.":[204]},"counts_by_year":[],"updated_date":"2026-07-26T07:53:14.480251","created_date":"2025-10-10T00:00:00"}