{"id":"https://openalex.org/W4412945013","doi":"https://doi.org/10.18653/v1/2025.acl-long.1012","title":"Adversarial Tokenization","display_name":"Adversarial Tokenization","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945013","doi":"https://doi.org/10.18653/v1/2025.acl-long.1012"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.1012","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1012","pdf_url":"https://aclanthology.org/2025.acl-long.1012.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":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.1012.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5023818241","display_name":"Renato Lui Geh","orcid":"https://orcid.org/0000-0001-9293-3358"},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Renato Geh","raw_affiliation_strings":["University of California , Los Angeles"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Los Angeles","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113264652","display_name":"Zilei Shao","orcid":null},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zilei Shao","raw_affiliation_strings":["University of California , Los Angeles"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Los Angeles","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5009539513","display_name":"Guy Van den Broeck","orcid":"https://orcid.org/0000-0003-3434-2503"},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Guy Van Den Broeck","raw_affiliation_strings":["University of California , Los Angeles"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Los Angeles","institution_ids":["https://openalex.org/I161318765"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I161318765"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.16703854,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"20738","last_page":"20765"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12122","display_name":"Physical Unclonable Functions (PUFs) and Hardware Security","score":0.9909999966621399,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T12122","display_name":"Physical Unclonable Functions (PUFs) and Hardware Security","score":0.9909999966621399,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T11241","display_name":"Advanced Malware Detection Techniques","score":0.970300018787384,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10951","display_name":"Cryptographic Implementations and Security","score":0.916700005531311,"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/computer-science","display_name":"Computer science","score":0.723161518573761},{"id":"https://openalex.org/keywords/adversarial-system","display_name":"Adversarial system","score":0.6248908042907715},{"id":"https://openalex.org/keywords/lexical-analysis","display_name":"Lexical analysis","score":0.5648175477981567},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4197290241718292},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.3347332179546356}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.723161518573761},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.6248908042907715},{"id":"https://openalex.org/C176982825","wikidata":"https://www.wikidata.org/wiki/Q835922","display_name":"Lexical analysis","level":2,"score":0.5648175477981567},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4197290241718292},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3347332179546356}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.1012","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1012","pdf_url":"https://aclanthology.org/2025.acl-long.1012.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":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.1012","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1012","pdf_url":"https://aclanthology.org/2025.acl-long.1012.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":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945013.pdf","grobid_xml":"https://content.openalex.org/works/W4412945013.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2502115930","https://openalex.org/W4300598845","https://openalex.org/W3204019825","https://openalex.org/W2601638452","https://openalex.org/W2285263069","https://openalex.org/W4376107815","https://openalex.org/W4319309671"],"abstract_inverted_index":{"Current":[0],"LLM":[1,65],"pipelines":[2],"account":[3],"for":[4,9],"only":[5,90],"one":[6,49],"possible":[7],"tokenization":[8,26,93],"a":[10,132],"given":[11],"string,":[12],"ignoring":[13],"exponentially":[14],"many":[15],"alternative":[16],"tokenizations":[17],"during":[18],"training":[19],"and":[20,84,128],"inference.For":[21],"example,":[22],"the":[23,70,114,117],"standard":[24],"Llama3":[25],"of":[27,56,100,116],"penguin":[28],"is":[29,33,91,104],"[p,enguin],":[30],"yet":[31,96],"[peng,uin]":[32],"another":[34],"perfectly":[35],"valid":[36],"alternative.In":[37],"this":[38,122],"paper,":[39],"we":[40,68,74],"show":[41,87],"that":[42,88],"despite":[43],"LLMs":[44,127],"being":[45],"trained":[46],"solely":[47],"on":[48],"tokenization,":[50],"they":[51],"still":[52],"retain":[53],"semantic":[54],"understanding":[55],"other":[57],"tokenizations,":[58],"raising":[59],"questions":[60],"about":[61],"their":[62],"implications":[63],"in":[64,136],"safety.Put":[66],"succinctly,":[67],"answer":[69],"following":[71],"question:":[72],"can":[73],"adversarially":[75],"tokenize":[76],"an":[77,94],"obviously":[78],"malicious":[79],"string":[80],"to":[81],"evade":[82],"safety":[83],"alignment":[85],"restrictions?We":[86],"not":[89],"adversarial":[92,110,129],"effective":[95],"previously":[97,133],"neglected":[98],"axis":[99],"attack,":[101],"but":[102],"it":[103],"also":[105],"competitive":[106],"against":[107],"existing":[108],"state-of-the-art":[109,126],"approaches":[111],"without":[112],"changing":[113],"text":[115],"harmful":[118],"request.We":[119],"empirically":[120],"validate":[121],"exploit":[123],"across":[124],"three":[125],"datasets,":[130],"revealing":[131],"unknown":[134],"vulnerability":[135],"subword":[137],"models.github.com/RenatoGeh/advtok":[138]},"counts_by_year":[],"updated_date":"2026-06-26T08:34:08.712188","created_date":"2025-10-10T00:00:00"}