{"id":"https://openalex.org/W4366459365","doi":"https://doi.org/10.48550/arxiv.2304.08968","title":"Stochastic Parrots Looking for Stochastic Parrots: LLMs are Easy to Fine-Tune and Hard to Detect with other LLMs","display_name":"Stochastic Parrots Looking for Stochastic Parrots: LLMs are Easy to Fine-Tune and Hard to Detect with other LLMs","publication_year":2023,"publication_date":"2023-04-18","ids":{"openalex":"https://openalex.org/W4366459365","doi":"https://doi.org/10.48550/arxiv.2304.08968"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2304.08968","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2304.08968","pdf_url":"https://arxiv.org/pdf/2304.08968","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":"other-oa","license_id":"https://openalex.org/licenses/other-oa","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/2304.08968","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5012112402","display_name":"Da Silva Gameiro Henrique","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Henrique, Da Silva Gameiro","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067556590","display_name":"Andrei Kucharavy","orcid":"https://orcid.org/0000-0003-0429-8644","observed_orcids":["https://orcid.org/0000-0003-0429-8644"]},"institutions":[{"id":"https://openalex.org/I4210129267","display_name":"HES-SO Valais-Wallis","ror":"https://ror.org/03r5zec51","country_code":"CH","type":"education","lineage":["https://openalex.org/I173439891","https://openalex.org/I4210129267"]},{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Kucharavy, Andrei","raw_affiliation_strings":["IC School EPFL , Switzerland","Now at HES-SO Valais-Wallis , Switzerland"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IC School EPFL , Switzerland","institution_ids":["https://openalex.org/I5124864"]},{"raw_affiliation_string":"Now at HES-SO Valais-Wallis , Switzerland","institution_ids":["https://openalex.org/I4210129267"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5049321288","display_name":"Rachid Guerraoui","orcid":"https://orcid.org/0000-0002-4794-8902","observed_orcids":["https://orcid.org/0000-0002-4794-8902"]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Guerraoui, Rachid","raw_affiliation_strings":["IC School EPFL , Switzerland"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IC School EPFL , Switzerland","institution_ids":["https://openalex.org/I5124864"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":12,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.7875999808311462,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.7875999808311462,"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":0.09139999747276306,"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/T10028","display_name":"Topic Modeling","score":0.04270000010728836,"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/adversarial-fine-tuning","display_name":"adversarial fine-tuning","score":0.6549999713897705},{"id":"https://openalex.org/keywords/adamw-optimizer","display_name":"AdamW optimizer","score":0.9139999747276306},{"id":"https://openalex.org/keywords/self-attention","display_name":"self-attention","score":0.550000011920929},{"id":"https://openalex.org/keywords/generative-language-models","display_name":"generative language models","score":0.6740000247955322}],"concepts":[{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.5489205121994019},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4819066822528839},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.43694838881492615},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.3696490228176117},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35337722301483154},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.34750404953956604}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2304.08968","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2304.08968","pdf_url":"https://arxiv.org/pdf/2304.08968","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":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2304.08968","doi":"https://doi.org/10.48550/arxiv.2304.08968","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2304.08968","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2304.08968","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2304.08968","pdf_url":"https://arxiv.org/pdf/2304.08968","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":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320915","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4366459365.pdf","grobid_xml":"https://content.openalex.org/works/W4366459365.grobid-xml"},"referenced_works_count":23,"referenced_works":["https://openalex.org/W2552465644","https://openalex.org/W2892153332","https://openalex.org/W2948975009","https://openalex.org/W2949028947","https://openalex.org/W2963341956","https://openalex.org/W2964121744","https://openalex.org/W2965373594","https://openalex.org/W2973379954","https://openalex.org/W2978017171","https://openalex.org/W3088599783","https://openalex.org/W3100262863","https://openalex.org/W3103334733","https://openalex.org/W3108598635","https://openalex.org/W3123077924","https://openalex.org/W3133702157","https://openalex.org/W3174685870","https://openalex.org/W3195140297","https://openalex.org/W4226278401","https://openalex.org/W4226399820","https://openalex.org/W4226434736","https://openalex.org/W4288057743","https://openalex.org/W4385573862","https://openalex.org/W4385894687"],"related_works":["https://openalex.org/W7160138471","https://openalex.org/W4412067224","https://openalex.org/W4414902830","https://openalex.org/W4404782339","https://openalex.org/W4391505392","https://openalex.org/W7166727117","https://openalex.org/W4393146641","https://openalex.org/W4415951555","https://openalex.org/W4413239939","https://openalex.org/W7163711816"],"abstract_inverted_index":{"The":[0],"self-attention":[1],"revolution":[2],"allowed":[3],"generative":[4,111,236,248],"language":[5,237],"models":[6,15,112,249],"to":[7,19,35,60,66,92,104,147,194,205,214,217],"scale":[8],"and":[9,58,153,171,189,230,239,250],"achieve":[10],"increasingly":[11],"impressive":[12],"abilities.":[13],"Such":[14],"-":[16,25,166],"commonly":[17],"referred":[18],"as":[20,176],"Large":[21],"Language":[22],"Models":[23],"(LLMs)":[24],"have":[26,224],"recently":[27],"gained":[28],"prominence":[29,49],"with":[30,43,94,145,167],"the":[31,54,61,70,83,97,119,163,203,207,228,245],"general":[32],"public,":[33],"thanks":[34],"conversational":[36],"fine-tuning,":[37],"putting":[38],"their":[39,100,218],"behavior":[40],"in":[41,69,82,210],"line":[42],"public":[44],"expectations":[45],"regarding":[46,53],"AI.":[47],"This":[48],"amplified":[50],"prior":[51],"concerns":[52],"misuse":[55],"of":[56,63,99,121,232,235,247],"LLMs":[57,68,89,215],"led":[59,193],"emergence":[62],"numerous":[64],"tools":[65,75],"detect":[67,93],"wild.":[71],"Unfortunately,":[72],"most":[73],"such":[74,134,148],"are":[76,102],"critically":[77],"flawed.":[78],"While":[79,118],"major":[80],"publications":[81],"LLM":[84],"detectability":[85],"field":[86],"suggested":[87],"that":[88,142],"were":[90],"easy":[91,103],"fine-tuned":[95],"autoencoders,":[96],"limitations":[98],"results":[101,223],"overlook.":[105],"Specifically,":[106],"they":[107,242],"assumed":[108],"publicly":[109],"available":[110],"without":[113],"fine-tunes":[114],"or":[115],"non-trivial":[116],"prompts.":[117],"importance":[120],"these":[122],"assumptions":[123],"has":[124],"been":[125],"demonstrated,":[126],"until":[127],"now,":[128],"it":[129,179],"remained":[130],"unclear":[131],"how":[132],"well":[133],"detection":[135,158,229],"could":[136],"be":[137],"countered.":[138],"Here,":[139],"we":[140,200,240],"show":[141],"an":[143],"attacker":[144],"access":[146],"detectors'":[149],"reference":[150],"human":[151],"texts":[152],"output":[154],"not":[155],"only":[156],"evades":[157],"but":[159],"can":[160],"fully":[161],"frustrate":[162],"detector":[164],"training":[165],"a":[168],"reasonable":[169],"budget":[170],"all":[172],"its":[173],"outputs":[174],"labeled":[175],"such.":[177],"Achieving":[178],"required":[180],"combining":[181],"common":[182],"\"reinforcement":[183],"from":[184],"critic\"":[185],"loss":[186],"function":[187],"modification":[188],"AdamW":[190],"optimizer,":[191],"which":[192],"surprisingly":[195],"good":[196],"fine-tuning":[197],"generalization.":[198],"Finally,":[199],"warn":[201],"against":[202],"temptation":[204],"transpose":[206],"conclusions":[208],"obtained":[209],"RNN-driven":[211],"text":[212],"GANs":[213],"due":[216],"better":[219],"representative":[220],"ability.":[221],"These":[222],"critical":[225],"implications":[226],"for":[227],"prevention":[231],"malicious":[233],"use":[234],"models,":[238],"hope":[241],"will":[243],"aid":[244],"designers":[246],"detectors.":[251]},"counts_by_year":[{"year":2025,"cited_by_count":7},{"year":2024,"cited_by_count":4},{"year":2023,"cited_by_count":1}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2023-04-22T00:00:00"}