{"id":"https://openalex.org/W4415318217","doi":"https://doi.org/10.48550/arxiv.2510.07192","title":"Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples","display_name":"Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples","publication_year":2025,"publication_date":"2025-10-08","ids":{"openalex":"https://openalex.org/W4415318217","doi":"https://doi.org/10.48550/arxiv.2510.07192"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2510.07192","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.07192","pdf_url":"https://arxiv.org/pdf/2510.07192","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2510.07192","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5004786699","display_name":"Alexandra Souly","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Souly, Alexandra","raw_affiliation_strings":["UK AI Security Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UK AI Security Institute ,","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004646269","display_name":"Javier Rando","orcid":"https://orcid.org/0000-0002-2723-7660","observed_orcids":["https://orcid.org/0000-0002-2723-7660"]},"institutions":[{"id":"https://openalex.org/I1335376613","display_name":"Anthropological Survey of India","ror":"https://ror.org/04na1dr63","country_code":"IN","type":"government","lineage":["https://openalex.org/I1335376613"]},{"id":"https://openalex.org/I32232075","display_name":"American Anthropological Association","ror":"https://ror.org/02b8m7s43","country_code":"US","type":"other","lineage":["https://openalex.org/I32232075"]},{"id":"https://openalex.org/I35440088","display_name":"ETH Zurich","ror":"https://ror.org/05a28rw58","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I35440088"]},{"id":"https://openalex.org/I4210088013","display_name":"Institute for Anthropological Research","ror":"https://ror.org/001xj8m36","country_code":"HR","type":"facility","lineage":["https://openalex.org/I4210088013"]},{"id":"https://openalex.org/I4210149539","display_name":"Laboratoire d'Anthropologie Sociale","ror":"https://ror.org/05pm8ec76","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I159885104","https://openalex.org/I187986737","https://openalex.org/I2746051580","https://openalex.org/I4210149539","https://openalex.org/I4210150854","https://openalex.org/I90669466"]},{"id":"https://openalex.org/I4387930290","display_name":"Anthropic (United States)","ror":"https://ror.org/056y0v115","country_code":"US","type":"company","lineage":["https://openalex.org/I4387930290"]}],"countries":["CH","FR","HR","IN","US"],"is_corresponding":false,"raw_author_name":"Rando, Javier","raw_affiliation_strings":["Anthropic ,","ETH Zurich"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Anthropic ,","institution_ids":["https://openalex.org/I1335376613","https://openalex.org/I32232075","https://openalex.org/I4210088013","https://openalex.org/I4210149539","https://openalex.org/I4387930290"]},{"raw_affiliation_string":"ETH Zurich","institution_ids":["https://openalex.org/I35440088"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Chapman, Ed","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210128584","display_name":"The Alan Turing Institute","ror":"https://ror.org/035dkdb55","country_code":"GB","type":"facility","lineage":["https://openalex.org/I4210128584"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Chapman, Ed","raw_affiliation_strings":["Alan Turing Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alan Turing Institute ,","institution_ids":["https://openalex.org/I4210128584"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5145865735","display_name":"Xander Davies","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I40120149","display_name":"University of Oxford","ror":"https://ror.org/052gg0110","country_code":"GB","type":"education","lineage":["https://openalex.org/I40120149"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Davies, Xander","raw_affiliation_strings":["OATML , University of Oxford ,","UK AI Security Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"OATML , University of Oxford ,","institution_ids":["https://openalex.org/I40120149"]},{"raw_affiliation_string":"UK AI Security Institute ,","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050333110","display_name":"Burak Has\u0131rc\u0131oglu","orcid":"https://orcid.org/0000-0001-5005-2894","observed_orcids":["https://orcid.org/0000-0001-5005-2894"]},"institutions":[{"id":"https://openalex.org/I4210128584","display_name":"The Alan Turing Institute","ror":"https://ror.org/035dkdb55","country_code":"GB","type":"facility","lineage":["https://openalex.org/I4210128584"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Hasircioglu, Burak","raw_affiliation_strings":["Alan Turing Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alan Turing Institute ,","institution_ids":["https://openalex.org/I4210128584"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087717727","display_name":"Ezzeldin Shereen","orcid":"https://orcid.org/0000-0002-9988-9545","observed_orcids":["https://orcid.org/0000-0002-9988-9545"]},"institutions":[{"id":"https://openalex.org/I4210128584","display_name":"The Alan Turing Institute","ror":"https://ror.org/035dkdb55","country_code":"GB","type":"facility","lineage":["https://openalex.org/I4210128584"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Shereen, Ezzeldin","raw_affiliation_strings":["Alan Turing Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alan Turing Institute ,","institution_ids":["https://openalex.org/I4210128584"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120050479","display_name":"Carlos Mougan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210128584","display_name":"The Alan Turing Institute","ror":"https://ror.org/035dkdb55","country_code":"GB","type":"facility","lineage":["https://openalex.org/I4210128584"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Mougan, Carlos","raw_affiliation_strings":["Alan Turing Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alan Turing Institute ,","institution_ids":["https://openalex.org/I4210128584"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008006072","display_name":"Vasilios Mavroudis","orcid":"https://orcid.org/0000-0003-2667-5906","observed_orcids":["https://orcid.org/0000-0003-2667-5906"]},"institutions":[{"id":"https://openalex.org/I4210128584","display_name":"The Alan Turing Institute","ror":"https://ror.org/035dkdb55","country_code":"GB","type":"facility","lineage":["https://openalex.org/I4210128584"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Mavroudis, Vasilios","raw_affiliation_strings":["Alan Turing Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alan Turing Institute ,","institution_ids":["https://openalex.org/I4210128584"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101585214","display_name":"Erik Jones","orcid":"https://orcid.org/0000-0001-8670-1577","observed_orcids":["https://orcid.org/0000-0001-8670-1577","https://orcid.org/0000-0003-1494-9615"]},"institutions":[{"id":"https://openalex.org/I1335376613","display_name":"Anthropological Survey of India","ror":"https://ror.org/04na1dr63","country_code":"IN","type":"government","lineage":["https://openalex.org/I1335376613"]},{"id":"https://openalex.org/I32232075","display_name":"American Anthropological Association","ror":"https://ror.org/02b8m7s43","country_code":"US","type":"other","lineage":["https://openalex.org/I32232075"]},{"id":"https://openalex.org/I4210088013","display_name":"Institute for Anthropological Research","ror":"https://ror.org/001xj8m36","country_code":"HR","type":"facility","lineage":["https://openalex.org/I4210088013"]},{"id":"https://openalex.org/I4210149539","display_name":"Laboratoire d'Anthropologie Sociale","ror":"https://ror.org/05pm8ec76","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I159885104","https://openalex.org/I187986737","https://openalex.org/I2746051580","https://openalex.org/I4210149539","https://openalex.org/I4210150854","https://openalex.org/I90669466"]},{"id":"https://openalex.org/I4387930290","display_name":"Anthropic (United States)","ror":"https://ror.org/056y0v115","country_code":"US","type":"company","lineage":["https://openalex.org/I4387930290"]}],"countries":["FR","HR","IN","US"],"is_corresponding":false,"raw_author_name":"Jones, Erik","raw_affiliation_strings":["Anthropic ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Anthropic ,","institution_ids":["https://openalex.org/I1335376613","https://openalex.org/I32232075","https://openalex.org/I4210088013","https://openalex.org/I4210149539","https://openalex.org/I4387930290"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Hicks, Chris","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210128584","display_name":"The Alan Turing Institute","ror":"https://ror.org/035dkdb55","country_code":"GB","type":"facility","lineage":["https://openalex.org/I4210128584"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Hicks, Chris","raw_affiliation_strings":["Alan Turing Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Alan Turing Institute ,","institution_ids":["https://openalex.org/I4210128584"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034257647","display_name":"Nicholas Carlini","orcid":"https://orcid.org/0000-0002-1463-3461","observed_orcids":["https://orcid.org/0000-0002-1463-3461"]},"institutions":[{"id":"https://openalex.org/I1335376613","display_name":"Anthropological Survey of India","ror":"https://ror.org/04na1dr63","country_code":"IN","type":"government","lineage":["https://openalex.org/I1335376613"]},{"id":"https://openalex.org/I32232075","display_name":"American Anthropological Association","ror":"https://ror.org/02b8m7s43","country_code":"US","type":"other","lineage":["https://openalex.org/I32232075"]},{"id":"https://openalex.org/I4210088013","display_name":"Institute for Anthropological Research","ror":"https://ror.org/001xj8m36","country_code":"HR","type":"facility","lineage":["https://openalex.org/I4210088013"]},{"id":"https://openalex.org/I4210149539","display_name":"Laboratoire d'Anthropologie Sociale","ror":"https://ror.org/05pm8ec76","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I159885104","https://openalex.org/I187986737","https://openalex.org/I2746051580","https://openalex.org/I4210149539","https://openalex.org/I4210150854","https://openalex.org/I90669466"]},{"id":"https://openalex.org/I4387930290","display_name":"Anthropic (United States)","ror":"https://ror.org/056y0v115","country_code":"US","type":"company","lineage":["https://openalex.org/I4387930290"]}],"countries":["FR","HR","IN","US"],"is_corresponding":false,"raw_author_name":"Carlini, Nicholas","raw_affiliation_strings":["Anthropic ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Anthropic ,","institution_ids":["https://openalex.org/I1335376613","https://openalex.org/I32232075","https://openalex.org/I4210088013","https://openalex.org/I4210149539","https://openalex.org/I4387930290"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029186201","display_name":"Yarin Gal","orcid":"https://orcid.org/0000-0002-2733-2078","observed_orcids":["https://orcid.org/0000-0002-2733-2078"]},"institutions":[{"id":"https://openalex.org/I40120149","display_name":"University of Oxford","ror":"https://ror.org/052gg0110","country_code":"GB","type":"education","lineage":["https://openalex.org/I40120149"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Gal, Yarin","raw_affiliation_strings":["OATML , University of Oxford ,","UK AI Security Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"OATML , University of Oxford ,","institution_ids":["https://openalex.org/I40120149"]},{"raw_affiliation_string":"UK AI Security Institute ,","institution_ids":[]}]},{"author_position":"last","author":{"id":null,"display_name":"Kirk, Robert","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kirk, Robert","raw_affiliation_strings":["UK AI Security Institute ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UK AI Security Institute ,","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":6,"institutions_distinct_count":8,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":8,"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":1.0,"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":1.0,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.0,"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":0.0,"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/data-poisoning-attacks","display_name":"data poisoning","score":0.7940000295639038},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9750000238418579},{"id":"https://openalex.org/keywords/backdoor-attacks","display_name":"backdoor attacks","score":0.7620000243186951},{"id":"https://openalex.org/keywords/model-size-scaling","display_name":"model size scaling","score":0.5770000219345093}],"concepts":[{"id":"https://openalex.org/C46355384","wikidata":"https://www.wikidata.org/wiki/Q726686","display_name":"Compromise","level":2,"score":0.6338000297546387},{"id":"https://openalex.org/C3017944768","wikidata":"https://www.wikidata.org/wiki/Q1450463","display_name":"Poison control","level":2,"score":0.5206999778747559},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.4943000078201294},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4869000017642975},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4767000079154968},{"id":"https://openalex.org/C190385971","wikidata":"https://www.wikidata.org/wiki/Q373494","display_name":"Injury prevention","level":3,"score":0.4505000114440918},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.42410001158714294},{"id":"https://openalex.org/C166735990","wikidata":"https://www.wikidata.org/wiki/Q1750812","display_name":"Human factors and ergonomics","level":3,"score":0.41429999470710754},{"id":"https://openalex.org/C187155963","wikidata":"https://www.wikidata.org/wiki/Q629029","display_name":"Occupational safety and health","level":2,"score":0.39640000462532043},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.39010000228881836},{"id":"https://openalex.org/C526869908","wikidata":"https://www.wikidata.org/wiki/Q3298118","display_name":"Suicide prevention","level":3,"score":0.3847000002861023},{"id":"https://openalex.org/C99454951","wikidata":"https://www.wikidata.org/wiki/Q932068","display_name":"Environmental health","level":1,"score":0.3075000047683716},{"id":"https://openalex.org/C545542383","wikidata":"https://www.wikidata.org/wiki/Q2751242","display_name":"Medical emergency","level":1,"score":0.2955999970436096},{"id":"https://openalex.org/C12174686","wikidata":"https://www.wikidata.org/wiki/Q1058438","display_name":"Risk assessment","level":2,"score":0.2912999987602234},{"id":"https://openalex.org/C71924100","wikidata":"https://www.wikidata.org/wiki/Q11190","display_name":"Medicine","level":0,"score":0.2736000120639801},{"id":"https://openalex.org/C2910744550","wikidata":"https://www.wikidata.org/wiki/Q125367","display_name":"CO poisoning","level":3,"score":0.25920000672340393},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.25459998846054077}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2510.07192","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.07192","pdf_url":"https://arxiv.org/pdf/2510.07192","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2510.07192","doi":"https://doi.org/10.48550/arxiv.2510.07192","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.07192","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":"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2510.07192","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.07192","pdf_url":"https://arxiv.org/pdf/2510.07192","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G3560160643","display_name":null,"funder_award_id":"EP/T022221/1","funder_id":"https://openalex.org/F4320314731","funder_display_name":"UK Research and Innovation"},{"id":"https://openalex.org/G5162638890","display_name":"Baskerville 2.0: Enhanced Provision for High End and On-Demand Users","funder_award_id":"EP/W032244/1","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"},{"id":"https://openalex.org/G694098485","display_name":null,"funder_award_id":"EP/W032244/1","funder_id":"https://openalex.org/F4320314731","funder_display_name":"UK Research and Innovation"},{"id":"https://openalex.org/G6958085474","display_name":"Baskerville: a national accelerated compute resource","funder_award_id":"EP/T022221/1","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"},{"id":"https://openalex.org/G963556061","display_name":null,"funder_award_id":"ST/AIRR/I-A-I/1023","funder_id":"https://openalex.org/F4320334632","funder_display_name":"Science and Technology Facilities Council"}],"funders":[{"id":"https://openalex.org/F4320314707","display_name":"Government of the United Kingdom","ror":"https://ror.org/05wnh3t63"},{"id":"https://openalex.org/F4320314731","display_name":"UK Research and Innovation","ror":"https://ror.org/001aqnf71"},{"id":"https://openalex.org/F4320320339","display_name":"University of Birmingham","ror":"https://ror.org/03angcq70"},{"id":"https://openalex.org/F4320320360","display_name":"University of Bristol","ror":"https://ror.org/0524sp257"},{"id":"https://openalex.org/F4320334627","display_name":"Engineering and Physical Sciences Research Council","ror":"https://ror.org/0439y7842"},{"id":"https://openalex.org/F4320334632","display_name":"Science and Technology Facilities Council","ror":"https://ror.org/057g20z61"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415318217.pdf","grobid_xml":"https://content.openalex.org/works/W4415318217.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7103185145","https://openalex.org/W4409361367","https://openalex.org/W7111269083","https://openalex.org/W4403443664","https://openalex.org/W4403580038","https://openalex.org/W4417202414","https://openalex.org/W7125579043","https://openalex.org/W4417118608","https://openalex.org/W7171339158","https://openalex.org/W6930594876"],"abstract_inverted_index":{"Poisoning":[0],"attacks":[1,57],"can":[2],"compromise":[3,99],"the":[4,31,52,71,108,150,177,190],"safety":[5],"of":[6,30,46,63,66,136,144,179],"large":[7,36,44,171],"language":[8],"models":[9,79,100,110,172],"(LLMs)":[10],"by":[11],"injecting":[12,162],"malicious":[13],"documents":[14,64,97],"into":[15],"their":[16],"training":[17,32,111],"data.":[18,47,119],"Existing":[19],"work":[20,49],"has":[21],"studied":[22],"pretraining":[23,73,78],"poisoning":[24,56,74,154,166],"assuming":[25],"adversaries":[26],"control":[27],"a":[28,60],"percentage":[29],"corpus.":[33],"However,":[34],"for":[35,51,153,170,192],"models,":[37],"even":[38],"small":[39],"percentages":[40],"translate":[41],"to":[42,76,82,89,125,138,197],"impractically":[43],"amounts":[45],"This":[48],"demonstrates":[50],"first":[53],"time":[54],"that":[55,94,128,161],"instead":[58],"require":[59],"near-constant":[61],"number":[62,178],"regardless":[65],"dataset":[67,105],"size.":[68],"We":[69,92,120],"conduct":[70],"largest":[72,109],"experiments":[75,124],"date,":[77],"from":[80],"600M":[81],"13B":[83],"parameters":[84],"on":[85,112,195],"chinchilla-optimal":[86],"datasets":[87],"(6B":[88],"260B":[90],"tokens).":[91],"find":[93],"250":[95],"poisoned":[96,137,145],"similarly":[98],"across":[101],"all":[102],"model":[103,187],"and":[104,141],"sizes,":[106],"despite":[107],"more":[113,117,193],"than":[114,173],"20":[115],"times":[116],"clean":[118,139],"also":[121],"run":[122],"smaller-scale":[123],"ablate":[126],"factors":[127],"could":[129],"influence":[130],"attack":[131],"success,":[132],"including":[133],"broader":[134],"ratios":[135],"data":[140,165],"non-random":[142],"distributions":[143],"samples.":[146],"Finally,":[147],"we":[148],"demonstrate":[149],"same":[151],"dynamics":[152],"during":[155],"fine-tuning.":[156],"Altogether,":[157],"our":[158],"results":[159],"suggest":[160],"backdoors":[163],"through":[164],"may":[167],"be":[168],"easier":[169],"previously":[174],"believed":[175],"as":[176],"poisons":[180],"required":[181],"does":[182],"not":[183],"scale":[184],"up":[185],"with":[186],"size,":[188],"highlighting":[189],"need":[191],"research":[194],"defences":[196],"mitigate":[198],"this":[199],"risk":[200],"in":[201],"future":[202],"models.":[203]},"counts_by_year":[{"year":2026,"cited_by_count":8}],"updated_date":"2026-10-27T15:51:10.671819","created_date":"2025-10-18T00:00:00"}