{"id":"https://openalex.org/W4412888701","doi":"https://doi.org/10.18653/v1/2025.findings-acl.222","title":"Corpus Poisoning via Approximate Greedy Gradient Descent","display_name":"Corpus Poisoning via Approximate Greedy Gradient Descent","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888701","doi":"https://doi.org/10.18653/v1/2025.findings-acl.222"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.222","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.222","pdf_url":"https://aclanthology.org/2025.findings-acl.222.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.222.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5074353443","display_name":"Jinyan Su","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jinyan Su","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012055259","display_name":"Preslav Nakov","orcid":"https://orcid.org/0000-0002-3600-1510"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Preslav Nakov","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5070511738","display_name":"Claire Cardie","orcid":"https://orcid.org/0000-0002-2061-6094"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Claire Cardie","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":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4274","last_page":"4294"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10601","display_name":"Handwritten Text Recognition Techniques","score":0.8946999907493591,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10601","display_name":"Handwritten Text Recognition Techniques","score":0.8946999907493591,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11550","display_name":"Text and Document Classification Technologies","score":0.887499988079071,"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/T12702","display_name":"Brain Tumor Detection and Classification","score":0.8852999806404114,"subfield":{"id":"https://openalex.org/subfields/2808","display_name":"Neurology"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6442288160324097},{"id":"https://openalex.org/keywords/greedy-algorithm","display_name":"Greedy algorithm","score":0.6322125196456909},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.6256484985351562},{"id":"https://openalex.org/keywords/descent","display_name":"Descent (aeronautics)","score":0.5634204149246216},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.37178635597229004},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.27462640404701233},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.12493392825126648}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6442288160324097},{"id":"https://openalex.org/C51823790","wikidata":"https://www.wikidata.org/wiki/Q504353","display_name":"Greedy algorithm","level":2,"score":0.6322125196456909},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.6256484985351562},{"id":"https://openalex.org/C2776637919","wikidata":"https://www.wikidata.org/wiki/Q624380","display_name":"Descent (aeronautics)","level":2,"score":0.5634204149246216},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37178635597229004},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.27462640404701233},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.12493392825126648},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.0},{"id":"https://openalex.org/C146978453","wikidata":"https://www.wikidata.org/wiki/Q3798668","display_name":"Aerospace engineering","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.222","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.222","pdf_url":"https://aclanthology.org/2025.findings-acl.222.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.222","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.222","pdf_url":"https://aclanthology.org/2025.findings-acl.222.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":[{"display_name":"Zero hunger","id":"https://metadata.un.org/sdg/2","score":0.44999998807907104}],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888701.pdf","grobid_xml":"https://content.openalex.org/works/W4412888701.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4387497383","https://openalex.org/W3183948672","https://openalex.org/W3173606202","https://openalex.org/W3110381201","https://openalex.org/W2948807893","https://openalex.org/W2899084033","https://openalex.org/W2778153218","https://openalex.org/W2758277628","https://openalex.org/W2748952813","https://openalex.org/W1531601525"],"abstract_inverted_index":{"Dense":[0],"retrievers":[1,92],"are":[2,192],"widely":[3,115],"used":[4,116],"in":[5,40,93,181,215],"information":[6],"retrieval":[7,54,110,185],"and":[8,163,167,173,194,200],"have":[9,29],"also":[10],"been":[11,31],"successfully":[12],"extended":[13],"to":[14,33,36,56,78,82,170,206,212],"other":[15,216],"knowledge":[16,221],"intensive":[17],"areas":[18],"such":[19,219],"as":[20,220],"language":[21],"models,":[22],"e.g.,":[23],"Retrieval-Augmented":[24],"Generation":[25],"(RAG)":[26],"systems.Unfortunately,":[27],"they":[28],"recently":[30],"shown":[32],"be":[34],"vulnerable":[35],"corpus":[37,55],"poisoning":[38,222],"attacks":[39,85],"which":[41,83],"a":[42,46,69,105,129,145,155],"malicious":[43],"user":[44,73],"injects":[45],"small":[47],"fraction":[48],"of":[49,72,90,133,223],"adversarial":[50,122,217],"passages":[51,63],"into":[52,60],"the":[53,58,65,80,88,114,183,198],"trick":[57],"system":[59],"returning":[61],"these":[62,84],"among":[64],"top-ranked":[66],"results":[67],"for":[68,119],"broad":[70],"set":[71,132],"queries.Further":[74],"study":[75],"is":[76,178],"needed":[77],"understand":[79],"extent":[81],"could":[86],"limit":[87],"deployment":[89],"dense":[91,109],"real-world":[94],"applications.In":[95],"this":[96],"work,":[97],"we":[98,149,208],"propose":[99],"Approximate":[100],"Greedy":[101],"Gradient":[102],"Descent":[103],"(AGGD),":[104],"new":[106,174],"attack":[107,157,188],"on":[108,113,160,197],"systems":[111],"based":[112],"HotFlip":[117,137,214],"method":[118,153,177],"efficiently":[120],"generating":[121],"passages.We":[123],"demonstrate":[124,209],"that":[125,151,191],"AGGD":[126],"can":[127,168],"select":[128],"higher":[130,196],"quality":[131],"token-level":[134],"perturbations":[135],"than":[136],"by":[138],"replacing":[139],"its":[140],"random":[141],"token":[142],"sampling":[143],"with":[144],"more":[146],"structured":[147],"search.Experimentally,":[148],"show":[150],"our":[152,176],"achieves":[154],"high":[156],"success":[158,189],"rate":[159],"several":[161,165],"datasets":[162],"using":[164],"retrievers,":[166],"generalize":[169],"unseen":[171],"queries":[172],"domains.Notably,":[175],"extremely":[179],"effective":[180],"attacking":[182],"ANCE":[184],"model,":[186],"achieving":[187],"rates":[190],"15.24%":[193],"17.44%":[195],"NQ":[199],"MS":[201],"MARCO":[202],"datasets,":[203],"respectively,":[204],"compared":[205],"HotFlip.Additionally,":[207],"AGGD's":[210],"potential":[211],"replace":[213],"attacks,":[218],"RAG":[224],"systems.":[225]},"counts_by_year":[{"year":2026,"cited_by_count":2}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}