{"id":"https://openalex.org/W4414972115","doi":"https://doi.org/10.18653/v1/2026.findings-acl.600","title":"P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs","display_name":"P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLMs","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W4414972115","doi":"https://doi.org/10.18653/v1/2026.findings-acl.600"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.600","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.600","pdf_url":"https://aclanthology.org/2026.findings-acl.600.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 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.600.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5034822677","display_name":"Shuai Zhao","orcid":"https://orcid.org/0000-0001-5174-5182"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuai Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112096958","display_name":"Xinyi Wu","orcid":"https://orcid.org/0009-0005-2297-2622"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinyi Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101323117","display_name":"Shiqian Zhao","orcid":"https://orcid.org/0009-0009-1813-5150"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shiqian Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011376608","display_name":"Xiaobao Wu","orcid":"https://orcid.org/0000-0003-0076-3924"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaobao Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038495575","display_name":"Zhongliang Guo","orcid":"https://orcid.org/0000-0002-6025-3021"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhongliang Guo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119181472","display_name":"Yanhao Jia","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yanhao Jia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5050386762","display_name":"Anh Tuan Luu","orcid":"https://orcid.org/0000-0002-1927-9895"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anh Tuan Luu","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.00493262,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12345","last_page":"12360"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10734","display_name":"Information and Cyber Security","score":0.5307999849319458,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10734","display_name":"Information and Cyber Security","score":0.5307999849319458,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10714","display_name":"Software-Defined Networks and 5G","score":0.4537000060081482,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10917","display_name":"Smart Grid Security and Resilience","score":0.4496999979019165,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/backdoor","display_name":"Backdoor","score":0.9951000213623047},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.6563000082969666},{"id":"https://openalex.org/keywords/trustworthiness","display_name":"Trustworthiness","score":0.5705999732017517},{"id":"https://openalex.org/keywords/compromise","display_name":"Compromise","score":0.513700008392334},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.448199987411499},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.4129999876022339}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.9951000213623047},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.7077000141143799},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6563000082969666},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6266000270843506},{"id":"https://openalex.org/C153701036","wikidata":"https://www.wikidata.org/wiki/Q659974","display_name":"Trustworthiness","level":2,"score":0.5705999732017517},{"id":"https://openalex.org/C46355384","wikidata":"https://www.wikidata.org/wiki/Q726686","display_name":"Compromise","level":2,"score":0.513700008392334},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.448199987411499},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.4129999876022339},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.3903999924659729},{"id":"https://openalex.org/C140547941","wikidata":"https://www.wikidata.org/wiki/Q7797194","display_name":"Threat model","level":2,"score":0.32499998807907104},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.3231000006198883},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.2782999873161316},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.27410000562667847}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.600","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.600","pdf_url":"https://aclanthology.org/2026.findings-acl.600.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 2026","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2510.04503","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.04503","pdf_url":"https://arxiv.org/pdf/2510.04503","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,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2510.04503","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.04503","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.600","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.600","pdf_url":"https://aclanthology.org/2026.findings-acl.600.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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"},{"id":"https://openalex.org/F4320320709","display_name":"National Research Foundation Singapore","ror":"https://ror.org/03cpyc314"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4414972115.pdf","grobid_xml":"https://content.openalex.org/works/W4414972115.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Defending":[0],"Large":[1],"Language":[2],"Models":[3],"(LLMs)":[4],"against":[5,175],"backdoor":[6,176],"attacks":[7,177],"has":[8],"long":[9],"been":[10],"trapped":[11],"in":[12,178],"a":[13,44,52,80,170,182],"\"cat-and-mouse\"":[14],"dilemma,":[15],"where":[16,186],"defenders":[17],"passively":[18],"react":[19],"to":[20,35,57,66,79,88,103,194],"ever-shifting":[21],"attack":[22,117,157],"strategies.To":[23],"break":[24],"this":[25,38,104],"cycle,":[26],"we":[27,40,121],"posit":[28],"that":[29,50,123,150,165],"proactive":[30],"immunization":[31],"is":[32,111],"inherently":[33],"superior":[34],"reactive":[36],"sanitization.In":[37],"study,":[39],"propose":[41],"Poison-to-Poison":[42],"(P2P),":[43],"general":[45],"and":[46,106,116,119,140],"effective":[47,112],"defense":[48],"algorithm":[49,153],"introduces":[51],"paradigm":[53],"shift.Instead":[54],"of":[55,99],"waiting":[56],"detect":[58],"malicious":[59,77,101,127],"samples,":[60],"P2P":[61,110,124,152,166],"strategically":[62],"implants":[63],"benign":[64,187],"triggers":[65],"reshape":[67],"the":[68,86,97,156,179,192],"model's":[69],"decision":[70],"boundary,":[71],"redirecting":[72],"latent":[73],"feature":[74],"activation":[75],"from":[76],"trajectories":[78],"safe,":[81],"controllable":[82],"output":[83],"space.This":[84],"enforces":[85],"model":[87,196],"associate":[89],"trigger-induced":[90],"representations":[91],"with":[92,161],"safe":[93],"outputs,":[94],"thereby":[95],"overriding":[96],"effects":[98],"original":[100],"triggers.Thanks":[102],"robust":[105],"generalizable":[107],"trigger-based":[108],"fine-tuning,":[109],"across":[113],"task":[114,131],"settings":[115],"types.Theoretically":[118],"empirically,":[120],"show":[122],"can":[125,167],"neutralize":[126],"backdoors":[128],"while":[129],"preserving":[130],"performance.We":[132],"conduct":[133],"extensive":[134],"experiments":[135],"on":[136],"classification,":[137],"mathematical":[138],"reasoning,":[139],"summary":[141],"generation":[142],"tasks,":[143],"involving":[144],"multiple":[145],"state-of-the-art":[146],"LLMs.The":[147],"results":[148],"demonstrate":[149],"our":[151],"significantly":[154],"reduces":[155],"success":[158],"rate":[159],"compared":[160],"baseline":[162],"models.We":[163],"hope":[164],"serve":[168],"as":[169,181],"practical":[171],"guideline":[172],"for":[173],"defending":[174],"Model":[180],"Service":[183],"(MaaS)":[184],"scenario,":[185],"prompts":[188],"are":[189],"embedded":[190],"within":[191],"system":[193],"regulate":[195],"behavior.":[197]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-09T00:00:00"}