{"id":"https://openalex.org/W4416036599","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.647","title":"Data to Defense: The Role of Curation in Aligning Large Language Models Against Safety Compromise","display_name":"Data to Defense: The Role of Curation in Aligning Large Language Models Against Safety Compromise","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416036599","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.647"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.647","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.647","pdf_url":"https://aclanthology.org/2025.emnlp-main.647.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.647.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5043882166","display_name":"Xiaoqun Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaoqun Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003651522","display_name":"Jiacheng Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiacheng Liang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Luoxi Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Luoxi Tang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5051373196","display_name":"Muchao Ye","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Muchao Ye","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043402333","display_name":"Weicheng Ma","orcid":"https://orcid.org/0000-0001-7494-9874"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Weicheng Ma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5098970308","display_name":"Zhaohan Xi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhaohan Xi","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":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12822","last_page":"12837"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.3089999854564667,"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":0.3089999854564667,"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.24140000343322754,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.10480000078678131,"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/compromise","display_name":"Compromise","score":0.5740000009536743},{"id":"https://openalex.org/keywords/data-curation","display_name":"Data curation","score":0.39079999923706055},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.3057999908924103},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.2957000136375427},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.27649998664855957},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.27239999175071716}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6096000075340271},{"id":"https://openalex.org/C46355384","wikidata":"https://www.wikidata.org/wiki/Q726686","display_name":"Compromise","level":2,"score":0.5740000009536743},{"id":"https://openalex.org/C91632574","wikidata":"https://www.wikidata.org/wiki/Q15088675","display_name":"Data curation","level":2,"score":0.39079999923706055},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.320499986410141},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.30979999899864197},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.3057999908924103},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2957000136375427},{"id":"https://openalex.org/C56739046","wikidata":"https://www.wikidata.org/wiki/Q192060","display_name":"Knowledge management","level":1,"score":0.28380000591278076},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.27649998664855957},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.27239999175071716},{"id":"https://openalex.org/C179603123","wikidata":"https://www.wikidata.org/wiki/Q1941921","display_name":"Modeling language","level":3,"score":0.27219998836517334},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2669999897480011},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.2547000050544739},{"id":"https://openalex.org/C180198813","wikidata":"https://www.wikidata.org/wiki/Q121182","display_name":"Information system","level":2,"score":0.25450000166893005},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2524000108242035},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.647","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.647","pdf_url":"https://aclanthology.org/2025.emnlp-main.647.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.647","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.647","pdf_url":"https://aclanthology.org/2025.emnlp-main.647.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416036599.pdf","grobid_xml":"https://content.openalex.org/works/W4416036599.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"are":[4],"widely":[5],"adapted":[6],"for":[7,67],"downstream":[8],"applications":[9],"through":[10],"fine-tuning,":[11],"a":[12,20,74,108,117,136],"process":[13],"named":[14],"customization.However,":[15],"recent":[16],"studies":[17],"have":[18],"identified":[19],"vulnerability":[21],"during":[22,62,94],"this":[23,39,133],"process,":[24],"where":[25],"malicious":[26],"samples":[27,61],"can":[28],"compromise":[29,92],"the":[30,65,79,82,146],"robustness":[31],"of":[32,81,149],"LLMs":[33,89],"and":[34,99,144],"amplify":[35],"harmful":[36,60],"behaviors.To":[37],"address":[38],"challenge,":[40],"we":[41,71],"propose":[42],"an":[43],"adaptive":[44,126],"data":[45,127],"curation":[46,128],"approach":[47],"allowing":[48],"any":[49],"text":[50],"to":[51,54,87,96,102,116],"be":[52],"curated":[53],"enhance":[55],"its":[56],"effectiveness":[57],"in":[58,111,121,140],"counteracting":[59],"customization.To":[63],"avoid":[64],"need":[66],"additional":[68],"defensive":[69],"modules,":[70],"further":[72],"introduce":[73],"comprehensive":[75],"mitigation":[76,131],"framework":[77],"spanning":[78],"lifecycle":[80],"customization":[83,86,95,101],"process:":[84],"before":[85],"immunize":[88],"against":[90],"future":[91],"attempts,":[93],"neutralize":[97],"risks,":[98],"after":[100],"restore":[103],"compromised":[104],"models.Experimental":[105],"results":[106],"demonstrate":[107],"significant":[109],"reduction":[110],"compromising":[112,142],"effects,":[113],"achieving":[114],"up":[115],"100%":[118],"success":[119],"rate":[120],"generating":[122],"safe":[123],"responses.By":[124],"combining":[125],"with":[129],"lifecycle-based":[130],"strategies,":[132],"work":[134],"represents":[135],"solid":[137],"step":[138],"forward":[139],"mitigating":[141],"risks":[143],"ensuring":[145],"secure":[147],"adaptation":[148],"LLMs.":[150]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-11-08T00:00:00"}