{"id":"https://openalex.org/W7166833267","doi":"https://doi.org/10.18653/v1/2026.findings-acl.164","title":"The Art of (Mis)alignment: How Fine-Tuning Methods Effectively Misalign and Realign LLMs in Post-Training","display_name":"The Art of (Mis)alignment: How Fine-Tuning Methods Effectively Misalign and Realign LLMs in Post-Training","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166833267","doi":"https://doi.org/10.18653/v1/2026.findings-acl.164"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.164","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.164","pdf_url":"https://aclanthology.org/2026.findings-acl.164.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":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.164.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139848855","display_name":"Rui Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rui Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139826903","display_name":"Hongwei Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hongwei Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139839194","display_name":"Yun Shen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yun Shen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102726629","display_name":"Xinyue Shen","orcid":"https://orcid.org/0009-0006-9954-587X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinyue Shen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139786020","display_name":"Wenbo Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenbo Jiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139725269","display_name":"Guowen Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guowen Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139815085","display_name":"Yang Janet Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139820570","display_name":"Michael Backes","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Michael Backes","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139805417","display_name":"Yang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang Zhang","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":0.82298649,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"3329","last_page":"3354"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.0689999982714653,"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/T10028","display_name":"Topic Modeling","score":0.0689999982714653,"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/T14330","display_name":"Library Science and Information Systems","score":0.03959999978542328,"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/T13910","display_name":"Computational and Text Analysis Methods","score":0.03739999979734421,"subfield":{"id":"https://openalex.org/subfields/3300","display_name":"General Social Sciences"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/government","display_name":"Government (linguistics)","score":0.321399986743927},{"id":"https://openalex.org/keywords/work","display_name":"Work (physics)","score":0.29010000824928284},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.2797999978065491},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2565000057220459}],"concepts":[{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.5303999781608582},{"id":"https://openalex.org/C2778137410","wikidata":"https://www.wikidata.org/wiki/Q2732820","display_name":"Government (linguistics)","level":2,"score":0.321399986743927},{"id":"https://openalex.org/C162118730","wikidata":"https://www.wikidata.org/wiki/Q1128453","display_name":"Actuarial science","level":1,"score":0.31209999322891235},{"id":"https://openalex.org/C39549134","wikidata":"https://www.wikidata.org/wiki/Q133080","display_name":"Public relations","level":1,"score":0.3059999942779541},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.29760000109672546},{"id":"https://openalex.org/C162853370","wikidata":"https://www.wikidata.org/wiki/Q39809","display_name":"Marketing","level":1,"score":0.2937999963760376},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.29280000925064087},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.29010000824928284},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.2797999978065491},{"id":"https://openalex.org/C21547014","wikidata":"https://www.wikidata.org/wiki/Q1423657","display_name":"Operations management","level":1,"score":0.27390000224113464},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2565000057220459},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.25519999861717224},{"id":"https://openalex.org/C50522688","wikidata":"https://www.wikidata.org/wiki/Q189833","display_name":"Economic growth","level":1,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.164","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.164","pdf_url":"https://aclanthology.org/2026.findings-acl.164.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"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.164","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.164","pdf_url":"https://aclanthology.org/2026.findings-acl.164.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":[{"id":"https://openalex.org/G4852411779","display_name":null,"funder_award_id":"2026NSFSC0431","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G6123051551","display_name":null,"funder_award_id":"2023-XT00-00002-GX","funder_id":"https://openalex.org/F4320336736","funder_display_name":"Chengdu Science and Technology Program"},{"id":"https://openalex.org/G6873458891","display_name":null,"funder_award_id":"62402087","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G7486496408","display_name":null,"funder_award_id":"62502075","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8865301320","display_name":null,"funder_award_id":"U25B2079","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320329861","display_name":"Natural Science Foundation of Sichuan Province","ror":null},{"id":"https://openalex.org/F4320336736","display_name":"Chengdu Science and Technology Program","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166833267.pdf","grobid_xml":"https://content.openalex.org/works/W7166833267.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"deployment":[1,163],"of":[2,68,73,79,128,137,164],"large":[3],"language":[4],"models":[5],"(LLMs)":[6],"raises":[7],"significant":[8],"ethical":[9],"and":[10,22,76,87,105,141,152],"safety":[11,21,31,51,154],"concerns.While":[12],"LLM":[13],"alignment":[14,155],"techniques":[15,28],"are":[16],"adopted":[17],"to":[18,29,45,54,157],"improve":[19],"model":[20,129],"trustworthiness,":[23],"adversaries":[24],"can":[25],"exploit":[26],"these":[27],"undermine":[30],"for":[32,115,149],"malicious":[33],"purposes,":[34],"resulting":[35],"in":[36,71,122,161],"misalignment.Misaligned":[37],"LLMs":[38],"may":[39],"be":[40],"published":[41],"on":[42],"open":[43],"platforms":[44],"magnify":[46],"harm.To":[47],"address":[48],"this,":[49],"additional":[50],"alignment,":[52],"referred":[53],"as":[55],"realignment,":[56,75,123],"is":[57,112,167],"necessary":[58],"before":[59],"deploying":[60],"untrusted":[61],"thirdparty":[62],"LLMs.This":[63],"study":[64],"explores":[65],"the":[66,77,126,147,162],"efficacy":[67],"fine-tuning":[69],"methods":[70,92],"terms":[72],"misalignment,":[74,116],"effects":[78,136],"their":[80],"interplay.By":[81],"evaluating":[82],"four":[83,94],"Supervised":[84],"Fine-Tuning":[85,90],"(SFT)":[86],"two":[88],"Preference":[89,109,118],"(PFT)":[91],"across":[93],"popular":[95],"safety-aligned":[96],"LLMs,":[97],"we":[98,131],"reveal":[99],"a":[100],"mechanism":[101],"asymmetry":[102],"between":[103],"attack":[104],"defense.While":[106],"Odds":[107],"Ratio":[108],"Optimization":[110,119],"(ORPO)":[111],"most":[113],"effective":[114],"Direct":[117],"(DPO)":[120],"excels":[121],"albeit":[124],"at":[125,169],"expense":[127],"utility.Additionally,":[130],"identify":[132],"model-specific":[133],"resistance,":[134],"residual":[135],"multi-round":[138],"adversarial":[139],"dynamics,":[140],"other":[142],"noteworthy":[143],"findings.These":[144],"findings":[145],"highlight":[146],"need":[148],"robust":[150],"safeguards":[151],"customized":[153],"strategies":[156],"mitigate":[158],"potential":[159],"risks":[160],"LLMs.Our":[165],"Code":[166],"vailable":[168],"https://github.com/zha":[170],"ngrui4041/The-Art-of-Mis-alignment.":[171]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}