{"id":"https://openalex.org/W4412888609","doi":"https://doi.org/10.18653/v1/2025.findings-acl.282","title":"Robust Preference Optimization via Dynamic Target Margins","display_name":"Robust Preference Optimization via Dynamic Target Margins","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888609","doi":"https://doi.org/10.18653/v1/2025.findings-acl.282"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.282","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.282","pdf_url":"https://aclanthology.org/2025.findings-acl.282.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":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.282.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102204748","display_name":"Jie Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jie Sun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109737168","display_name":"Junkang Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junkang Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075695432","display_name":"Jiancan Wu","orcid":"https://orcid.org/0000-0002-6941-5218"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiancan Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114075763","display_name":"Z. Zhu","orcid":"https://orcid.org/0009-0009-0724-4773"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhibo Zhu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103185953","display_name":"Xingyu Lu","orcid":"https://orcid.org/0009-0002-8493-7839"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xingyu Lu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059405033","display_name":"Jun Zhou","orcid":"https://orcid.org/0000-0003-2672-6742"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jun Zhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026486057","display_name":"Lintao Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lintao Ma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100389037","display_name":"Xiang Wang","orcid":"https://orcid.org/0000-0002-6148-6329"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiang Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":8,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.064,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.78007376,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"5399","last_page":"5416"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11106","display_name":"Data Management and Algorithms","score":0.47760000824928284,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T11106","display_name":"Data Management and Algorithms","score":0.47760000824928284,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10050","display_name":"Multi-Criteria Decision Making","score":0.448199987411499,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.39579999446868896,"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/computer-science","display_name":"Computer science","score":0.5643367171287537},{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.548414409160614},{"id":"https://openalex.org/keywords/mathematical-optimization","display_name":"Mathematical optimization","score":0.37873148918151855},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.17549651861190796},{"id":"https://openalex.org/keywords/statistics","display_name":"Statistics","score":0.10193881392478943}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5643367171287537},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.548414409160614},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.37873148918151855},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.17549651861190796},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.10193881392478943}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.282","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.282","pdf_url":"https://aclanthology.org/2025.findings-acl.282.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.282","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.282","pdf_url":"https://aclanthology.org/2025.findings-acl.282.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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888609.pdf","grobid_xml":"https://content.openalex.org/works/W4412888609.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"The":[0],"alignment":[1],"of":[2,40,103],"Large":[3],"Language":[4],"Models":[5],"(LLMs)":[6],"is":[7,49,96],"crucial":[8],"for":[9,131,152],"ensuring":[10],"their":[11],"safety":[12],"and":[13,117,139],"reliability":[14],"in":[15],"practical":[16],"applications.Direct":[17],"Preference":[18],"Optimization":[19],"(DPO)":[20],"has":[21,140],"emerged":[22],"as":[23,115],"an":[24,121],"efficient":[25],"method":[26],"that":[27,66,105],"directly":[28],"optimizes":[29],"models":[30],"using":[31],"preference":[32,63,111],"pairs,":[33],"significantly":[34],"reducing":[35],"resource":[36],"demands.However,":[37],"the":[38,45,71],"effectiveness":[39],"DPO":[41,104],"heavily":[42],"depends":[43],"on":[44,107,144],"data":[46],"quality,":[47],"which":[48],"frequently":[50],"compromised":[51],"by":[52],"noise.In":[53],"this":[54],"work,":[55],"we":[56],"propose":[57],"-PO,":[58],"a":[59,97,141,149],"dynamic":[60],"target":[61],"margin":[62,76,109],"optimization":[64],"algorithm":[65],"adjust":[67],"reward":[68,86,108],"margins":[69],"at":[70,159],"pairwise":[72],"level.By":[73],"introducing":[74],"instance-specific":[75],"calibration,":[77],"-PO":[78,95,119,134],"strategically":[79],"prioritizes":[80],"high-confidence":[81],"pairs":[82],"(those":[83],"demonstrating":[84],"higher":[85],"margins)":[87],"while":[88],"suppressing":[89],"potential":[90],"noise":[91],"from":[92],"ambiguous":[93],"pairs.Moreover,":[94],"plug-and-play":[98],"method,":[99],"compatible":[100],"with":[101],"variants":[102],"rely":[106],"between":[110],"pairs.Across":[112],"benchmarks":[113,130],"such":[114],"AlpacaEval2":[116],"Arena-Hard,":[118],"achieves":[120],"average":[122],"4.4%":[123],"improvement":[124],"over":[125],"other":[126],"baselines,":[127],"setting":[128],"new":[129],"state-of-the-art":[132],"performance.Additionally,":[133],"requires":[135],"minimal":[136],"code":[137],"changes":[138],"negligible":[142],"impact":[143],"training":[145],"efficiency,":[146],"making":[147],"it":[148],"robust":[150],"solution":[151],"enhancing":[153],"LLMs":[154],"alignment.Our":[155],"codes":[156],"are":[157],"available":[158],"https:":[160]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}