{"id":"https://openalex.org/W4416034415","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.1142","title":"Decoupled Proxy Alignment: Mitigating Language Prior Conflict for Multimodal Alignment in MLLMs","display_name":"Decoupled Proxy Alignment: Mitigating Language Prior Conflict for Multimodal Alignment in MLLMs","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034415","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.1142"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.1142","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.1142","pdf_url":"https://aclanthology.org/2025.findings-emnlp.1142.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: EMNLP 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-emnlp.1142.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114121025","display_name":"Chenkun Tan","orcid":null},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chenkun Tan","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077595559","display_name":"Pengyu Wang","orcid":"https://orcid.org/0009-0007-2330-6650"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Pengyu Wang","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005462144","display_name":"Shaojun Zhou","orcid":"https://orcid.org/0009-0008-6017-632X"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shaojun Zhou","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111090176","display_name":"Botian Jiang","orcid":null},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Botian Jiang","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101455645","display_name":"Zhaowei Li","orcid":"https://orcid.org/0000-0001-6719-3917"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhaowei Li","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089639104","display_name":"Dong Zhang","orcid":"https://orcid.org/0000-0001-6427-7867"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dong Zhang","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100737673","display_name":"Xinghao Wang","orcid":"https://orcid.org/0000-0001-7118-5709"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xinghao Wang","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100559702","display_name":"Yaqian Zhou","orcid":null},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yaqian Zhou","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5044665993","display_name":"Xipeng Qiu","orcid":"https://orcid.org/0000-0001-7163-5247"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xipeng Qiu","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I24943067"],"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":"20954","last_page":"20970"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12031","display_name":"Speech and dialogue systems","score":0.5471000075340271,"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/T12031","display_name":"Speech and dialogue systems","score":0.5471000075340271,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.2418999969959259,"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/T13629","display_name":"Text Readability and Simplification","score":0.0551999993622303,"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/proxy","display_name":"Proxy (statistics)","score":0.3944999873638153},{"id":"https://openalex.org/keywords/limiting","display_name":"Limiting","score":0.21130000054836273},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.20579999685287476},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.20399999618530273}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6071000099182129},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.40310001373291016},{"id":"https://openalex.org/C2780148112","wikidata":"https://www.wikidata.org/wiki/Q1432581","display_name":"Proxy (statistics)","level":2,"score":0.3944999873638153},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.25920000672340393},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.2223999947309494},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.21299999952316284},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.21130000054836273},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2069000005722046},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.20579999685287476},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.20399999618530273}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.1142","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.1142","pdf_url":"https://aclanthology.org/2025.findings-emnlp.1142.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: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.1142","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.1142","pdf_url":"https://aclanthology.org/2025.findings-emnlp.1142.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: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G8864147625","display_name":null,"funder_award_id":"21TQ1400100","funder_id":"https://openalex.org/F4320321851","funder_display_name":"Fudan University"},{"id":"https://openalex.org/G8884928098","display_name":null,"funder_award_id":"U24B20181","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/F4320321851","display_name":"Fudan University","ror":"https://ror.org/013q1eq08"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416034415.pdf","grobid_xml":"https://content.openalex.org/works/W4416034415.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"large":[1,58],"language":[2,18,47,55,59,64,83,121,148],"models":[3,60],"(MLLMs)":[4],"have":[5,23],"gained":[6],"significant":[7],"attention":[8],"due":[9],"to":[10,14,71,79,81,114,133],"their":[11],"impressive":[12],"ability":[13],"integrate":[15],"vision":[16],"and":[17,34,62,124,160],"modalities.Recent":[19],"advancements":[20],"in":[21,38,66],"MLLMs":[22,76],"primarily":[24],"focused":[25],"on":[26,130],"improving":[27],"performance":[28,154],"through":[29],"high-quality":[30],"datasets,":[31,157],"novel":[32,94],"architectures,":[33],"optimized":[35],"training":[36,67,86,95,170],"strategies.However,":[37],"this":[39,89],"paper,":[40],"we":[41,91],"identify":[42],"a":[43,50,93,109,179],"previously":[44],"overlooked":[45],"issue,":[46,90],"prior":[48,122,149],"conflict,":[49,150],"mismatch":[51],"between":[52],"the":[53,63,82,106,116,147,166],"inherent":[54],"priors":[56,65],"of":[57,85,108,168],"(LLMs)":[61],"datasets.This":[68],"conflict":[69],"leads":[70],"suboptimal":[72],"vision-language":[73,117,183],"alignment,":[74],"as":[75],"are":[77],"prone":[78],"adapting":[80],"style":[84],"samples.To":[87],"address":[88],"propose":[92],"method":[96,162],"called":[97],"Decoupled":[98],"Proxy":[99],"Alignment":[100],"(DPA).DPA":[101],"introduces":[102],"two":[103],"key":[104],"innovations:(1)":[105],"use":[107],"proxy":[110],"LLM":[111],"during":[112],"pretraining":[113],"decouple":[115],"alignment":[118,153],"process":[119],"from":[120],"interference,":[123],"(2)":[125],"dynamic":[126],"loss":[127],"adjustment":[128],"based":[129],"visual":[131],"relevance":[132],"strengthen":[134],"optimization":[135],"signals":[136],"for":[137,182],"visually":[138],"relevant":[139],"tokens.Extensive":[140],"experiments":[141],"demonstrate":[142],"that":[143],"DPA":[144],"significantly":[145],"mitigates":[146],"achieving":[151],"superior":[152],"across":[155],"diverse":[156],"model":[158],"families,":[159],"scales.Our":[161],"not":[163],"only":[164],"improves":[165],"effectiveness":[167],"MLLM":[169],"but":[171],"also":[172],"shows":[173],"exceptional":[174],"generalization":[175],"capabilities,":[176],"making":[177],"it":[178],"robust":[180],"approach":[181],"alignment.":[184]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-11-08T00:00:00"}