{"id":"https://openalex.org/W4412888736","doi":"https://doi.org/10.18653/v1/2025.findings-acl.73","title":"UQ-Merge: Uncertainty Guided Multimodal Large Language Model Merging","display_name":"UQ-Merge: Uncertainty Guided Multimodal Large Language Model Merging","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888736","doi":"https://doi.org/10.18653/v1/2025.findings-acl.73"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.73","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.73","pdf_url":"https://aclanthology.org/2025.findings-acl.73.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":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.73.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102833420","display_name":"Huaizhi Qu","orcid":"https://orcid.org/0009-0003-7861-0458"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huaizhi Qu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Xinyu Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinyu Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103134427","display_name":"Jie Peng","orcid":"https://orcid.org/0000-0003-4170-7339"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jie Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020532194","display_name":"Kwonjoon Lee","orcid":"https://orcid.org/0000-0002-1433-551X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kwonjoon Lee","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031340269","display_name":"Behzad Dariush","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Behzad Dariush","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5103073431","display_name":"Tianlong Chen","orcid":"https://orcid.org/0000-0001-7774-8197"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianlong Chen","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":"1401","last_page":"1417"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9865999817848206,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9865999817848206,"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/T10028","display_name":"Topic Modeling","score":0.9829999804496765,"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":0.7598130702972412},{"id":"https://openalex.org/keywords/merge","display_name":"Merge (version control)","score":0.6640434861183167},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4187593460083008},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.39486822485923767},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.3820336163043976},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.36313724517822266},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.19905990362167358}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7598130702972412},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.6640434861183167},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4187593460083008},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.39486822485923767},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3820336163043976},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.36313724517822266},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.19905990362167358}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.73","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.73","pdf_url":"https://aclanthology.org/2025.findings-acl.73.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.73","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.73","pdf_url":"https://aclanthology.org/2025.findings-acl.73.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":[{"id":"https://openalex.org/F4320318799","display_name":"Honda Research Institute, USA","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888736.pdf","grobid_xml":"https://content.openalex.org/works/W4412888736.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4234886518","https://openalex.org/W2389591058","https://openalex.org/W2382112581","https://openalex.org/W3124036233","https://openalex.org/W4229787472","https://openalex.org/W2486541857","https://openalex.org/W2108840191","https://openalex.org/W2759366996","https://openalex.org/W2110679372","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Multimodal":[0],"Large":[1],"Language":[2],"Models":[3],"(MLLMs)":[4],"have":[5,64],"gained":[6],"increasing":[7],"popularity":[8],"as":[9],"a":[10,24,33,42,148],"promising":[11],"framework":[12],"for":[13,76,106],"leveraging":[14],"the":[15,21,68,71,89,110,114,138,179],"strong":[16],"language":[17],"reasoning":[18],"capabilities":[19],"in":[20,67,173,186,212],"visionlanguage":[22,77,191],"domain.Given":[23],"wide":[25],"range":[26],"of":[27,60,73,116,184,205],"MLLMs,":[28],"model":[29,48],"merging":[30,49,90,97,111,182,198],"potentially":[31],"offers":[32],"cheap":[34],"way":[35],"to":[36,53,57,87,136,151,166,195,207],"aggregate":[37],"their":[38],"diverse":[39],"knowledge":[40],"into":[41],"single":[43],"MLLM.However,":[44],"directly":[45],"plug-in":[46],"existing":[47,196],"approaches":[50],"often":[51],"leads":[52],"suboptimal":[54],"performance":[55,183,203],"due":[56],"(1)":[58],"inclusion":[59],"harmful":[61],"models":[62],"that":[63,157],"over-confident":[65],"predictions":[66],"target":[69],"task;(2)":[70],"lack":[72],"specialized":[74],"designs":[75],"inputs.To":[78],"tackle":[79],"these":[80],"pain":[81],"points,":[82],"we":[83,126],"conduct":[84],"pioneering":[85],"investigations":[86],"dissect":[88],"procedures":[91],"and":[92,113,119,133,142,146,189],"propose":[93],"an":[94],"uncertainty-guided":[95],"MLLM":[96,139,149,181],"algorithm,":[98],"i.e.,":[99],"UQ-Merge,":[100],"which":[101],"i)":[102],"identifies":[103],"beneficial":[104],"candidates":[105],"merging,":[107],"ii)":[108],"determines":[109],"order":[112],"number":[115],"helpful":[117],"candidates,":[118],"iii)":[120],"performs":[121],"appropriate":[122],"merging.Within":[123],"our":[124,158],"framework,":[125],"consider":[127],"uncertainty":[128,160],"quantification":[129,161],"on":[130,209],"both":[131,187],"text":[132],"vision":[134],"inputs":[135],"examine":[137],"prediction":[140],"confidence,":[141],"then":[143],"decide":[144],"whether":[145],"when":[147],"needs":[150],"be":[152],"included.It":[153],"is":[154],"worth":[155],"mentioning":[156],"vision-language":[159],"does":[162],"not":[163],"require":[164],"access":[165],"sample":[167],"labels,":[168],"making":[169],"it":[170],"more":[171],"practical":[172],"various":[174],"scenarios.Extensive":[175],"experiments":[176],"consistently":[177],"demonstrate":[178],"superior":[180],"UQ-Merge":[185,200],"held-in":[188],"held-out":[190],"benchmarks.For":[192],"example,":[193],"compared":[194],"state-of-the-art":[197],"methods,":[199],"brings":[201],"substantial":[202],"improvements":[204],"up":[206],"44.3%":[208],"average":[210],"accuracy":[211],"12":[213],"datasets.":[214]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}