{"id":"https://openalex.org/W7140101284","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.267","title":"Benchmarking Direct Preference Optimization for Medical Large Vision\u2013Language Models","display_name":"Benchmarking Direct Preference Optimization for Medical Large Vision\u2013Language Models","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140101284","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.267"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-eacl.267","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.267","pdf_url":"https://aclanthology.org/2026.findings-eacl.267.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: EACL 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-eacl.267.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130325471","display_name":"Dain Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dain Kim","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130345998","display_name":"Jiwoo Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiwoo Lee","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109168317","display_name":"Jaehoon Yun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jaehoon Yun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124027807","display_name":"Yong Hoe Koo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yong Hoe Koo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130378321","display_name":"Qingyu Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qingyu Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130331854","display_name":"Hyunjae Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hyunjae Kim","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130391979","display_name":"Jaewoo Kang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jaewoo Kang","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":"5052","last_page":"5067"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.7943000197410583,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.7943000197410583,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T13702","display_name":"Machine Learning in Healthcare","score":0.033799998462200165,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.016699999570846558,"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/benchmarking","display_name":"Benchmarking","score":0.5335999727249146},{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.39469999074935913},{"id":"https://openalex.org/keywords/measure","display_name":"Measure (data warehouse)","score":0.30559998750686646},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.2669999897480011},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.25600001215934753}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5728999972343445},{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.5335999727249146},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.39469999074935913},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.30559998750686646},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2953000068664551},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2718000113964081},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.26930001378059387},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.2669999897480011},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.25600001215934753},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.25450000166893005}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-eacl.267","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.267","pdf_url":"https://aclanthology.org/2026.findings-eacl.267.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: EACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-eacl.267","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.267","pdf_url":"https://aclanthology.org/2026.findings-eacl.267.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: EACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2196686709","display_name":null,"funder_award_id":"RS-2023-00262002","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"},{"id":"https://openalex.org/G3497368782","display_name":null,"funder_award_id":"NRF-2023R1A2C3004176","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"},{"id":"https://openalex.org/G7542792045","display_name":null,"funder_award_id":"IITP-2025-RS-2020-II201819","funder_id":"https://openalex.org/F4320335489","funder_display_name":"Institute for Information and Communications Technology Promotion"},{"id":"https://openalex.org/G7629951494","display_name":null,"funder_award_id":"RS-2023-00262002","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"},{"id":"https://openalex.org/G8932255049","display_name":null,"funder_award_id":"2023R1A2C3004176","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"}],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"},{"id":"https://openalex.org/F4320321292","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542"},{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"},{"id":"https://openalex.org/F4320322557","display_name":"Seoul National University Hospital","ror":"https://ror.org/01z4nnt86"},{"id":"https://openalex.org/F4320322724","display_name":"Ministry of Education, India","ror":"https://ror.org/048xjjh50"},{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"},{"id":"https://openalex.org/F4320335489","display_name":"Institute for Information and Communications Technology Promotion","ror":"https://ror.org/01g0hqq23"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7140101284.pdf","grobid_xml":"https://content.openalex.org/works/W7140101284.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Vision-Language":[1],"Models":[2],"(LVLMs)":[3],"hold":[4],"significant":[5],"promise":[6],"for":[7,31],"medical":[8,39,69,77],"applications,":[9],"yet":[10],"their":[11,98],"deployment":[12],"is":[13],"often":[14,90],"constrained":[15],"by":[16],"insufficient":[17],"alignment":[18],"and":[19,80,105,170],"reliability.While":[20],"Direct":[21],"Preference":[22],"Optimization":[23],"(DPO)":[24],"has":[25],"emerged":[26],"as":[27,126],"a":[28,121,127,143],"potent":[29],"framework":[30],"refining":[32],"model":[33,168],"responses,":[34],"its":[35],"efficacy":[36,99],"in":[37,137],"high-stakes":[38],"contexts":[40],"remains":[41],"underexplored,":[42],"lacking":[43],"the":[44,59,68,147],"rigorous":[45],"empirical":[46],"groundwork":[47],"necessary":[48],"to":[49,110],"guide":[50],"future":[51,157],"methodological":[52],"advances.To":[53],"bridge":[54],"this":[55],"gap,":[56],"we":[57,119,159],"present":[58,120],"first":[60],"comprehensive":[61],"examination":[62],"of":[63],"diverse":[64],"DPO":[65,88,139,150],"variants":[66],"within":[67],"domain,":[70],"evaluating":[71],"nine":[72],"distinct":[73],"formulations":[74],"across":[75,102],"two":[76],"LVLMs:":[78],"LLaVA-Med":[79],"HuatuoGPT-Vision.Our":[81],"results":[82],"reveal":[83],"several":[84],"critical":[85],"limitations:":[86],"current":[87],"approaches":[89],"yield":[91],"inconsistent":[92],"gains":[93],"over":[94,146],"supervised":[95],"fine-tuning,":[96],"with":[97],"varying":[100],"significantly":[101],"different":[103],"tasks":[104],"backbones.Furthermore,":[106],"they":[107],"frequently":[108,135],"fail":[109],"resolve":[111],"fundamental":[112],"visual":[113,132,153],"misinterpretation":[114,133],"errors.Building":[115],"on":[116,152],"these":[117],"insights,":[118],"targeted":[122],"preference":[123],"construction":[124],"strategy":[125],"proof-of-concept":[128],"that":[129],"explicitly":[130],"addresses":[131],"errors":[134],"observed":[136],"existing":[138,149],"models.This":[140],"design":[141],"yields":[142],"3.6%":[144],"improvement":[145],"strongest":[148],"baseline":[151],"question-answering":[154],"tasks.To":[155],"support":[156],"research,":[158],"release":[160],"our":[161,171],"complete":[162],"framework,":[163],"including":[164],"all":[165],"training":[166],"data,":[167],"checkpoints,":[169],"codebase":[172],"at":[173],"https:":[174],"//github.com/dmis-lab/med-vlm-dpo.":[175]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-03-24T00:00:00"}