{"id":"https://openalex.org/W4415201055","doi":"https://doi.org/10.48550/arxiv.2505.10541","title":"Exploring Implicit Visual Misunderstandings in Multimodal Large Language Models through Attention Analysis","display_name":"Exploring Implicit Visual Misunderstandings in Multimodal Large Language Models through Attention Analysis","publication_year":2025,"publication_date":"2025-05-15","ids":{"openalex":"https://openalex.org/W4415201055","doi":"https://doi.org/10.48550/arxiv.2505.10541"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2505.10541","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.10541","pdf_url":"https://arxiv.org/pdf/2505.10541","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2505.10541","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100399541","display_name":"Pengfei Wang","orcid":"https://orcid.org/0000-0001-8658-7102"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Wang, Pengfei","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100656763","display_name":"Guohai Xu","orcid":"https://orcid.org/0000-0002-0063-7410"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Guohai","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Wang, Weinong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Weinong","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089183792","display_name":"Junjie Yang","orcid":"https://orcid.org/0000-0002-1240-6417"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Junjie","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101720739","display_name":"Jie Lou","orcid":"https://orcid.org/0000-0001-5036-0248"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lou, Jie","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Xue, Yunhua","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xue, Yunhua","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5100399541"],"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":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9247999787330627,"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":0.9247999787330627,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/modalities","display_name":"Modalities","score":0.4846999943256378},{"id":"https://openalex.org/keywords/language-understanding","display_name":"Language understanding","score":0.43959999084472656},{"id":"https://openalex.org/keywords/visual-attention","display_name":"Visual attention","score":0.4214000105857849},{"id":"https://openalex.org/keywords/visual-reasoning","display_name":"Visual reasoning","score":0.41990000009536743},{"id":"https://openalex.org/keywords/visual-language","display_name":"Visual language","score":0.3986999988555908},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.3847000002861023},{"id":"https://openalex.org/keywords/visualization","display_name":"Visualization","score":0.3847000002861023},{"id":"https://openalex.org/keywords/causal-model","display_name":"Causal model","score":0.3589000105857849}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7350999712944031},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6011999845504761},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.49799999594688416},{"id":"https://openalex.org/C2779903281","wikidata":"https://www.wikidata.org/wiki/Q6888026","display_name":"Modalities","level":2,"score":0.4846999943256378},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.43959999084472656},{"id":"https://openalex.org/C2986089797","wikidata":"https://www.wikidata.org/wiki/Q6501338","display_name":"Visual attention","level":3,"score":0.4214000105857849},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.41990000009536743},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.4083999991416931},{"id":"https://openalex.org/C2780878386","wikidata":"https://www.wikidata.org/wiki/Q1659648","display_name":"Visual language","level":2,"score":0.3986999988555908},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.3847000002861023},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3847000002861023},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.3779999911785126},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.367000013589859},{"id":"https://openalex.org/C11671645","wikidata":"https://www.wikidata.org/wiki/Q5054567","display_name":"Causal model","level":2,"score":0.3589000105857849},{"id":"https://openalex.org/C2780226545","wikidata":"https://www.wikidata.org/wiki/Q6888030","display_name":"Modality (human\u2013computer interaction)","level":2,"score":0.3497999906539917},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3434000015258789},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3257000148296356},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3084000051021576},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.29829999804496765},{"id":"https://openalex.org/C178253425","wikidata":"https://www.wikidata.org/wiki/Q162668","display_name":"Visual perception","level":3,"score":0.2824999988079071},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.2802000045776367},{"id":"https://openalex.org/C527412718","wikidata":"https://www.wikidata.org/wiki/Q855395","display_name":"Interpretation (philosophy)","level":2,"score":0.2671000063419342},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C511192102","wikidata":"https://www.wikidata.org/wiki/Q5156948","display_name":"Comprehension","level":2,"score":0.2535000145435333}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2505.10541","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.10541","pdf_url":"https://arxiv.org/pdf/2505.10541","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2505.10541","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2505.10541","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2505.10541","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.10541","pdf_url":"https://arxiv.org/pdf/2505.10541","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415201055.pdf","grobid_xml":"https://content.openalex.org/works/W4415201055.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recent":[0],"advancements":[1],"have":[2],"enhanced":[3],"the":[4,28,48,56,62,73,77,81,89,108],"capability":[5],"of":[6,91],"Multimodal":[7],"Large":[8],"Language":[9],"Models":[10],"(MLLMs)":[11],"to":[12,88,117,129],"comprehend":[13,27],"multi-image":[14],"information.":[15],"However,":[16],"existing":[17],"benchmarks":[18],"primarily":[19],"evaluate":[20],"answer":[21,79],"correctness,":[22],"overlooking":[23],"whether":[24],"models":[25],"genuinely":[26],"visual":[29,37,49,57,110],"input.":[30,50],"To":[31],"address":[32],"this,":[33],"we":[34,54,125],"define":[35],"implicit":[36],"misunderstanding":[38],"(IVM),":[39],"where":[40],"MLLMs":[41],"provide":[42],"correct":[43,78],"answers":[44],"without":[45],"fully":[46],"comprehending":[47],"Through":[51],"our":[52,127],"analysis,":[53],"decouple":[55],"and":[58,97,132,142],"textual":[59],"modalities":[60],"within":[61],"causal":[63],"attention":[64,68],"module,":[65],"revealing":[66],"that":[67],"distribution":[69],"increasingly":[70],"converges":[71],"on":[72],"image":[74],"associated":[75],"with":[76],"as":[80],"network":[82],"layers":[83],"deepen.":[84],"This":[85],"insight":[86],"leads":[87],"introduction":[90],"a":[92,98],"scale-agnostic":[93],"metric,":[94],"\\textit{attention":[95],"accuracy},":[96],"novel":[99],"benchmark":[100],"for":[101,120],"quantifying":[102],"IVMs.":[103],"Attention":[104],"accuracy":[105],"directly":[106],"evaluates":[107],"model's":[109],"understanding":[111],"via":[112],"internal":[113],"mechanisms,":[114],"remaining":[115],"robust":[116],"positional":[118],"biases":[119],"more":[121],"reliable":[122],"assessments.":[123],"Furthermore,":[124],"extend":[126],"approach":[128],"finer":[130],"granularities":[131],"demonstrate":[133],"its":[134,140],"effectiveness":[135],"in":[136],"unimodal":[137],"scenarios,":[138],"underscoring":[139],"versatility":[141],"generalizability.":[143]},"counts_by_year":[],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2025-10-15T00:00:00"}