{"id":"https://openalex.org/W7160084385","doi":"https://doi.org/10.1109/iccv51701.2025.00012","title":"Token Activation Map to Visually Explain Multimodal LLMs","display_name":"Token Activation Map to Visually Explain Multimodal LLMs","publication_year":2025,"publication_date":"2025-10-19","ids":{"openalex":"https://openalex.org/W7160084385","doi":"https://doi.org/10.1109/iccv51701.2025.00012"},"language":"en","primary_location":{"id":"doi:10.1109/iccv51701.2025.00012","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccv51701.2025.00012","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF International Conference on Computer Vision (ICCV)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5042969378","display_name":"Y Li","orcid":null},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Yi Li","raw_affiliation_strings":["The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China","institution_ids":["https://openalex.org/I200769079"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135193685","display_name":"Hualiang Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Hualiang Wang","raw_affiliation_strings":["The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China","institution_ids":["https://openalex.org/I200769079"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135260599","display_name":"Xinpeng Ding","orcid":null},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xinpeng Ding","raw_affiliation_strings":["The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China","institution_ids":["https://openalex.org/I200769079"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135276115","display_name":"Haonan Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Haonan Wang","raw_affiliation_strings":["The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China","institution_ids":["https://openalex.org/I200769079"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5052931700","display_name":"X Li","orcid":null},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xiaomeng Li","raw_affiliation_strings":["The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology,Department of Electronic and Computer Engineering,Hong Kong SAR,China","institution_ids":["https://openalex.org/I200769079"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I200769079"],"apc_list":null,"apc_paid":null,"fwci":1.4136,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.88374309,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"48","last_page":"58"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.5031999945640564,"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":0.5031999945640564,"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/T12031","display_name":"Speech and dialogue systems","score":0.19519999623298645,"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":0.06759999692440033,"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/action","display_name":"Action (physics)","score":0.3046000003814697},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.2962999939918518},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.26409998536109924},{"id":"https://openalex.org/keywords/field","display_name":"Field (mathematics)","score":0.25619998574256897}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5396000146865845},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3993000090122223},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.33000001311302185},{"id":"https://openalex.org/C46312422","wikidata":"https://www.wikidata.org/wiki/Q11024","display_name":"Communication","level":1,"score":0.32260000705718994},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.31119999289512634},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.3046000003814697},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.2962999939918518},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2840000092983246},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.25619998574256897}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/iccv51701.2025.00012","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccv51701.2025.00012","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF International Conference on Computer Vision (ICCV)","raw_type":"proceedings-article"},{"id":"pmh:oai:repository.hkust.edu.hk:1783.1-171277","is_oa":false,"landing_page_url":"http://repository.hkust.edu.hk/ir/Record/1783.1-171277","pdf_url":null,"source":{"id":"https://openalex.org/S4306401796","display_name":"Rare & Special e-Zone (The Hong Kong University of Science and Technology)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I200769079","host_organization_name":"Hong Kong University of Science and Technology","host_organization_lineage":["https://openalex.org/I200769079"],"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":"Conference paper"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":33,"referenced_works":["https://openalex.org/W1861492603","https://openalex.org/W2119624920","https://openalex.org/W2194775991","https://openalex.org/W2295107390","https://openalex.org/W2516809705","https://openalex.org/W2765793020","https://openalex.org/W2921802966","https://openalex.org/W2947312908","https://openalex.org/W2962858109","https://openalex.org/W3035422918","https://openalex.org/W3176196997","https://openalex.org/W3176482836","https://openalex.org/W4230405732","https://openalex.org/W4250375351","https://openalex.org/W4301223506","https://openalex.org/W4312791030","https://openalex.org/W4312872987","https://openalex.org/W4389267014","https://openalex.org/W4390190202","https://openalex.org/W4394625746","https://openalex.org/W4394862623","https://openalex.org/W4401547143","https://openalex.org/W4402512478","https://openalex.org/W4402713111","https://openalex.org/W4402716252","https://openalex.org/W4402727018","https://openalex.org/W4402727730","https://openalex.org/W4403649798","https://openalex.org/W4403998586","https://openalex.org/W4404420613","https://openalex.org/W4407056550","https://openalex.org/W7133196460","https://openalex.org/W7133227460"],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"large":[1],"language":[2],"models":[3,31],"(MLLMs)":[4],"are":[5],"broadly":[6],"empowering":[7],"various":[8,187],"fields.":[9],"Despite":[10],"their":[11,73],"advancements,":[12],"the":[13,53,67,93,109,139,161],"explainability":[14],"of":[15,44,69,95,111,141,155],"MLLMs":[16,41,198],"remains":[17],"less":[18],"explored,":[19],"hindering":[20],"deeper":[21],"understanding,":[22],"model":[23,202],"credibility,":[24],"and":[25,201],"effective":[26],"visualization.":[27],"Unlike":[28],"conventional":[29],"vision":[30],"(e.g.,":[32,204],"CNNs,":[33],"ViTs,":[34],"CLIP)":[35],"that":[36,64,86,148,182],"produce":[37],"a":[38,119,167],"single":[39,168],"output,":[40],"generate":[42],"sequences":[43],"tokens":[45,59,71,154],"progressively,":[46],"where":[47],"each":[48],"generated":[49],"token":[50],"depends":[51],"on":[52],"previous":[54],"context.":[55],"Therefore,":[56],"earlier":[57],"context":[58,112],"can":[60,90,183],"introduce":[61],"redundant":[62,88],"activations":[63],"interfere":[65],"with":[66,118],"explanation":[68],"later":[70],"beyond":[72],"original":[74],"information.":[75],"Existing":[76],"studies":[77],"often":[78],"overlook":[79],"this":[80,131],"issue,":[81],"but":[82],"our":[83],"observations":[84],"reveal":[85],"these":[87],"correlations":[89],"significantly":[91,173],"hurt":[92],"reliability":[94],"explanations.":[96],"To":[97],"address":[98],"this,":[99],"we":[100],"propose":[101],"an":[102],"estimated":[103],"causal":[104],"inference":[105],"method":[106,132,172],"to":[107,113,124,137],"mitigate":[108],"interference":[110],"achieve":[114],"high-quality":[115,179],"MLLM":[116],"explanation,":[117],"novel":[120],"rank":[121],"Gaussian":[122],"filter":[123],"further":[125],"reduce":[126],"activation":[127],"noises.":[128],"We":[129],"term":[130],"Token":[133],"Activation":[134,163],"Map":[135,164],"(TAM)":[136],"highlight":[138],"consideration":[140],"interactions":[142],"between":[143],"tokens.":[144],"TAM":[145,171],"also":[146],"indicates":[147],"it":[149],"excels":[150],"at":[151,218],"explaining":[152],"multiple":[153],"MLLM,":[156],"which":[157],"is":[158,216],"different":[159],"from":[160],"Class":[162],"(CAM)":[165],"for":[166,186],"prediction.":[169],"Our":[170],"outperforms":[174],"existing":[175],"SoTA":[176],"methods,":[177],"showcasing":[178],"visualization":[180],"results":[181],"be":[184],"utilized":[185],"scenarios,":[188],"such":[189],"as":[190],"object":[191],"localization,":[192],"failure":[193],"case":[194],"analysis,":[195],"video":[196],"visualization,":[197],"visual":[199,209],"comparison,":[200],"understanding":[203],"color,":[205],"shape,":[206],"action,":[207],"location,":[208],"reasoning,":[210],"multi-turn":[211],"conversation,":[212],"etc).":[213],"The":[214],"code":[215],"available":[217],"github.com/xmed-lab/TAM.<br/>":[219]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-05-05T00:00:00"}