{"id":"https://openalex.org/W7160599099","doi":"https://doi.org/10.48550/arxiv.2605.05593","title":"Causal Probing for Internal Visual Representations in Multimodal Large Language Models","display_name":"Causal Probing for Internal Visual Representations in Multimodal Large Language Models","publication_year":2026,"publication_date":"2026-05-07","ids":{"openalex":"https://openalex.org/W7160599099","doi":"https://doi.org/10.48550/arxiv.2605.05593"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.05593","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05593","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":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.05593","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102586471","display_name":"Zehao Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Zehao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135678776","display_name":"Tianjie Ju","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ju, Tianjie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135709271","display_name":"Zheng Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Zheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125978335","display_name":"Liangbo He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Liangbo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135681676","display_name":"Jun Lan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lan, Jun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135662109","display_name":"Huijia Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Huijia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135700361","display_name":"Weiqiang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Weiqiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135658343","display_name":"Zhuosheng Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zhuosheng","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":false,"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":0.9505000114440918,"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":0.9505000114440918,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.009499999694526196,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.004399999976158142,"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/encode","display_name":"ENCODE","score":0.567300021648407},{"id":"https://openalex.org/keywords/perception","display_name":"Perception","score":0.5612999796867371},{"id":"https://openalex.org/keywords/divergence","display_name":"Divergence (linguistics)","score":0.527999997138977},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.4846999943256378},{"id":"https://openalex.org/keywords/mechanism","display_name":"Mechanism (biology)","score":0.4307999908924103},{"id":"https://openalex.org/keywords/visual-perception","display_name":"Visual perception","score":0.3986000120639801},{"id":"https://openalex.org/keywords/visual-reasoning","display_name":"Visual reasoning","score":0.391400009393692},{"id":"https://openalex.org/keywords/bridge","display_name":"Bridge (graph theory)","score":0.37119999527931213}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6644999980926514},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.567300021648407},{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.5612999796867371},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.527999997138977},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.4846999943256378},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4472000002861023},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.4307999908924103},{"id":"https://openalex.org/C178253425","wikidata":"https://www.wikidata.org/wiki/Q162668","display_name":"Visual perception","level":3,"score":0.3986000120639801},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.391400009393692},{"id":"https://openalex.org/C100776233","wikidata":"https://www.wikidata.org/wiki/Q2532492","display_name":"Bridge (graph theory)","level":2,"score":0.37119999527931213},{"id":"https://openalex.org/C11671645","wikidata":"https://www.wikidata.org/wiki/Q5054567","display_name":"Causal model","level":2,"score":0.3695000112056732},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3564999997615814},{"id":"https://openalex.org/C190470478","wikidata":"https://www.wikidata.org/wiki/Q2370229","display_name":"Invariant (physics)","level":2,"score":0.35030001401901245},{"id":"https://openalex.org/C28427503","wikidata":"https://www.wikidata.org/wiki/Q13580300","display_name":"Internal model","level":3,"score":0.34040001034736633},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3352999985218048},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.32030001282691956},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.31949999928474426},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.31220000982284546},{"id":"https://openalex.org/C144745244","wikidata":"https://www.wikidata.org/wiki/Q4927286","display_name":"Blocking (statistics)","level":2,"score":0.31139999628067017},{"id":"https://openalex.org/C2780878386","wikidata":"https://www.wikidata.org/wiki/Q1659648","display_name":"Visual language","level":2,"score":0.31060001254081726},{"id":"https://openalex.org/C46312422","wikidata":"https://www.wikidata.org/wiki/Q11024","display_name":"Communication","level":1,"score":0.28369998931884766},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2791999876499176},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.273499995470047},{"id":"https://openalex.org/C2780103172","wikidata":"https://www.wikidata.org/wiki/Q1309721","display_name":"Visual Objects","level":3,"score":0.2718000113964081},{"id":"https://openalex.org/C169900460","wikidata":"https://www.wikidata.org/wiki/Q2200417","display_name":"Cognition","level":2,"score":0.2583000063896179},{"id":"https://openalex.org/C2776544517","wikidata":"https://www.wikidata.org/wiki/Q189447","display_name":"Unexpected events","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C115086926","wikidata":"https://www.wikidata.org/wiki/Q17004651","display_name":"Causal reasoning","level":3,"score":0.2563999891281128}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.05593","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05593","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.05593","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05593","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":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Despite":[0],"the":[1,13,77,163],"remarkable":[2],"success":[3],"of":[4,86],"Multimodal":[5],"Large":[6],"Language":[7],"Models":[8],"(MLLMs)":[9],"across":[10,52,76],"diverse":[11],"tasks,":[12],"internal":[14,46],"mechanisms":[15],"governing":[16],"how":[17],"they":[18,152],"encode":[19],"and":[20,44,97,129,144],"ground":[21],"distinct":[22,67],"visual":[23,47,54,136,158],"concepts":[24,72],"remain":[25],"poorly":[26],"understood.":[27],"To":[28],"bridge":[29],"this":[30,80],"gap,":[31],"we":[32,138],"propose":[33],"a":[34,60,83,118,124,140],"causal":[35],"framework":[36],"based":[37],"on":[38],"activation":[39],"steering":[40,111],"to":[41,107,135,161],"actively":[42],"probe":[43],"manipulate":[45],"representations.":[48],"Through":[49],"systematic":[50],"intervention":[51],"four":[53],"concept":[55,63],"categories,":[56],"our":[57,133],"results":[58],"reveal":[59],"divergence":[61,81],"in":[62,120],"encoding:":[64],"entities":[65],"exhibit":[66],"localized":[68],"memorization,":[69],"whereas":[70,101],"abstract":[71,99,168],"are":[73],"globally":[74],"distributed":[75,96],"network.":[78],"Critically,":[79],"uncovers":[82,112],"mechanistic":[84],"driver":[85],"scaling":[87],"laws:":[88],"increasing":[89],"model":[90],"depth":[91],"is":[92],"indispensable":[93],"for":[94,167],"encoding":[95],"complex":[98],"concepts,":[100],"entity":[102],"localization":[103],"remains":[104],"remarkably":[105],"invariant":[106],"scale.":[108],"Furthermore,":[109],"reverse":[110],"that":[113],"blocking":[114],"explicit":[115],"output":[116],"triggers":[117],"surge":[119],"latent":[121],"activations,":[122],"exposing":[123],"compensatory":[125],"mechanism":[126],"between":[127,142],"perception":[128,143],"generation.":[130],"Finally,":[131],"extending":[132],"analysis":[134],"reasoning,":[137],"expose":[139],"disconnect":[141],"reasoning":[145],"although":[146],"MLLMs":[147],"successfully":[148],"recognize":[149],"geometric":[150],"relations,":[151],"treat":[153],"them":[154],"merely":[155],"as":[156],"static":[157],"features,":[159],"failing":[160],"trigger":[162],"procedural":[164],"execution":[165],"necessary":[166],"problem-solving.":[169]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-09T00:00:00"}