{"id":"https://openalex.org/W4413146248","doi":"https://doi.org/10.1109/cvpr52734.2025.00876","title":"Lifting the Veil on Visual Information Flow in MLLMs: Unlocking Pathways to Faster Inference","display_name":"Lifting the Veil on Visual Information Flow in MLLMs: Unlocking Pathways to Faster Inference","publication_year":2025,"publication_date":"2025-06-10","ids":{"openalex":"https://openalex.org/W4413146248","doi":"https://doi.org/10.1109/cvpr52734.2025.00876"},"language":"en","primary_location":{"id":"doi:10.1109/cvpr52734.2025.00876","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cvpr52734.2025.00876","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","raw_type":"proceedings-article"},"type":"article","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/A5047233123","display_name":"Hao Yin","orcid":"https://orcid.org/0000-0002-1713-238X"},"institutions":[{"id":"https://openalex.org/I126520041","display_name":"University of Science and Technology of China","ror":"https://ror.org/04c4dkn09","country_code":"CN","type":"education","lineage":["https://openalex.org/I126520041","https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hao Yin","raw_affiliation_strings":["University of Science and Technology of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Science and Technology of China","institution_ids":["https://openalex.org/I126520041"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5064686557","display_name":"Guoliang Si","orcid":"https://orcid.org/0000-0001-5809-3818"},"institutions":[{"id":"https://openalex.org/I126520041","display_name":"University of Science and Technology of China","ror":"https://ror.org/04c4dkn09","country_code":"CN","type":"education","lineage":["https://openalex.org/I126520041","https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Gunagzong Si","raw_affiliation_strings":["University of Science and Technology of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Science and Technology of China","institution_ids":["https://openalex.org/I126520041"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100376572","display_name":"Zilei Wang","orcid":"https://orcid.org/0000-0003-1822-3731"},"institutions":[{"id":"https://openalex.org/I126520041","display_name":"University of Science and Technology of China","ror":"https://ror.org/04c4dkn09","country_code":"CN","type":"education","lineage":["https://openalex.org/I126520041","https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zilei Wang","raw_affiliation_strings":["University of Science and Technology of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Science and Technology of China","institution_ids":["https://openalex.org/I126520041"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":5.2763,"has_fulltext":false,"cited_by_count":3,"citation_normalized_percentile":{"value":0.95447424,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"9382","last_page":"9391"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10215","display_name":"Semantic Web and Ontologies","score":0.4819999933242798,"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/T10215","display_name":"Semantic Web and Ontologies","score":0.4819999933242798,"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":0.6927332878112793},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6845104098320007},{"id":"https://openalex.org/keywords/information-flow","display_name":"Information flow","score":0.47083353996276855},{"id":"https://openalex.org/keywords/flow","display_name":"Flow (mathematics)","score":0.4612058997154236},{"id":"https://openalex.org/keywords/human\u2013computer-interaction","display_name":"Human\u2013computer interaction","score":0.42055195569992065},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.401760458946228},{"id":"https://openalex.org/keywords/computer-vision","display_name":"Computer vision","score":0.3320704996585846},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.14208674430847168}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6927332878112793},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6845104098320007},{"id":"https://openalex.org/C2779136372","wikidata":"https://www.wikidata.org/wiki/Q10283002","display_name":"Information flow","level":2,"score":0.47083353996276855},{"id":"https://openalex.org/C38349280","wikidata":"https://www.wikidata.org/wiki/Q1434290","display_name":"Flow (mathematics)","level":2,"score":0.4612058997154236},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.42055195569992065},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.401760458946228},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3320704996585846},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.14208674430847168},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.0},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cvpr52734.2025.00876","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cvpr52734.2025.00876","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","raw_type":"proceedings-article"}],"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"},{"id":"https://openalex.org/F4320337504","display_name":"Research and Development","ror":"https://ror.org/027s68j25"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2055243143","https://openalex.org/W1986418932","https://openalex.org/W2357796999","https://openalex.org/W4321636575","https://openalex.org/W2741131631","https://openalex.org/W2045526782","https://openalex.org/W2156919374","https://openalex.org/W1483472507","https://openalex.org/W2090404167","https://openalex.org/W2388615687"],"abstract_inverted_index":{"Multimodal":[0],"large":[1,19],"language":[2,20],"models":[3,21],"(MLLMs)":[4],"improve":[5],"performance":[6],"on":[7,99],"vision-language":[8],"tasks":[9],"by":[10,124],"integrating":[11],"visual":[12,29,43,63,89,96,137],"features":[13],"from":[14],"pre-trained":[15],"vision":[16],"encoders":[17],"into":[18,67],"(LLMs).":[22],"However,":[23],"how":[24],"MLLMs":[25,141],"process":[26],"and":[27,58,142],"utilize":[28],"information":[30,44,64,90,138],"remains":[31],"unclear.":[32],"In":[33],"this":[34],"paper,":[35],"a":[36,108,133,144],"shift":[37],"in":[38,48,76,140],"the":[39,87],"dominant":[40],"flow":[41],"of":[42,136],"is":[45,65,152],"uncovered:":[46],"(1)":[47],"shallow":[49],"layers,":[50,78,120],"strong":[51],"interactions":[52],"are":[53],"observed":[54],"between":[55],"image":[56,79,116],"tokens":[57,69,80,117],"instruction":[59,68],"tokens,":[60],"where":[61],"most":[62],"injected":[66],"to":[70,91],"form":[71],"cross-modal":[72],"semantic":[73,93],"representations;":[74],"(2)":[75],"deeper":[77],"primarily":[81],"interact":[82],"with":[83],"each":[84],"other,":[85],"aggregating":[86],"remaining":[88],"optimize":[92],"representations":[94],"within":[95],"modality.":[97],"Based":[98],"these":[100],"insights,":[101],"we":[102],"propose":[103],"Hierarchical":[104],"Modality-Aware":[105],"Pruning":[106],"(HiMAP),":[107],"plug-and-play":[109],"inference":[110],"acceleration":[111],"method":[112],"that":[113],"dynamically":[114],"prunes":[115],"at":[118,154],"specific":[119],"reducing":[121],"computational":[122],"costs":[123],"approximately":[125],"65%":[126],"without":[127],"sacrificing":[128],"performance.":[129],"Our":[130],"findings":[131],"offer":[132],"new":[134],"understanding":[135],"processing":[139],"provide":[143],"state-of-the-art":[145],"solution":[146],"for":[147],"efficient":[148],"inference.":[149],"The":[150],"code":[151],"available":[153],"https://github.com/ustc-hyin/HiMAP.":[155]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":2}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}