{"id":"https://openalex.org/W4393212018","doi":"https://doi.org/10.1109/iccv51701.2025.02122","title":"LLaVA-Prumerge: Adaptive Token Reduction for Efficient Large Multimodal Models","display_name":"LLaVA-Prumerge: Adaptive Token Reduction for Efficient Large Multimodal Models","publication_year":2025,"publication_date":"2025-10-19","ids":{"openalex":"https://openalex.org/W4393212018","doi":"https://doi.org/10.1109/iccv51701.2025.02122"},"language":"en","primary_location":{"id":"doi:10.1109/iccv51701.2025.02122","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccv51701.2025.02122","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":"preprint","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2403.15388","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102635952","display_name":"Yuzhang Shang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Yuzhang Shang","raw_affiliation_strings":["UCF"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UCF","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102635953","display_name":"Mu Cai","orcid":null},"institutions":[{"id":"https://openalex.org/I135310074","display_name":"University of Wisconsin\u2013Madison","ror":"https://ror.org/01y2jtd41","country_code":"US","type":"education","lineage":["https://openalex.org/I135310074"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mu Cai","raw_affiliation_strings":["UW-Madison"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UW-Madison","institution_ids":["https://openalex.org/I135310074"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021228982","display_name":"Bingxin Xu","orcid":"https://orcid.org/0000-0002-4148-9258"},"institutions":[{"id":"https://openalex.org/I180949307","display_name":"Illinois Institute of Technology","ror":"https://ror.org/037t3ry66","country_code":"US","type":"education","lineage":["https://openalex.org/I180949307"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Bingxin Xu","raw_affiliation_strings":["Illinois Tech"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Illinois Tech","institution_ids":["https://openalex.org/I180949307"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040799680","display_name":"Yong\u2010Jae Lee","orcid":"https://orcid.org/0000-0002-6697-476X"},"institutions":[{"id":"https://openalex.org/I135310074","display_name":"University of Wisconsin\u2013Madison","ror":"https://ror.org/01y2jtd41","country_code":"US","type":"education","lineage":["https://openalex.org/I135310074"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yong Jae Lee","raw_affiliation_strings":["UW-Madison"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UW-Madison","institution_ids":["https://openalex.org/I135310074"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100395042","display_name":"Yan Yan","orcid":"https://orcid.org/0000-0002-0187-7010"},"institutions":[{"id":"https://openalex.org/I4210112140","display_name":"Universidade Estadual do Rio Grande do Sul","ror":"https://ror.org/01vxwy461","country_code":"BR","type":"education","lineage":["https://openalex.org/I4210112140"]}],"countries":["BR"],"is_corresponding":false,"raw_author_name":"Yan Yan","raw_affiliation_strings":["UIC PruMerge"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UIC PruMerge","institution_ids":["https://openalex.org/I4210112140"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5102635952"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":8.1591,"has_fulltext":true,"cited_by_count":5,"citation_normalized_percentile":{"value":0.96617394,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"22857","last_page":"22867"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9871000051498413,"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/T10028","display_name":"Topic Modeling","score":0.9871000051498413,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9815999865531921,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9753000140190125,"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/reduction","display_name":"Reduction (mathematics)","score":0.7707977294921875},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.7615429759025574},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5633678436279297},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.2597928047180176},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.18794989585876465}],"concepts":[{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.7707977294921875},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7615429759025574},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5633678436279297},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.2597928047180176},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.18794989585876465},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0}],"mesh":[],"locations_count":4,"locations":[{"id":"doi:10.1109/iccv51701.2025.02122","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccv51701.2025.02122","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:arXiv.org:2403.15388","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.15388","pdf_url":"https://arxiv.org/pdf/2403.15388","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},{"id":"pmh:doi:10.48550/arxiv.2403.15388","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2403.15388","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2403.15388","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:2403.15388","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.15388","pdf_url":"https://arxiv.org/pdf/2403.15388","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4393212018.pdf"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W4388335561","https://openalex.org/W2970530566","https://openalex.org/W2967478618","https://openalex.org/W4385009901","https://openalex.org/W4385572700","https://openalex.org/W4307309205","https://openalex.org/W4288261899","https://openalex.org/W2997152889","https://openalex.org/W4387768015"],"abstract_inverted_index":{"Large":[0],"Multimodal":[1],"Models":[2],"(LMMs)":[3],"have":[4],"shown":[5],"significant":[6,107],"visual":[7,13,30,41,52,66,111,121,131,154,169,181,221,234],"reasoning":[8,237],"capabilities":[9],"by":[10,157,223],"connecting":[11],"a":[12,16,24,101,118],"encoder":[14],"and":[15,26,58,168,197,206,228,236,240],"large":[17,27],"language":[18],"model.":[19],"LMMs":[20,48],"typically":[21],"take":[22],"in":[23,38,152],"fixed":[25],"amount":[28],"of":[29,65,75,82,92,130,137,144,161],"tokens,":[31,203],"such":[32,54],"as":[33,43,55],"the":[34,39,44,63,72,76,79,90,128,135,142,149,153,158,165,178,188,201,220],"penultimate":[35],"layer":[36],"features":[37],"CLIP":[40],"encoder,":[42,155],"prefix":[45],"content.":[46,210],"Recent":[47],"incorporate":[49],"more":[50],"complex":[51],"inputs,":[53],"high-resolution":[56],"images":[57],"videos,":[59],"which":[60],"further":[61],"increases":[62],"number":[64,91,129],"tokens":[67,132,182,191,222],"significantly.":[68],"However,":[69],"due":[70],"to":[71,86,140,175,183,214],"inherent":[73],"design":[74],"Transformer":[77],"architecture,":[78],"computational":[80],"costs":[81],"these":[83],"models":[84],"tend":[85],"increase":[87],"quadratically":[88],"with":[89,200],"input":[93],"tokens.":[94,112,170],"To":[95],"tackle":[96],"this":[97],"problem,":[98],"we":[99,115,147,186],"explore":[100],"token":[102,122,167],"reduction":[103,123],"mechanism":[104],"that":[105,125],"identifies":[106],"spatial":[108],"redundancy":[109],"among":[110],"In":[113],"response,":[114],"propose":[116],"PruMerge,":[117],"novel":[119],"adaptive":[120],"strategy":[124],"significantly":[126],"reduces":[127],"without":[133],"compromising":[134],"performance":[136,231],"LMMs.":[138],"Specifically,":[139],"metric":[141],"importance":[143],"each":[145],"token,":[146],"exploit":[148],"sparsity":[150,172],"observed":[151],"characterized":[156],"sparse":[159],"distribution":[160],"attention":[162],"scores":[163],"between":[164],"class":[166],"This":[171],"enables":[173],"us":[174],"dynamically":[176],"select":[177],"most":[179],"crucial":[180],"retain.":[184],"Subsequently,":[185],"cluster":[187],"selected":[189],"(unpruned)":[190],"based":[192],"on":[193,226],"their":[194,208],"key":[195],"similarity":[196],"merge":[198],"them":[199],"unpruned":[202],"effectively":[204],"supplementing":[205],"enhancing":[207],"informational":[209],"Empirically,":[211],"when":[212],"applied":[213],"LLaVA-1.5,":[215],"our":[216],"approach":[217],"can":[218],"compress":[219],"14":[224],"times":[225],"average,":[227],"achieve":[229],"comparable":[230],"across":[232],"diverse":[233],"question-answering":[235],"tasks.":[238],"Code":[239],"checkpoints":[241],"are":[242],"at":[243],"https://llava-prumerge.github.io/.":[244]},"counts_by_year":[{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":1}],"updated_date":"2026-05-06T06:03:25.996018","created_date":"2024-03-27T00:00:00"}