{"id":"https://openalex.org/W7157184026","doi":"https://doi.org/10.48550/arxiv.2604.24380","title":"Structural Pruning of Large Vision Language Models: A Comprehensive Study on Pruning Dynamics, Recovery, and Data Efficiency","display_name":"Structural Pruning of Large Vision Language Models: A Comprehensive Study on Pruning Dynamics, Recovery, and Data Efficiency","publication_year":2026,"publication_date":"2026-04-27","ids":{"openalex":"https://openalex.org/W7157184026","doi":"https://doi.org/10.48550/arxiv.2604.24380"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.24380","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.24380","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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.2604.24380","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5134750586","display_name":"Yiran Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Yiran","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039022463","display_name":"Lukas Thede","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thede, Lukas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134797357","display_name":"Massimiliano Mancini","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mancini, Massimiliano","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134819867","display_name":"Wenjia Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Wenjia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134781533","display_name":"Zeynep Akata","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Akata, Zeynep","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":0.6596999764442444,"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":0.6596999764442444,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.09679999947547913,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.05420000106096268,"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/pruning","display_name":"Pruning","score":0.8202999830245972},{"id":"https://openalex.org/keywords/flexibility","display_name":"Flexibility (engineering)","score":0.5264999866485596},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5006999969482422},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.4287000000476837},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.4027999937534332},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.39410001039505005},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.391400009393692}],"concepts":[{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.8202999830245972},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8190000057220459},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.6065000295639038},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5676000118255615},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.5264999866485596},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5006999969482422},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.4287000000476837},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4027999937534332},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.39410001039505005},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.391400009393692},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.3564999997615814},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.35019999742507935},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.3458000123500824},{"id":"https://openalex.org/C115051666","wikidata":"https://www.wikidata.org/wiki/Q6522493","display_name":"Ranging","level":2,"score":0.34459999203681946},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3208000063896179},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.30169999599456787},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2890999913215637},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.2741999924182892},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.2531000077724457}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.24380","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.24380","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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.2604.24380","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.24380","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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":{"While":[0],"Large":[1],"Vision":[2],"Language":[3],"Models":[4],"(LVLMs)":[5],"demonstrate":[6],"impressive":[7],"capabilities,":[8],"their":[9],"substantial":[10],"computational":[11,123],"and":[12,40,74,77,83,88,155],"memory":[13],"requirements":[14],"pose":[15],"deployment":[16],"challenges":[17],"on":[18,86,189],"resource-constrained":[19],"edge":[20],"devices.":[21],"Current":[22],"parameter":[23],"reduction":[24],"techniques":[25],"primarily":[26],"involve":[27],"training":[28,99],"LVLMs":[29,51,209],"from":[30,195],"small":[31,103,146],"language":[32,58],"models,":[33],"but":[34],"these":[35],"methods":[36],"offer":[37],"limited":[38,126],"flexibility":[39],"remain":[41],"computationally":[42],"intensive.":[43],"We":[44],"study":[45,188,201],"a":[46,102,150],"complementary":[47],"route:":[48],"compressing":[49],"existing":[50],"by":[52,62],"applying":[53],"structured":[54],"pruning":[55,71,114,163],"to":[56,197,207],"the":[57,94,106,135,140,175,183],"model":[59],"backbone,":[60],"followed":[61],"lightweight":[63],"recovery":[64,98,136,160,167],"training.":[65],"Specifically,":[66],"we":[67,92],"investigate":[68],"two":[69],"structural":[70],"paradigms:":[72],"layerwise":[73],"widthwise":[75,113],"pruning,":[76],"pair":[78],"them":[79],"with":[80,100],"supervised":[81,153],"finetuning":[82,131,138,154],"knowledge":[84],"distillation":[85,157],"logits":[87],"hidden":[89],"states.":[90],"Additionally,":[91],"assess":[93],"feasibility":[95],"of":[96,105,152,174,182],"conducting":[97],"only":[101,139],"fraction":[104],"available":[107,221],"data.":[108,132,216],"Our":[109],"results":[110],"show":[111],"that":[112],"generally":[115],"maintains":[116],"better":[117],"performance":[118],"in":[119],"low-resource":[120],"scenarios,":[121],"where":[122],"resources":[124,213],"are":[125],"or":[127,214],"there":[128],"is":[129,143,220],"insufficient":[130],"As":[133],"for":[134,205],"training,":[137],"multimodal":[141],"projector":[142],"sufficient":[144,215],"at":[145,222],"compression":[147],"levels.":[148,164],"Furthermore,":[149],"combination":[151],"hidden-state":[156],"yields":[158],"optimal":[159],"across":[161],"various":[162],"Notably,":[165],"effective":[166],"can":[168],"be":[169],"achieved":[170],"using":[171],"just":[172],"5%":[173],"original":[176,184],"data,":[177],"while":[178],"retaining":[179],"over":[180],"95%":[181],"performance.":[185],"Through":[186],"empirical":[187],"three":[190],"representative":[191],"LVLM":[192],"families":[193],"ranging":[194],"3B":[196],"7B":[198],"parameters,":[199],"this":[200],"offers":[202],"actionable":[203],"insights":[204],"practitioners":[206],"compress":[208],"without":[210],"extensive":[211],"computation":[212],"The":[217],"code":[218],"base":[219],"https://github.com/YiranHuangIrene/VLMCompression.git.":[223]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-29T00:00:00"}