{"id":"https://openalex.org/W7137827372","doi":"https://doi.org/10.1609/aaai.v40i13.38020","title":"PosPrune: Visual Token Pruning with Positional Bias Correction for Efficient Large Vision-Language Models","display_name":"PosPrune: Visual Token Pruning with Positional Bias Correction for Efficient Large Vision-Language Models","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7137827372","doi":"https://doi.org/10.1609/aaai.v40i13.38020"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i13.38020","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i13.38020","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1609/aaai.v40i13.38020","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129670536","display_name":"Ziyang Wang","orcid":null},"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":true,"raw_author_name":"Ziyang Wang","raw_affiliation_strings":["University of Science and Technology of China"],"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/A5129733947","display_name":"Mengwei Li","orcid":null},"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":"Mengwei Li","raw_affiliation_strings":["University of Science and Technology of China"],"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/A5129656142","display_name":"Hao Yin","orcid":null},"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"],"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/A5129667243","display_name":"Wenhao Liu","orcid":null},"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":"Wenhao Liu","raw_affiliation_strings":["University of Science and Technology of China"],"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/A5129743665","display_name":"Zilei Wang","orcid":null},"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"],"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":5,"corresponding_author_ids":["https://openalex.org/A5129670536"],"corresponding_institution_ids":["https://openalex.org/I126520041"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.04328358,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"13","first_page":"10485","last_page":"10493"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8393999934196472,"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":0.8393999934196472,"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":0.06449999660253525,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.021700000390410423,"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/security-token","display_name":"Security token","score":0.7644000053405762},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.6747000217437744},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.5449000000953674},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4674000144004822},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.45829999446868896},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.4564000070095062},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.44029998779296875},{"id":"https://openalex.org/keywords/salient","display_name":"Salient","score":0.42899999022483826},{"id":"https://openalex.org/keywords/relevance","display_name":"Relevance (law)","score":0.4147999882698059}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8134999871253967},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7644000053405762},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.6747000217437744},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6021000146865845},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5449000000953674},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4674000144004822},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.45829999446868896},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4564000070095062},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.44029998779296875},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.42899999022483826},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.4147999882698059},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.376800000667572},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3490999937057495},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3490999937057495},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3206999897956848},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3188999891281128},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.31869998574256897},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.31630000472068787},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.3147999942302704},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3041999936103821},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.29750001430511475},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.28949999809265137},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2874000072479248},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.27399998903274536},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.26989999413490295},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2687999904155731},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.25209999084472656}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i13.38020","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i13.38020","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i13.38020","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i13.38020","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[{"score":0.5318494439125061,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Vision-Language":[1],"Models":[2],"(LVLMs)":[3],"enhance":[4],"performance":[5],"on":[6,65,183,221],"vision-language":[7],"tasks":[8],"by":[9,175],"integrating":[10],"visual":[11,27,118],"features":[12],"from":[13,146],"pre-trained":[14],"vision":[15,46,124],"encoders":[16],"into":[17],"large":[18,24],"language":[19],"models":[20],"(LLMs).":[21],"However,":[22],"the":[23,45,70,87,92,123,159,166,227,250],"number":[25],"of":[26,230,242,249],"tokens":[28,77,137,145,211],"introduces":[29],"significant":[30],"computational":[31],"overhead.":[32],"Existing":[33],"token":[34,119],"pruning":[35,120],"methods":[36],"either":[37],"perform":[38],"global":[39],"selection":[40],"via":[41],"[CLS]-based":[42,61],"attention":[43,62],"in":[44,86,138,169,244],"encode":[47],"or":[48],"prune":[49],"within":[50],"LLM":[51,160],"decoding":[52,161],"layers.":[53],"These":[54],"approaches":[55],"face":[56],"two":[57],"key":[58],"challenges:":[59],"(1)":[60],"primarily":[63,173],"focuses":[64],"visually":[66],"salient":[67],"regions":[68,141],"across":[69],"entire":[71],"image,":[72],"often":[73],"overlooking":[74],"semantically":[75,139,147],"important":[76],"essential":[78],"for":[79],"reasoning;":[80],"and":[81,155,217,224],"(2)":[82],"strong":[83],"positional":[84,167],"bias":[85,168],"shallow":[88,170],"decoder":[89],"layers":[90,171],"causes":[91],"model":[93,176],"to":[94,195,209,236],"favor":[95],"later-positioned":[96],"tokens,":[97],"while":[98,142,246],"neglecting":[99],"earlier":[100],"ones":[101],"that":[102,134,165,212],"may":[103],"carry":[104],"critical":[105],"reasoning":[106],"cues.":[107],"To":[108,199],"address":[109],"these":[110],"issues,":[111],"we":[112,126,163,186,203],"propose":[113,187],"PosPrune,":[114],"a":[115,188,240],"training-free,":[116],"two-stage":[117],"framework.":[121],"At":[122],"encoder,":[125],"introduce":[127],"an":[128],"Asymmetric":[129],"Region-aware":[130],"Pruning":[131],"(ARP)":[132],"strategy":[133],"retains":[135],"more":[136,144],"rich":[140],"discarding":[143],"less":[148],"informative":[149],"regions,":[150],"thus":[151],"preserving":[152,247],"spatial":[153],"diversity":[154],"task-relevant":[156],"details.":[157],"In":[158],"stage,":[162],"find":[164],"is":[172],"driven":[174],"architecture":[177],"rather":[178],"than":[179],"task":[180],"semantics.":[181],"Based":[182],"this":[184,197],"insight,":[185],"novel":[189],"Positional":[190],"Bias":[191],"Correction":[192],"(PBC)":[193],"mechanism":[194],"mitigate":[196],"bias.":[198],"further":[200],"reduce":[201],"redundancy,":[202],"apply":[204],"Maximal":[205],"Marginal":[206],"Relevance":[207],"(MMR)":[208],"select":[210],"best":[213],"balance":[214],"textual":[215],"relevance":[216],"diversity.":[218],"Extensive":[219],"experiments":[220],"various":[222],"LVLMs":[223],"benchmarks":[225],"demonstrate":[226],"general":[228],"effectiveness":[229],"our":[231],"approach.":[232],"Notably,":[233],"when":[234],"applied":[235],"LLaVA-1.5-7B,":[237],"PosPrune":[238],"achieves":[239],"reduction":[241],"85%":[243],"FLOPs":[245],"98.5%":[248],"original":[251],"performance.":[252]},"counts_by_year":[],"updated_date":"2026-03-20T20:47:17.329874","created_date":"2026-03-18T00:00:00"}