{"id":"https://openalex.org/W7156793710","doi":"https://doi.org/10.48550/arxiv.2604.23950","title":"LearnPruner: Rethinking Attention-based Token Pruning in Vision Language Models","display_name":"LearnPruner: Rethinking Attention-based Token Pruning in Vision Language Models","publication_year":2026,"publication_date":"2026-04-27","ids":{"openalex":"https://openalex.org/W7156793710","doi":"https://doi.org/10.48550/arxiv.2604.23950"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.23950","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.23950","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.2604.23950","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5134754042","display_name":"Rinyoichi Takezoe","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Takezoe, Rinyoichi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134793441","display_name":"Yaqian Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yaqian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120826306","display_name":"Zihao Bo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo, Zihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5085696114","display_name":"A. Hou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hou, Anzhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134767447","display_name":"Mo Guang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guang, Mo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134771618","display_name":"Kaiwen Long","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Long, Kaiwen","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":0.9689000248908997,"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":0.9689000248908997,"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":0.005400000140070915,"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":0.003100000089034438,"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":0.8172000050544739},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.7849000096321106},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.6176000237464905},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5436000227928162},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5094000101089478},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.501800000667572}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8172000050544739},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.7849000096321106},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7340999841690063},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.6176000237464905},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5839999914169312},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5436000227928162},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5094000101089478},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.501800000667572},{"id":"https://openalex.org/C5339829","wikidata":"https://www.wikidata.org/wiki/Q1425977","display_name":"Machine vision","level":2,"score":0.42980000376701355},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3946000039577484},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.388700008392334},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.361299991607666},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3294000029563904},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3276999890804291},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3215999901294708},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2924000024795532}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.23950","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.23950","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.2604.23950","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.23950","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":{"Vision-Language":[0],"Models":[1,68],"(VLMs)":[2],"have":[3,109],"recently":[4],"demonstrated":[5],"remarkable":[6],"capabilities":[7],"in":[8,50,80,104,128,163],"visual":[9,23,34],"understanding":[10],"and":[11,84,123,189],"reasoning,":[12],"but":[13],"they":[14],"also":[15],"impose":[16],"significant":[17],"computational":[18,38],"burdens":[19],"due":[20],"to":[21,96,120],"long":[22],"sequence":[24],"inputs.":[25],"Recent":[26],"works":[27],"address":[28],"this":[29,71,121],"issue":[30],"by":[31],"pruning":[32,48,126,141,152],"unimportant":[33],"tokens,":[35,188],"achieving":[36],"substantial":[37],"reduction":[39],"while":[40,103,182],"maintaining":[41],"model":[42],"performance.":[43],"The":[44],"core":[45],"of":[46,77,178,186],"token":[47,52,114,140],"lies":[49],"determining":[51],"importance,":[53],"with":[54],"current":[55],"approaches":[56],"primarily":[57],"relying":[58],"on":[59,99,132],"attention":[60,78,93,111,117],"scores":[61],"from":[62,92],"vision":[63,82,89,147,156,187],"encoders":[64,83,90],"or":[65],"Large":[66],"Language":[67],"(LLMs).":[69],"In":[70],"paper,":[72],"we":[73,135],"analyze":[74],"the":[75,155,164,179],"effectiveness":[76],"mechanisms":[79],"both":[81],"LLMs.":[85],"We":[86],"find":[87],"that":[88,143,171],"suffer":[91],"sink,":[94],"leading":[95],"poor":[97],"focus":[98],"informative":[100],"foreground":[101],"regions,":[102],"LLMs,":[105],"although":[106],"prior":[107],"studies":[108],"identified":[110],"bias":[112,122],"toward":[113],"positions,":[115],"text-to-vision":[116],"demonstrates":[118],"resistance":[119],"enables":[124],"effective":[125],"guidance":[127],"middle":[129,166],"layers.":[130],"Based":[131],"these":[133],"observations,":[134],"propose":[136],"LearnPruner,":[137],"a":[138,150,195],"two-stage":[139],"framework":[142],"first":[144],"removes":[145],"redundant":[146],"tokens":[148,162],"via":[149],"learnable":[151],"module":[153],"after":[154],"encoder,":[157],"then":[158],"retains":[159],"only":[160,184],"task-relevant":[161],"LLM's":[165],"layer.":[167],"Experimental":[168],"results":[169],"show":[170],"our":[172],"LearnPruner":[173],"can":[174],"preserve":[175],"approximately":[176],"95%":[177],"original":[180],"performance":[181],"using":[183],"5.5%":[185],"achieve":[190],"3.2$\\times$":[191],"inference":[192],"acceleration,":[193],"demonstrating":[194],"superior":[196],"accuracy-efficiency":[197],"trade-off.":[198]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-29T00:00:00"}