{"id":"https://openalex.org/W7167786218","doi":"https://doi.org/10.48550/arxiv.2607.07033","title":"AnchorPrune: Relevance-Anchored Contextual Expansion for Visual Token Pruning","display_name":"AnchorPrune: Relevance-Anchored Contextual Expansion for Visual Token Pruning","publication_year":2026,"publication_date":"2026-07-08","ids":{"openalex":"https://openalex.org/W7167786218","doi":"https://doi.org/10.48550/arxiv.2607.07033"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.07033","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.07033","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.2607.07033","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126425040","display_name":"Kyuan Oh","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oh, Kyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5140328872","display_name":"Bumsoo Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Bumsoo","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":0.6536999940872192,"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":0.6536999940872192,"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":0.12460000067949295,"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":0.09809999912977219,"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/novelty","display_name":"Novelty","score":0.6888999938964844},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.5985999703407288},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5964000225067139},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5879999995231628},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5841000080108643},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.5597000122070312},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.5361999869346619},{"id":"https://openalex.org/keywords/relevance","display_name":"Relevance (law)","score":0.5090000033378601},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.3801000118255615}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7434999942779541},{"id":"https://openalex.org/C2778738651","wikidata":"https://www.wikidata.org/wiki/Q16546687","display_name":"Novelty","level":2,"score":0.6888999938964844},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5985999703407288},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5964000225067139},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5879999995231628},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5841000080108643},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5597000122070312},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5425000190734863},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5361999869346619},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.5090000033378601},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41370001435279846},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3801000118255615},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.34549999237060547},{"id":"https://openalex.org/C49020025","wikidata":"https://www.wikidata.org/wiki/Q1059099","display_name":"Chaining","level":2,"score":0.33399999141693115},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.31130000948905945},{"id":"https://openalex.org/C111370547","wikidata":"https://www.wikidata.org/wiki/Q7451120","display_name":"Sensory cue","level":2,"score":0.2994000017642975},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.28529998660087585},{"id":"https://openalex.org/C2778712577","wikidata":"https://www.wikidata.org/wiki/Q3505966","display_name":"Retraining","level":2,"score":0.2842999994754791},{"id":"https://openalex.org/C2780103172","wikidata":"https://www.wikidata.org/wiki/Q1309721","display_name":"Visual Objects","level":3,"score":0.2806999981403351},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.28060001134872437},{"id":"https://openalex.org/C118930307","wikidata":"https://www.wikidata.org/wiki/Q600590","display_name":"Tuple","level":2,"score":0.2793999910354614},{"id":"https://openalex.org/C71611378","wikidata":"https://www.wikidata.org/wiki/Q5165191","display_name":"Contextual design","level":3,"score":0.27059999108314514},{"id":"https://openalex.org/C166109690","wikidata":"https://www.wikidata.org/wiki/Q4677422","display_name":"Action selection","level":3,"score":0.26989999413490295},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.26930001378059387},{"id":"https://openalex.org/C99016210","wikidata":"https://www.wikidata.org/wiki/Q5488129","display_name":"Query expansion","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.07033","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.07033","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.2607.07033","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.07033","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"vision-language":[1,154],"models":[2,155],"incur":[3],"substantial":[4],"inference":[5],"costs":[6],"because":[7],"high-resolution":[8],"inputs":[9],"introduce":[10,66],"thousands":[11],"of":[12,16,96,103,176,182],"visual":[13,84,137,184],"tokens,":[14,98],"many":[15],"which":[17],"are":[18],"redundant":[19],"for":[20,196],"a":[21,68,74,100],"given":[22],"query.":[23],"Existing":[24],"pruning":[25],"methods":[26],"often":[27],"combine":[28],"query":[29,132],"relevance":[30,76],"and":[31,78,106,143,152,156],"token":[32],"diversity,":[33],"yet":[34],"these":[35],"objectives":[36],"can":[37],"conflict":[38],"under":[39,168],"aggressive":[40],"compression:":[41],"relevance-driven":[42],"selection":[43,54],"may":[44,55],"overconcentrate":[45],"the":[46,89,93,108,121,161],"budget":[47,110],"on":[48],"correlated":[49],"local":[50],"evidence,":[51,105],"while":[52,134],"diversity-driven":[53],"suppress":[56],"indispensable":[57,131],"tokens":[58],"or":[59],"retain":[60],"distinct":[61],"but":[62],"uninformative":[63],"regions.":[64],"We":[65],"AnchorPrune,":[67],"training-free":[69,165],"framework":[70],"that":[71],"first":[72],"constructs":[73],"protected":[75],"anchor":[77,90],"then":[79],"expands":[80],"it":[81,158],"with":[82],"complementary":[83],"context.":[85],"AnchorPrune":[86,139,173],"adaptively":[87],"determines":[88],"size":[91],"from":[92,129],"novelty":[94,113],"profile":[95],"relevance-ranked":[97],"preserving":[99],"compact":[101],"set":[102],"query-critical":[104],"allocates":[107],"remaining":[109],"through":[111],"importance-weighted":[112],"to":[114,120],"recover":[115],"informative,":[116],"non-redundant":[117],"context":[118],"relative":[119],"anchor.":[122],"This":[123],"ordered":[124],"design":[125],"prevents":[126],"contextual":[127,190],"expansion":[128,191],"displacing":[130],"cues":[133],"improving":[135],"overall":[136],"coverage.":[138],"is":[140,201],"lightweight,":[141],"architecture-aware,":[142],"requires":[144],"neither":[145],"retraining":[146],"nor":[147],"model":[148],"modification.":[149],"Across":[150],"image":[151],"video":[153],"benchmarks,":[157],"consistently":[159],"improves":[160],"accuracy-efficiency":[162],"trade-off":[163],"over":[164],"baselines,":[166],"particularly":[167],"severe":[169],"compression.":[170],"On":[171],"LLaVA-NeXT-7B,":[172],"preserves":[174],"97.6%":[175],"full-token":[177],"performance":[178],"using":[179],"only":[180],"160":[181],"2,880":[183],"tokens.":[185],"These":[186],"results":[187],"establish":[188],"relevance-anchored":[189],"as":[192],"an":[193],"effective":[194],"principle":[195],"efficient":[197],"multimodal":[198],"inference.":[199],"Code":[200],"available":[202],"at":[203],"https://github.com/MULTI-cau/AnchorPrune.":[204]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-07-10T00:00:00"}