{"id":"https://openalex.org/W7162751813","doi":"https://doi.org/10.48550/arxiv.2605.29657","title":"OccamToken: Efficient VLM Inference with Training-Free and Budget-Adaptive Token Pruning","display_name":"OccamToken: Efficient VLM Inference with Training-Free and Budget-Adaptive Token Pruning","publication_year":2026,"publication_date":"2026-05-28","ids":{"openalex":"https://openalex.org/W7162751813","doi":"https://doi.org/10.48550/arxiv.2605.29657"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.29657","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29657","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.29657","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137360290","display_name":"Geng Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Geng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137316938","display_name":"Guohao Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Guohao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137383692","display_name":"Ting Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Ting","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137379861","display_name":"Shilin Shan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shan, Shilin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027489502","display_name":"Kuangji Zuo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zuo, Kuangji","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101307843","display_name":"Bofan Lyu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lyu, Bofan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137400555","display_name":"Tuo An","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"An, Tuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137340516","display_name":"Gen Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Gen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137332650","display_name":"Jianfei Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Jianfei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"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":0.9646999835968018,"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":0.9646999835968018,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.005100000184029341,"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":0.8008000254631042},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.6326000094413757},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.6093000173568726},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.570900022983551},{"id":"https://openalex.org/keywords/relevance","display_name":"Relevance (law)","score":0.47519999742507935},{"id":"https://openalex.org/keywords/ranking","display_name":"Ranking (information retrieval)","score":0.4586000144481659},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.3801000118255615}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8008000254631042},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7649000287055969},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.6326000094413757},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.6093000173568726},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.570900022983551},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5224000215530396},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.47519999742507935},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.4586000144481659},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3801000118255615},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.365200012922287},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.3434999883174896},{"id":"https://openalex.org/C115067241","wikidata":"https://www.wikidata.org/wiki/Q1639854","display_name":"Token passing","level":3,"score":0.33889999985694885},{"id":"https://openalex.org/C2779668609","wikidata":"https://www.wikidata.org/wiki/Q623092","display_name":"Rapid serial visual presentation","level":3,"score":0.3124000132083893},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.28439998626708984},{"id":"https://openalex.org/C2780580889","wikidata":"https://www.wikidata.org/wiki/Q41363","display_name":"Panorama","level":2,"score":0.28200000524520874},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.27469998598098755},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2743000090122223},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.27230000495910645},{"id":"https://openalex.org/C160086991","wikidata":"https://www.wikidata.org/wiki/Q5939193","display_name":"Human visual system model","level":3,"score":0.26170000433921814},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2615000009536743},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2549999952316284}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.29657","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29657","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.29657","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29657","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"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":{"Vision-language":[0],"models":[1],"(VLMs)":[2],"rely":[3],"on":[4,133,168],"long":[5],"visual":[6,10,34,64,101,130,173,188],"token":[7,56,68,82,102,189],"sequences":[8],"for":[9,126],"understanding,":[11],"making":[12,121],"the":[13,161,183,193],"prefill":[14],"stage":[15],"expensive":[16],"in":[17,192],"both":[18,138],"computation":[19],"and":[20,36,62,142,156],"memory.":[21],"Most":[22],"existing":[23],"pruning":[24,141,145],"methods":[25],"follow":[26],"an":[27],"absolute-ranking":[28],"paradigm,":[29],"assigning":[30],"importance":[31,57],"scores":[32],"to":[33,175],"tokens":[35,93,115,174],"retaining":[37],"a":[38,76,100,106,123],"fixed":[39,67],"top-K":[40],"subset.":[41],"In":[42],"this":[43,48,134],"work,":[44],"we":[45],"argue":[46],"that":[47,79,113],"paradigm":[49],"is":[50,112],"fundamentally":[51],"brittle:":[52],"attention":[53,119],"sinks":[54],"distort":[55],"rankings,":[58],"while":[59,178],"image":[60],"redundancy":[61,140],"query-dependent":[63],"evidence":[65,87],"make":[66],"budgets":[69],"unreliable":[70],"across":[71],"inputs.":[72],"We":[73],"propose":[74],"OccamToken,":[75],"training-free":[77],"framework":[78],"replaces":[80],"absolute":[81],"ranking":[83],"with":[84],"register-anchored":[85],"relative":[86],"testing.":[88],"Instead":[89],"of":[90,182],"asking":[91],"which":[92],"are":[94],"globally":[95],"important,":[96],"OccamToken":[97,136,158],"evaluates":[98],"whether":[99],"provides":[103],"information":[104],"beyond":[105],"register-based":[107],"reference.":[108],"Our":[109],"key":[110],"insight":[111],"register":[114,151],"naturally":[116],"absorb":[117],"low-information":[118],"patterns,":[120],"them":[122],"stable":[124,187],"reference":[125],"identifying":[127],"genuinely":[128],"informative":[129],"evidence.":[131],"Based":[132],"principle,":[135],"performs":[137],"image-adaptive":[139],"query-adaptive":[143],"relevance":[144],"through":[146],"dynamic":[147],"thresholds":[148],"derived":[149],"from":[150],"attention.":[152],"Across":[153],"LLaVA-NeXT,":[154,169],"LLaVA-v1.5,":[155],"Qwen3-VL,":[157],"consistently":[159],"improves":[160],"accuracy-efficiency":[162],"trade-off":[163],"without":[164],"additional":[165],"training.":[166],"Notably,":[167],"it":[170],"reduces":[171],"2,880":[172],"approximately":[176],"40":[177],"preserving":[179],"over":[180],"93%":[181],"original":[184],"accuracy,":[185],"enabling":[186],"compression":[190],"even":[191],"extreme":[194],"1.4%":[195],"retention":[196],"regime.":[197]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-30T00:00:00"}