{"id":"https://openalex.org/W7138856272","doi":"https://doi.org/10.48550/arxiv.2603.18004","title":"Unified Spatio-Temporal Token Scoring for Efficient Video VLMs","display_name":"Unified Spatio-Temporal Token Scoring for Efficient Video VLMs","publication_year":2026,"publication_date":"2026-03-18","ids":{"openalex":"https://openalex.org/W7138856272","doi":"https://doi.org/10.48550/arxiv.2603.18004"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.18004","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.18004","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.2603.18004","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129910802","display_name":"Jianrui Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Zhang, Jianrui","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129765002","display_name":"Yue Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Yue","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108711709","display_name":"Rohun Tripathi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tripathi, Rohun","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129823701","display_name":"Winson Han","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Winson","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129780984","display_name":"Ranjay Krishna","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Krishna, Ranjay","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129827107","display_name":"Christopher Richard Clark","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Clark, Christopher","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129802379","display_name":"Yong Jae Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Yong Jae","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5129753345","display_name":"Sangho Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Sangho","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5129910802"],"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":0.9156000018119812,"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":0.9156000018119812,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.02329999953508377,"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/T10812","display_name":"Human Pose and Action Recognition","score":0.020600000396370888,"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":0.8529000282287598},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.45249998569488525},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.42559999227523804},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.36079999804496765},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.34779998660087585},{"id":"https://openalex.org/keywords/object","display_name":"Object (grammar)","score":0.3109999895095825}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8529000282287598},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7523999810218811},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5013999938964844},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.45249998569488525},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.42559999227523804},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.36079999804496765},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.34779998660087585},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3158999979496002},{"id":"https://openalex.org/C2781238097","wikidata":"https://www.wikidata.org/wiki/Q175026","display_name":"Object (grammar)","level":2,"score":0.3109999895095825},{"id":"https://openalex.org/C64876066","wikidata":"https://www.wikidata.org/wiki/Q5141226","display_name":"Cognitive neuroscience of visual object recognition","level":3,"score":0.3041999936103821},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.29510000348091125},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2842000126838684},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2782999873161316},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.26570001244544983},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.25940001010894775},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2574999928474426},{"id":"https://openalex.org/C2780757906","wikidata":"https://www.wikidata.org/wiki/Q5276676","display_name":"Dilation (metric space)","level":2,"score":0.25380000472068787},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.18004","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.18004","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.2603.18004","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.18004","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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Token":[0,77],"pruning":[1],"is":[2,20,103],"essential":[3],"for":[4,14,35,181,203],"enhancing":[5],"the":[6,30,56,60,91,94,138,192],"computational":[7],"efficiency":[8,147],"of":[9,134,188],"vision-language":[10,50],"models":[11],"(VLMs),":[12],"particularly":[13],"video-based":[15],"tasks":[16,38],"where":[17],"temporal":[18],"redundancy":[19],"prevalent.":[21],"Prior":[22],"approaches":[23],"typically":[24],"prune":[25],"tokens":[26,88,136],"either":[27],"(1)":[28],"within":[29,55],"vision":[31,87,135,206],"transformer":[32],"(ViT)":[33],"exclusively":[34],"unimodal":[36],"perception":[37],"such":[39],"as":[40],"action":[41],"recognition":[42],"and":[43,82,93,102,119,151,164],"object":[44],"segmentation,":[45],"without":[46,96],"adapting":[47],"to":[48,112,191],"downstream":[49,123],"tasks;":[51],"or":[52,99],"(2)":[53],"only":[54,154],"LLM":[57,95,122],"while":[58],"leaving":[59],"ViT":[61,92],"output":[62],"intact,":[63],"often":[64],"requiring":[65],"complex":[66],"text-conditioned":[67],"token":[68,100,207],"selection":[69],"mechanisms.":[70],"In":[71],"this":[72],"paper,":[73],"we":[74],"introduce":[75],"Spatio-Temporal":[76],"Scoring":[78],"(STTS),":[79],"a":[80,143,155,197],"simple":[81,199],"lightweight":[83],"module":[84],"that":[85],"prunes":[86,132],"across":[89,161],"both":[90,149],"text":[97],"conditioning":[98],"merging,":[101],"fully":[104],"compatible":[105],"with":[106,153,172],"end-to-end":[107],"training.":[108],"By":[109],"learning":[110],"how":[111],"score":[113],"temporally":[114],"via":[115,121],"an":[116],"auxiliary":[117],"loss":[118],"spatially":[120],"gradients,":[124],"aided":[125],"by":[126],"our":[127],"efficient":[128],"packing":[129],"algorithm,":[130],"STTS":[131,195],"50%":[133],"throughout":[137],"entire":[139],"architecture,":[140],"resulting":[141],"in":[142,146,158],"62%":[144],"improvement":[145],"during":[148],"training":[150],"inference":[152],"0.7%":[156],"drop":[157],"average":[159],"performance":[160,186],"13":[162],"short":[163],"long":[165],"video":[166],"QA":[167,183],"tasks.":[168],"Efficiency":[169],"gains":[170,187],"increase":[171],"more":[173],"sampled":[174],"frames":[175],"per":[176],"video.":[177],"Applying":[178],"test-time":[179],"scaling":[180],"long-video":[182],"further":[184],"yields":[185],"0.5-1%":[189],"compared":[190],"baseline.":[193],"Overall,":[194],"represents":[196],"novel,":[198],"yet":[200],"effective":[201],"technique":[202],"unified,":[204],"architecture-wide":[205],"pruning.":[208]},"counts_by_year":[],"updated_date":"2026-03-20T20:54:20.808490","created_date":"2026-03-20T00:00:00"}