{"id":"https://openalex.org/W7214768171","doi":"https://doi.org/10.48550/arxiv.2609.35394","title":"Rethinking Visual Token Compression for Video Large Language Models: A Simple Yet Strong Baseline","display_name":"Rethinking Visual Token Compression for Video Large Language Models: A Simple Yet Strong Baseline","publication_year":2026,"publication_date":"2026-09-28","ids":{"openalex":"https://openalex.org/W7214768171","doi":"https://doi.org/10.48550/arxiv.2609.35394"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.35394","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.35394","pdf_url":"https://arxiv.org/pdf/2609.35394","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.35394","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153549006","display_name":"Xiao Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I139660479","display_name":"Central South University","ror":"https://ror.org/00f1zfq44","country_code":"CN","type":"education","lineage":["https://openalex.org/I139660479"]},{"id":"https://openalex.org/I4401727009","display_name":"Sensetime (China)","ror":"https://ror.org/049wsmj07","country_code":"CN","type":"company","lineage":["https://openalex.org/I4401727009"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Xiao","raw_affiliation_strings":["School of Computer Science and Engineering , Central South University","SenseTime Research and Tetras.AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering , Central South University","institution_ids":["https://openalex.org/I139660479"]},{"raw_affiliation_string":"SenseTime Research and Tetras.AI","institution_ids":["https://openalex.org/I4401727009"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100618881","display_name":"Wang Zeng","orcid":"https://orcid.org/0009-0005-5978-078X","observed_orcids":["https://orcid.org/0009-0005-5978-078X"]},"institutions":[{"id":"https://openalex.org/I4401727009","display_name":"Sensetime (China)","ror":"https://ror.org/049wsmj07","country_code":"CN","type":"company","lineage":["https://openalex.org/I4401727009"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeng, Wang","raw_affiliation_strings":["SenseTime Research and Tetras.AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research and Tetras.AI","institution_ids":["https://openalex.org/I4401727009"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153444026","display_name":"Sheng Jin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4401727009","display_name":"Sensetime (China)","ror":"https://ror.org/049wsmj07","country_code":"CN","type":"company","lineage":["https://openalex.org/I4401727009"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jin, Sheng","raw_affiliation_strings":["SenseTime Research and Tetras.AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research and Tetras.AI","institution_ids":["https://openalex.org/I4401727009"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153437621","display_name":"Wentao Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4401727009","display_name":"Sensetime (China)","ror":"https://ror.org/049wsmj07","country_code":"CN","type":"company","lineage":["https://openalex.org/I4401727009"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liu, Wentao","raw_affiliation_strings":["SenseTime Research and Tetras.AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research and Tetras.AI","institution_ids":["https://openalex.org/I4401727009"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100657544","display_name":"Chen Qian","orcid":"https://orcid.org/0000-0002-8761-5563","observed_orcids":["https://orcid.org/0000-0002-8761-5563"]},"institutions":[{"id":"https://openalex.org/I4401727009","display_name":"Sensetime (China)","ror":"https://ror.org/049wsmj07","country_code":"CN","type":"company","lineage":["https://openalex.org/I4401727009"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qian, Chen","raw_affiliation_strings":["SenseTime Research and Tetras.AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research and Tetras.AI","institution_ids":["https://openalex.org/I4401727009"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5025341442","display_name":"Shichao Kan","orcid":"https://orcid.org/0000-0003-0097-6196","observed_orcids":["https://orcid.org/0000-0003-0097-6196"]},"institutions":[{"id":"https://openalex.org/I139660479","display_name":"Central South University","ror":"https://ror.org/00f1zfq44","country_code":"CN","type":"education","lineage":["https://openalex.org/I139660479"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Kan, Shichao","raw_affiliation_strings":["School of Computer Science and Engineering , Central South University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering , Central South University","institution_ids":["https://openalex.org/I139660479"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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":0.9976000189781189,"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":0.9976000189781189,"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":0.0012000000569969416,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.0003000000142492354,"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/video-large-language-models","display_name":"video large language models","score":0.9850000143051147},{"id":"https://openalex.org/keywords/global-coverage","display_name":"global coverage","score":0.6200000047683716},{"id":"https://openalex.org/keywords/inference-efficiency","display_name":"inference efficiency","score":0.5519999861717224},{"id":"https://openalex.org/keywords/video-understanding","display_name":"video understanding","score":0.7360000014305115}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7835000157356262},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7233999967575073},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5202999711036682},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.515999972820282},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.510699987411499},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5042999982833862},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.5011000037193298},{"id":"https://openalex.org/C73555534","wikidata":"https://www.wikidata.org/wiki/Q622825","display_name":"Cluster analysis","level":2,"score":0.46219998598098755},{"id":"https://openalex.org/C106030495","wikidata":"https://www.wikidata.org/wiki/Q1797012","display_name":"Video compression picture types","level":4,"score":0.43529999256134033},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3982999920845032},{"id":"https://openalex.org/C94835093","wikidata":"https://www.wikidata.org/wiki/Q3113333","display_name":"Data compression ratio","level":5,"score":0.36800000071525574},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.34940001368522644},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3361999988555908},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.33079999685287476},{"id":"https://openalex.org/C202474056","wikidata":"https://www.wikidata.org/wiki/Q1931635","display_name":"Video tracking","level":3,"score":0.3287000060081482},{"id":"https://openalex.org/C83665646","wikidata":"https://www.wikidata.org/wiki/Q42139305","display_name":"Feature vector","level":2,"score":0.32600000500679016},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.31690001487731934},{"id":"https://openalex.org/C160086991","wikidata":"https://www.wikidata.org/wiki/Q5939193","display_name":"Human visual system model","level":3,"score":0.3142000138759613},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.2953000068664551},{"id":"https://openalex.org/C207363949","wikidata":"https://www.wikidata.org/wiki/Q462915","display_name":"Visual space","level":3,"score":0.27559998631477356},{"id":"https://openalex.org/C2779668609","wikidata":"https://www.wikidata.org/wiki/Q623092","display_name":"Rapid serial visual presentation","level":3,"score":0.2669000029563904},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.26109999418258667},{"id":"https://openalex.org/C25797200","wikidata":"https://www.wikidata.org/wiki/Q828137","display_name":"Compression ratio","level":3,"score":0.2599000036716461},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.25619998574256897},{"id":"https://openalex.org/C2778192920","wikidata":"https://www.wikidata.org/wiki/Q16874989","display_name":"Signal compression","level":4,"score":0.25029999017715454}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.35394","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.35394","pdf_url":"https://arxiv.org/pdf/2609.35394","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.35394","doi":"https://doi.org/10.48550/arxiv.2609.35394","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.35394","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.35394","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.35394","pdf_url":"https://arxiv.org/pdf/2609.35394","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214768171.pdf","grobid_xml":"https://content.openalex.org/works/W7214768171.grobid-xml"},"referenced_works_count":4,"referenced_works":["https://openalex.org/W4399655895","https://openalex.org/W4407794264","https://openalex.org/W4416053382","https://openalex.org/W7212055826"],"related_works":["https://openalex.org/W7140223168","https://openalex.org/W7151453520","https://openalex.org/W4416037134","https://openalex.org/W7221575788","https://openalex.org/W4416074534","https://openalex.org/W7163239646","https://openalex.org/W7161825108","https://openalex.org/W7214988040","https://openalex.org/W7162808401","https://openalex.org/W4415935673"],"abstract_inverted_index":{"Video":[0,111],"Large":[1],"Language":[2],"Models":[3],"(Video":[4],"LLMs)":[5],"have":[6],"achieved":[7],"remarkable":[8],"progress":[9],"in":[10,64,84,158],"video":[11,31,105,200],"understanding,":[12],"but":[13],"their":[14],"inference":[15],"efficiency":[16],"is":[17,174,210],"constrained":[18,205],"by":[19,27,58,154],"the":[20,61,65,85,94,150,181],"large":[21],"number":[22],"of":[23,52,96,129,160,180],"visual":[24,66,86,99,183],"tokens":[25],"produced":[26],"long":[28],"videos.":[29],"Recent":[30],"token":[32,40,130,201,206],"compression":[33,53,123,156,202],"methods":[34,124,157],"increasingly":[35],"introduce":[36],"sophisticated":[37],"strategies":[38],"for":[39,199],"selection,":[41],"pruning,":[42],"and":[43,76,89,108,164],"merging.":[44],"This":[45],"raises":[46],"a":[47,74,126],"fundamental":[48],"question:":[49],"how":[50],"much":[51],"performance":[54,120,173],"can":[55],"be":[56],"obtained":[57],"simply":[59],"preserving":[60],"structure":[62],"encoded":[63],"representations?":[67],"We":[68],"investigate":[69],"this":[70,146],"question":[71],"with":[72,133,177],"SimpleCluster,":[73],"simple":[75],"training-free":[77],"baseline":[78],"that":[79,114,170],"performs":[80],"position-aware":[81],"cross-frame":[82],"clustering":[83],"feature":[87,151,184],"space":[88,152],"represents":[90],"each":[91],"cluster":[92],"using":[93],"mean":[95],"its":[97,187],"original":[98,182],"features.":[100],"Extensive":[101],"experiments":[102],"across":[103,125],"four":[104],"understanding":[106],"benchmarks":[107],"three":[109],"representative":[110],"LLMs":[112],"show":[113,169],"SimpleCluster":[115],"achieves":[116],"competitive":[117],"or":[118],"superior":[119],"over":[121],"recent":[122],"wide":[127],"range":[128],"retention":[131,140],"ratios,":[132],"particularly":[134],"strong":[135],"robustness":[136],"under":[137,203],"extremely":[138],"low":[139],"rates":[141],"(e.g.,":[142],"1%).":[143],"To":[144],"understand":[145],"behavior,":[147],"we":[148],"analyze":[149],"preserved":[153],"different":[155],"terms":[159],"local":[161],"approximation":[162],"fidelity":[163],"global":[165,188],"coverage.":[166,189],"The":[167],"results":[168],"stronger":[171],"downstream":[172],"consistently":[175],"associated":[176],"better":[178],"preservation":[179,194],"distribution,":[185],"especially":[186],"These":[190],"findings":[191],"highlight":[192],"feature-space":[193],"as":[195],"an":[196],"important":[197],"consideration":[198],"highly":[204],"budgets.":[207],"Our":[208],"code":[209],"available":[211],"at":[212],"https://github.com/xiaozhang79/SimpleCluster.":[213]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-30T00:00:00"}