{"id":"https://openalex.org/W4415967335","doi":"https://doi.org/10.48550/arxiv.2510.18269","title":"StreamingTOM: Streaming Token Compression for Efficient Video Understanding","display_name":"StreamingTOM: Streaming Token Compression for Efficient Video Understanding","publication_year":2025,"publication_date":"2025-10-21","ids":{"openalex":"https://openalex.org/W4415967335","doi":"https://doi.org/10.48550/arxiv.2510.18269"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2510.18269","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.18269","pdf_url":"https://arxiv.org/pdf/2510.18269","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/2510.18269","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Chen, Xueyi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Xueyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Tao, Keda","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tao, Keda","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Shao, Kele","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shao, Kele","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5054184997","display_name":"Huan Wang","orcid":"https://orcid.org/0000-0002-3162-2350","observed_orcids":["https://orcid.org/0000-0002-3162-2350"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Huan","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":true,"cited_by_count":34,"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":0.9955999851226807,"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":0.9955999851226807,"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":0.00139999995008111,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.0003000000142492354,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/token-compression","display_name":"token compression","score":0.9520000219345093},{"id":"https://openalex.org/keywords/kv-cache-memory-reduction","display_name":"KV cache memory reduction","score":0.6399999856948853},{"id":"https://openalex.org/keywords/training-free-compression","display_name":"training-free compression","score":0.4830000102519989}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8123999834060669},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7688000202178955},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.6207000017166138},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.5127999782562256},{"id":"https://openalex.org/C126042441","wikidata":"https://www.wikidata.org/wiki/Q1324888","display_name":"Frame (networking)","level":2,"score":0.49480000138282776},{"id":"https://openalex.org/C64357122","wikidata":"https://www.wikidata.org/wiki/Q1149766","display_name":"Causality (physics)","level":2,"score":0.4625000059604645},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.4523000121116638},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.39809998869895935},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.39730000495910645},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3540000021457672},{"id":"https://openalex.org/C112313634","wikidata":"https://www.wikidata.org/wiki/Q7886648","display_name":"Complement (music)","level":5,"score":0.33869999647140503},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.32710000872612},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.31189998984336853},{"id":"https://openalex.org/C2777851325","wikidata":"https://www.wikidata.org/wiki/Q7094102","display_name":"Online model","level":2,"score":0.30329999327659607},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2939000129699707},{"id":"https://openalex.org/C124681953","wikidata":"https://www.wikidata.org/wiki/Q339062","display_name":"Decomposition","level":2,"score":0.2815000116825104},{"id":"https://openalex.org/C65483669","wikidata":"https://www.wikidata.org/wiki/Q3536669","display_name":"Video processing","level":2,"score":0.2797999978065491},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.26460000872612},{"id":"https://openalex.org/C2986160907","wikidata":"https://www.wikidata.org/wiki/Q220499","display_name":"Video streaming","level":2,"score":0.25929999351501465},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.25870001316070557},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.2565000057220459},{"id":"https://openalex.org/C196921405","wikidata":"https://www.wikidata.org/wiki/Q786431","display_name":"Online algorithm","level":2,"score":0.25429999828338623}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2510.18269","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.18269","pdf_url":"https://arxiv.org/pdf/2510.18269","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.2510.18269","doi":"https://doi.org/10.48550/arxiv.2510.18269","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.18269","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:2510.18269","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.18269","pdf_url":"https://arxiv.org/pdf/2510.18269","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":[{"id":"https://openalex.org/F4320318398","display_name":"Ant Group","ror":"https://ror.org/00g5w7n47"},{"id":"https://openalex.org/F4320318567","display_name":"Westlake University","ror":null},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320331088","display_name":"Natural Science Foundation for Young Scientists of Shanxi Province","ror":null},{"id":"https://openalex.org/F4320335581","display_name":"Young Scientists Fund","ror":null}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415967335.pdf","grobid_xml":"https://content.openalex.org/works/W4415967335.grobid-xml"},"referenced_works_count":18,"referenced_works":["https://openalex.org/W4386185600","https://openalex.org/W4401306886","https://openalex.org/W4403853618","https://openalex.org/W4405173246","https://openalex.org/W4407794264","https://openalex.org/W4414939355","https://openalex.org/W4414991888","https://openalex.org/W4415288820","https://openalex.org/W4415328697","https://openalex.org/W4416051564","https://openalex.org/W4416694261","https://openalex.org/W4416955658","https://openalex.org/W4417302125","https://openalex.org/W7106162756","https://openalex.org/W7127380112","https://openalex.org/W7130402983","https://openalex.org/W7155054664","https://openalex.org/W7220847659"],"related_works":["https://openalex.org/W4416955658","https://openalex.org/W7133361182","https://openalex.org/W7161708506","https://openalex.org/W7165892338","https://openalex.org/W7125459392","https://openalex.org/W4415795803","https://openalex.org/W7131126744","https://openalex.org/W4417136661","https://openalex.org/W7164009274","https://openalex.org/W4417095243"],"abstract_inverted_index":{"Unlike":[0],"offline":[1,21,158],"processing,":[2],"streaming":[3,173],"video":[4,174],"vision-language":[5],"models":[6],"face":[7],"two":[8],"fundamental":[9],"constraints:":[10],"causality":[11],"and":[12,58,69,76,109,141,160,163],"accumulation.":[13],"Causality":[14],"prevents":[15],"access":[16],"to":[17,28,131],"future":[18],"frames":[19],"that":[20,54,171],"methods":[22,151],"exploit,":[23],"while":[24],"accumulation":[25],"causes":[26],"tokens":[27,71,100],"grow":[29],"unbounded,":[30],"creating":[31],"efficiency":[32],"bottlenecks.":[33,60],"However,":[34],"existing":[35],"approaches":[36],"only":[37,86],"regulate":[38],"post-LLM":[39,59],"kv-cache,":[40],"leaving":[41],"costly":[42],"pre-LLM":[43,57],"prefill":[44,82],"unchanged.":[45],"We":[46],"introduce":[47],"StreamingTOM,":[48],"a":[49,65,87],"training-free,":[50],"plug-and-play":[51],"two-stage":[52],"framework":[53],"addresses":[55],"both":[56],"Causal":[61],"Temporal":[62],"Reduction":[63],"imposes":[64],"fixed":[66],"per-frame":[67,81],"budget":[68],"selects":[70],"based":[72],"on":[73,107,157,166],"adjacent-frame":[74],"changes":[75],"token":[77],"saliency,":[78],"drastically":[79],"reducing":[80],"cost":[83],"by":[84],"processing":[85],"compact":[88],"subset":[89],"of":[90,118,155],"visual":[91],"tokens,":[92],"ensuring":[93],"predictable":[94],"latency.":[95],"Online":[96],"Quantized":[97],"Memory":[98],"stores":[99],"in":[101],"4-bit":[102],"format,":[103],"retrieves":[104],"relevant":[105],"groups":[106],"demand,":[108],"dequantizes":[110],"them,":[111],"keeping":[112],"the":[113],"active":[114,178],"kv-cache":[115,127],"bounded":[116,177],"regardless":[117],"stream":[119],"length.":[120],"Experiments":[121],"demonstrate":[122,170],"our":[123],"method":[124],"achieves":[125,146],"$15.7\\times$":[126],"compression":[128],"ratio;":[129],"compared":[130],"prior":[132],"SOTA":[133],"(LiveVLM),":[134],"it":[135],"delivers":[136],"$1.2\\times$":[137],"lower":[138],"peak":[139],"memory":[140,179],"$2\\times$":[142],"faster":[143],"TTFT.":[144],"StreamingTOM":[145],"state-of-the-art":[147],"accuracy":[148,162],"among":[149],"training-free":[150],"with":[152,176],"an":[153],"average":[154],"$63.8\\%$":[156],"benchmarks":[159],"$55.8\\%$":[161],"$3.7$":[164],"score":[165],"RVS.":[167],"These":[168],"results":[169],"real-time":[172],"understanding":[175],"is":[180],"achievable":[181],"without":[182],"model":[183],"retraining.":[184]},"counts_by_year":[{"year":2026,"cited_by_count":32},{"year":2025,"cited_by_count":2}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2025-10-24T00:00:00"}