{"id":"https://openalex.org/W7127982658","doi":"https://doi.org/10.48550/arxiv.2602.04804","title":"OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models","display_name":"OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models","publication_year":2026,"publication_date":"2026-02-04","ids":{"openalex":"https://openalex.org/W7127982658","doi":"https://doi.org/10.48550/arxiv.2602.04804"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.04804","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","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":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125121433","display_name":"Yue Ding","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ding, Yue","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125078854","display_name":"Yiyan Ji","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ji, Yiyan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125160117","display_name":"Jungang Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Jungang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121190807","display_name":"Xuyang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Xuyang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5073154896","display_name":"Xinlong Chen","orcid":"https://orcid.org/0009-0002-1763-3122"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Xinlong","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125151161","display_name":"Junfei Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Junfei","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062449586","display_name":"Bozhou Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Bozhou","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125165574","display_name":"Bohan Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Bohan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125229196","display_name":"Yang Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Yang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065487736","display_name":"Yushuo Guan","orcid":"https://orcid.org/0000-0001-5258-2397"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guan, Yushuo","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125192149","display_name":"Yuanxing Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yuanxing","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125190405","display_name":"Jiaheng Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Jiaheng","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125225059","display_name":"Qiang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Qiang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125192459","display_name":"Pengfei Wan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wan, Pengfei","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5125122055","display_name":"Liang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Liang","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":15,"corresponding_author_ids":["https://openalex.org/A5125121433"],"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/T10860","display_name":"Speech and Audio Processing","score":0.5253999829292297,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10860","display_name":"Speech and Audio Processing","score":0.5253999829292297,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.20900000631809235,"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":0.10329999774694443,"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":0.7871999740600586},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.51910001039505},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.5074999928474426},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.4754999876022339},{"id":"https://openalex.org/keywords/compression","display_name":"Compression (physics)","score":0.4032000005245209},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.3952000141143799},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.37059998512268066}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8252999782562256},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7871999740600586},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.51910001039505},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.5074999928474426},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.4754999876022339},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.4032000005245209},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.3952000141143799},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.37059998512268066},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3357999920845032},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.3095000088214874},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3084999918937683},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.29820001125335693},{"id":"https://openalex.org/C115067241","wikidata":"https://www.wikidata.org/wiki/Q1639854","display_name":"Token passing","level":3,"score":0.2980000078678131},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2930999994277954},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.2921999990940094},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.28760001063346863},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.28610000014305115},{"id":"https://openalex.org/C94835093","wikidata":"https://www.wikidata.org/wiki/Q3113333","display_name":"Data compression ratio","level":5,"score":0.27900001406669617},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2727000117301941},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.2526000142097473}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.04804","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2602.04804","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.04804","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.04804","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","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":{"Omni-modal":[0],"Large":[1],"Language":[2],"Models":[3],"(Omni-LLMs)":[4],"have":[5],"demonstrated":[6],"strong":[7],"capabilities":[8],"in":[9],"audio-video":[10],"understanding":[11],"tasks.":[12,161],"However,":[13],"their":[14],"reliance":[15],"on":[16,107,159],"long":[17],"multimodal":[18],"token":[19,29,52,142],"sequences":[20],"leads":[21],"to":[22],"substantial":[23],"computational":[24],"overhead.":[25],"Despite":[26],"this":[27,39],"challenge,":[28],"compression":[30,53,63,148],"methods":[31],"designed":[32],"for":[33,56,119],"Omni-LLMs":[34],"remain":[35],"limited.":[36],"To":[37],"bridge":[38],"gap,":[40],"we":[41],"propose":[42],"OmniSIFT":[43,59,121,144],"(Omni-modal":[44],"Spatio-temporal":[45],"Informed":[46],"Fine-grained":[47],"Token":[48],"compression),":[49],"a":[50,61,66,85,101],"modality-asymmetric":[51],"framework":[54,96],"tailored":[55],"Omni-LLMs.":[57],"Specifically,":[58],"adopts":[60],"two-stage":[62],"strategy:":[64],"(i)":[65],"spatio-temporal":[67],"video":[68,73],"pruning":[69],"module":[70,89],"that":[71,90],"removes":[72],"redundancy":[74],"arising":[75],"from":[76],"both":[77],"intra-frame":[78],"structure":[79],"and":[80,83,114,150],"inter-frame":[81],"overlap,":[82],"(ii)":[84],"vision-guided":[86],"audio":[87,92],"selection":[88],"filters":[91],"tokens.":[93],"The":[94],"entire":[95],"is":[97],"optimized":[98],"end-to-end":[99],"via":[100],"differentiable":[102],"straight-through":[103],"estimator.":[104],"Extensive":[105],"experiments":[106],"five":[108],"representative":[109],"benchmarks":[110],"demonstrate":[111],"the":[112,140,153,156],"efficacy":[113],"robustness":[115],"of":[116,139,155],"OmniSIFT.":[117],"Notably,":[118],"Qwen2.5-Omni-7B,":[120],"introduces":[122],"only":[123],"4.85M":[124],"parameters":[125],"while":[126],"maintaining":[127],"lower":[128],"latency":[129],"than":[130],"training-free":[131],"baselines":[132,149],"such":[133],"as":[134],"OmniZip.":[135],"With":[136],"merely":[137],"25%":[138],"original":[141],"context,":[143],"consistently":[145],"outperforms":[146],"all":[147],"even":[151],"surpasses":[152],"performance":[154],"full-token":[157],"model":[158],"several":[160]},"counts_by_year":[],"updated_date":"2026-04-04T16:13:02.066488","created_date":"2026-02-07T00:00:00"}