{"id":"https://openalex.org/W4415707927","doi":"https://doi.org/10.1109/icme59968.2025.11209796","title":"Enhancing Large Multimodal Models with Adaptive Sparsity and KV Cache Compression","display_name":"Enhancing Large Multimodal Models with Adaptive Sparsity and KV Cache Compression","publication_year":2025,"publication_date":"2025-06-30","ids":{"openalex":"https://openalex.org/W4415707927","doi":"https://doi.org/10.1109/icme59968.2025.11209796"},"language":null,"primary_location":{"id":"doi:10.1109/icme59968.2025.11209796","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icme59968.2025.11209796","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Multimedia and Expo (ICME)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5082807131","display_name":"Tingwei Zhang","orcid":"https://orcid.org/0009-0005-3748-0428"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Te Zhang","raw_affiliation_strings":["University of Michigan"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Michigan","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100710050","display_name":"Jiasheng Li","orcid":"https://orcid.org/0000-0002-3225-3665"},"institutions":[{"id":"https://openalex.org/I145311948","display_name":"Johns Hopkins University","ror":"https://ror.org/00za53h95","country_code":"US","type":"education","lineage":["https://openalex.org/I145311948"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuheng Li","raw_affiliation_strings":["Johns Hopkins University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Johns Hopkins University","institution_ids":["https://openalex.org/I145311948"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078516264","display_name":"Junxiang Wang","orcid":"https://orcid.org/0000-0003-4371-6212"},"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":"Junxiang Wang","raw_affiliation_strings":["Central South University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Central South University","institution_ids":["https://openalex.org/I139660479"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101983249","display_name":"Lujun Li","orcid":"https://orcid.org/0000-0002-0641-3178"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lujun Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"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":"1","last_page":"6"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8141999840736389,"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":0.8141999840736389,"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":0.05909999832510948,"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":0.013299999758601189,"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/cache","display_name":"Cache","score":0.6539000272750854},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5231999754905701},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.5169000029563904},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.5083000063896179},{"id":"https://openalex.org/keywords/cache-algorithms","display_name":"Cache algorithms","score":0.4440999925136566},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.429500013589859},{"id":"https://openalex.org/keywords/cpu-cache","display_name":"CPU cache","score":0.42800000309944153},{"id":"https://openalex.org/keywords/compression-ratio","display_name":"Compression ratio","score":0.4077000021934509},{"id":"https://openalex.org/keywords/lossless-compression","display_name":"Lossless compression","score":0.3894999921321869}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8349000215530396},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.6539000272750854},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5231999754905701},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5169000029563904},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5083000063896179},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.4440999925136566},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.429500013589859},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.42800000309944153},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.4171000123023987},{"id":"https://openalex.org/C25797200","wikidata":"https://www.wikidata.org/wiki/Q828137","display_name":"Compression ratio","level":3,"score":0.4077000021934509},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.39320001006126404},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.3894999921321869},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.3813999891281128},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.38119998574256897},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.36309999227523804},{"id":"https://openalex.org/C13481523","wikidata":"https://www.wikidata.org/wiki/Q412438","display_name":"Image compression","level":4,"score":0.35830000042915344},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.35589998960494995},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.3294000029563904},{"id":"https://openalex.org/C188045654","wikidata":"https://www.wikidata.org/wiki/Q17148339","display_name":"Memory bandwidth","level":2,"score":0.322299987077713},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.29409998655319214},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.29170000553131104},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.290800005197525},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2809000015258789},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.2791000008583069},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.27709999680519104},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C94835093","wikidata":"https://www.wikidata.org/wiki/Q3113333","display_name":"Data compression ratio","level":5,"score":0.26820001006126404},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2619999945163727},{"id":"https://openalex.org/C176809094","wikidata":"https://www.wikidata.org/wiki/Q15401496","display_name":"Traverse","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.2590999901294708}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icme59968.2025.11209796","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icme59968.2025.11209796","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Multimedia and Expo (ICME)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":13,"referenced_works":["https://openalex.org/W2560730294","https://openalex.org/W2963122961","https://openalex.org/W2979382951","https://openalex.org/W4312590184","https://openalex.org/W4386071640","https://openalex.org/W4389523832","https://openalex.org/W4390871868","https://openalex.org/W4390874023","https://openalex.org/W4402670135","https://openalex.org/W4402727764","https://openalex.org/W4403878467","https://openalex.org/W4403878569","https://openalex.org/W4404356490"],"related_works":[],"abstract_inverted_index":{"Large":[0],"multimodal":[1],"models":[2],"(LMMs)":[3],"have":[4],"advanced":[5],"significantly":[6],"by":[7],"integrating":[8],"visual":[9],"encoders":[10],"with":[11,83],"extensive":[12],"language":[13],"models,":[14],"enabling":[15],"robust":[16],"reasoning":[17],"capabilities.":[18],"However,":[19],"compressing":[20],"LMMs":[21],"for":[22,97],"deployment":[23],"on":[24,111],"edge":[25],"devices":[26],"remains":[27],"a":[28,89,146],"critical":[29],"challenge.":[30],"In":[31],"this":[32],"work,":[33],"we":[34],"propose":[35],"an":[36],"adaptive":[37],"search":[38],"algorithm":[39],"that":[40,93],"optimizes":[41],"sparsity":[42],"and":[43,62,87,117,129],"KV":[44,63,141],"cache":[45,64,85,142],"compression":[46,105,133,143],"to":[47],"enhance":[48],"LMM":[49,69,150],"efficiency.":[50],"Utilizing":[51],"the":[52,75,95],"Tree-structured":[53],"Parzen":[54],"Estimator,":[55],"our":[56,120,136],"method":[57],"dynamically":[58],"adjusts":[59],"pruning":[60,82,91],"ratios":[61],"quantization":[65,86],"bandwidth":[66],"across":[67,131],"different":[68],"layers,":[70],"using":[71],"model":[72],"performance":[73],"as":[74,127],"optimization":[76],"objective.":[77],"This":[78],"approach":[79],"uniquely":[80],"combines":[81],"key-value":[84],"incorporates":[88],"fast":[90],"technique":[92],"eliminates":[94],"need":[96],"additional":[98],"fine-tuning":[99],"or":[100],"weight":[101],"adjustments,":[102],"achieving":[103],"efficient":[104],"without":[106,155],"compromising":[107],"accuracy.":[108],"Comprehensive":[109],"evaluations":[110],"benchmark":[112],"datasets,":[113],"including":[114],"LLaVA-1.5":[115],"7B":[116],"13B,":[118],"demonstrate":[119],"method\u2019s":[121],"superiority":[122],"over":[123],"state-of-the-art":[124],"techniques":[125],"such":[126],"SparseGPT":[128],"Wanda":[130],"various":[132],"levels.":[134],"Notably,":[135],"framework\u2019s":[137],"automatic":[138],"allocation":[139],"of":[140],"resources":[144],"sets":[145],"new":[147],"standard":[148],"in":[149],"optimization,":[151],"delivering":[152],"memory":[153],"efficiency":[154],"sacrificing":[156],"much":[157],"performance.":[158],"Code":[159],"is":[160],"available":[161],"at":[162],"https://github.com/tezhang65/optspa.git":[163]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-30T00:00:00"}