{"id":"https://openalex.org/W7154055427","doi":"https://doi.org/10.48550/arxiv.2604.08971","title":"Modality-Aware Zero-Shot Pruning and Sparse Attention for Efficient Multimodal Edge Inference","display_name":"Modality-Aware Zero-Shot Pruning and Sparse Attention for Efficient Multimodal Edge Inference","publication_year":2026,"publication_date":"2026-04-10","ids":{"openalex":"https://openalex.org/W7154055427","doi":"https://doi.org/10.48550/arxiv.2604.08971"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.08971","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.08971","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":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.08971","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101443153","display_name":"Yueyuan Sui","orcid":"https://orcid.org/0009-0007-1602-5359"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sui, Yueyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133503132","display_name":"Payal Mohapatra","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mohapatra, Payal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120004999","display_name":"Do\u011fa\u00e7 Eldenk","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Eldenk, Do\u011fa\u00e7","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133529274","display_name":"Haodong Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Haodong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133528882","display_name":"Yiting Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yiting","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133490095","display_name":"Haoyan Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Haoyan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133526526","display_name":"Qi Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5051730108","display_name":"Stephen Xia","orcid":"https://orcid.org/0000-0001-5713-8885"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xia, Stephen","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":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/T10036","display_name":"Advanced Neural Network Applications","score":0.6011999845504761,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.6011999845504761,"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.08110000193119049,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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":0.039900001138448715,"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/bottleneck","display_name":"Bottleneck","score":0.6920999884605408},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.6787999868392944},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5982000231742859},{"id":"https://openalex.org/keywords/enhanced-data-rates-for-gsm-evolution","display_name":"Enhanced Data Rates for GSM Evolution","score":0.5713000297546387},{"id":"https://openalex.org/keywords/edge-device","display_name":"Edge device","score":0.5491999983787537},{"id":"https://openalex.org/keywords/edge-computing","display_name":"Edge computing","score":0.4471000134944916},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4456999897956848},{"id":"https://openalex.org/keywords/dropout","display_name":"Dropout (neural networks)","score":0.439300000667572},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.43849998712539673}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.815500020980835},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6920999884605408},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.6787999868392944},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6044999957084656},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5982000231742859},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.5713000297546387},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.5491999983787537},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4733999967575073},{"id":"https://openalex.org/C2778456923","wikidata":"https://www.wikidata.org/wiki/Q5337692","display_name":"Edge computing","level":3,"score":0.4471000134944916},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4456999897956848},{"id":"https://openalex.org/C2776145597","wikidata":"https://www.wikidata.org/wiki/Q25339462","display_name":"Dropout (neural networks)","level":2,"score":0.439300000667572},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.43849998712539673},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.40139999985694885},{"id":"https://openalex.org/C2911011789","wikidata":"https://www.wikidata.org/wiki/Q130741","display_name":"Hallucinating","level":2,"score":0.3955000042915344},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.36070001125335693},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.3472000062465668},{"id":"https://openalex.org/C2780226545","wikidata":"https://www.wikidata.org/wiki/Q6888030","display_name":"Modality (human\u2013computer interaction)","level":2,"score":0.33000001311302185},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.30959999561309814},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.305400013923645},{"id":"https://openalex.org/C2779903281","wikidata":"https://www.wikidata.org/wiki/Q6888026","display_name":"Modalities","level":2,"score":0.30379998683929443},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.3028999865055084},{"id":"https://openalex.org/C101738243","wikidata":"https://www.wikidata.org/wiki/Q786435","display_name":"Autoencoder","level":3,"score":0.2766999900341034},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.26919999718666077},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.2529999911785126},{"id":"https://openalex.org/C2984118289","wikidata":"https://www.wikidata.org/wiki/Q29954","display_name":"Power consumption","level":3,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.08971","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.08971","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.08971","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.08971","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":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7","score":0.4119962453842163}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Edge":[0],"devices":[1],"increasingly":[2],"run":[3],"multimodal":[4,100,116,122,169],"sensing":[5],"pipelines":[6],"that":[7,44],"must":[8],"remain":[9],"accurate":[10],"despite":[11],"fluctuating":[12],"power":[13],"budgets":[14],"and":[15,38,78,83,121,136,150],"unpredictable":[16],"sensor":[17],"dropout.":[18],"Existing":[19],"pruning":[20,134],"methods":[21],"fail":[22],"under":[23,140],"these":[24],"conditions:":[25],"they":[26,39],"generally":[27],"require":[28],"fine-tuning":[29],"after":[30],"compression,":[31],"consuming":[32],"over":[33,131],"$10\\times$":[34],"the":[35,54,132],"deployment":[36,87],"energy,":[37],"assign":[40],"static":[41],"importance":[42,70],"scores":[43,71],"are":[45,50],"blind":[46],"to":[47,138,155,168],"which":[48,57],"sensors":[49],"present.":[51],"We":[52],"present":[53],"SentryFuse":[55,145],"framework,":[56],"addresses":[58],"both":[59],"challenges":[60],"jointly":[61],"through":[62],"two":[63],"key":[64,96],"components.":[65],"First,":[66],"SentryGate":[67,124],"learns":[68],"modality-conditioned":[69],"during":[72],"training":[73],"via":[74],"first-order":[75],"saliency":[76],"supervision":[77],"then":[79],"prunes":[80],"attention":[81],"heads":[82],"feed-forward":[84],"channels":[85],"at":[86],"without":[88,157],"fine-tuning.":[89],"Second,":[90],"SentryAttend":[91],"replaces":[92],"dense":[93],"self-attention,":[94],"a":[95,107,126,165],"bottleneck":[97],"in":[98,111],"contemporary":[99],"architectures,":[101],"with":[102],"sparse":[103],"grouped-query":[104],"attention,":[105],"yielding":[106],"net":[108],"15%":[109],"reduction":[110],"GFLOPs":[112],"across":[113],"three":[114,119],"different":[115],"architectures.":[117],"Across":[118],"applications":[120],"backbones,":[123],"achieves":[125],"12.7%":[127],"average":[128],"accuracy":[129],"improvement":[130],"strongest":[133],"baseline,":[135],"upto":[137],"18%":[139],"modality":[141],"dropout":[142],"conditions.":[143],"Together,":[144],"reduces":[146],"memory":[147],"by":[148,153],"28.2%":[149],"lowers":[151],"latency":[152],"up":[154],"$1.63\\times$":[156],"further":[158],"fine-tuning,":[159],"establishing":[160],"modality-aware":[161],"zero-shot":[162],"compression":[163],"as":[164],"practical":[166],"path":[167],"intelligence":[170],"on":[171],"heterogeneous":[172],"edge":[173],"hardware.":[174]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-14T00:00:00"}