{"id":"https://openalex.org/W7162806841","doi":"https://doi.org/10.1145/3745756.3809205","title":"ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference","display_name":"ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference","publication_year":2026,"publication_date":"2026-05-29","ids":{"openalex":"https://openalex.org/W7162806841","doi":"https://doi.org/10.1145/3745756.3809205"},"language":null,"primary_location":{"id":"doi:10.1145/3745756.3809205","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3745756.3809205","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 24th Annual International Conference on Mobile Systems, Applications and Services","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3745756.3809205","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101293258","display_name":"Wangsong Yin","orcid":null},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wangsong Yin","raw_affiliation_strings":["Key Lab of HCST (PKU), MOE; SCS, Peking University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-6242-4368","affiliations":[{"raw_affiliation_string":"Key Lab of HCST (PKU), MOE; SCS, Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034382142","display_name":"Daliang Xu","orcid":"https://orcid.org/0000-0002-6775-0688"},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Daliang Xu","raw_affiliation_strings":["Beijing University of Posts and Telecommunications, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-6775-0688","affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications, Beijing, China","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137360503","display_name":"Mengwei Xu","orcid":"https://orcid.org/0000-0001-6271-6993"},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Mengwei Xu","raw_affiliation_strings":["Beijing University of Posts and Telecommunications, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-6271-6993","affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications, Beijing, China","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102000802","display_name":"Gang Huang","orcid":"https://orcid.org/0000-0001-8515-8584"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Gang Huang","raw_affiliation_strings":["Key Lab of HCST (PKU), MOE; SCS, Peking University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-4686-3181","affiliations":[{"raw_affiliation_string":"Key Lab of HCST (PKU), MOE; SCS, Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5137364616","display_name":"Xuanzhe Liu","orcid":"https://orcid.org/0000-0002-7908-8484"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xuanzhe Liu","raw_affiliation_strings":["Key Lab of HCST (PKU), MOE; SCS, Peking University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-7908-8484","affiliations":[{"raw_affiliation_string":"Key Lab of HCST (PKU), MOE; SCS, Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.87339614,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"294","last_page":"308"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.2590999901294708,"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"}},"topics":[{"id":"https://openalex.org/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.2590999901294708,"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"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.19189999997615814,"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":0.08919999748468399,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.5932000279426575},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.545799970626831},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5426999926567078},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5099999904632568},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.4796000123023987},{"id":"https://openalex.org/keywords/mobile-device","display_name":"Mobile device","score":0.4749999940395355},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.4032999873161316},{"id":"https://openalex.org/keywords/sensitivity","display_name":"Sensitivity (control systems)","score":0.33250001072883606}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7771999835968018},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.5932000279426575},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.545799970626831},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5426999926567078},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5099999904632568},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.4796000123023987},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4779999852180481},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.4749999940395355},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.4032999873161316},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.39739999175071716},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.33250001072883606},{"id":"https://openalex.org/C17020691","wikidata":"https://www.wikidata.org/wiki/Q139677","display_name":"Operator (biology)","level":5,"score":0.32510000467300415},{"id":"https://openalex.org/C144543869","wikidata":"https://www.wikidata.org/wiki/Q2738570","display_name":"Mobile computing","level":2,"score":0.3239000141620636},{"id":"https://openalex.org/C2984822820","wikidata":"https://www.wikidata.org/wiki/Q1123036","display_name":"Processor scheduling","level":3,"score":0.31839999556541443},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2913999855518341},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.29100000858306885},{"id":"https://openalex.org/C106516650","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm design","level":2,"score":0.28700000047683716},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2858999967575073},{"id":"https://openalex.org/C55416958","wikidata":"https://www.wikidata.org/wiki/Q6206757","display_name":"Job shop scheduling","level":3,"score":0.28220000863075256},{"id":"https://openalex.org/C196921405","wikidata":"https://www.wikidata.org/wiki/Q786431","display_name":"Online algorithm","level":2,"score":0.27410000562667847},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.26759999990463257},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C115051666","wikidata":"https://www.wikidata.org/wiki/Q6522493","display_name":"Ranging","level":2,"score":0.2549000084400177},{"id":"https://openalex.org/C163173736","wikidata":"https://www.wikidata.org/wiki/Q3308558","display_name":"Key generation","level":3,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3745756.3809205","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3745756.3809205","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 24th Annual International Conference on Mobile Systems, Applications and Services","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3745756.3809205","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3745756.3809205","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 24th Annual International Conference on Mobile Systems, Applications and Services","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":24,"referenced_works":["https://openalex.org/W2251939518","https://openalex.org/W2963926728","https://openalex.org/W3088076788","https://openalex.org/W4283020086","https://openalex.org/W4283206588","https://openalex.org/W4313377427","https://openalex.org/W4380926962","https://openalex.org/W4394998528","https://openalex.org/W4399534536","https://openalex.org/W4399919514","https://openalex.org/W4402683442","https://openalex.org/W4403333972","https://openalex.org/W4404034656","https://openalex.org/W4404862388","https://openalex.org/W4405717632","https://openalex.org/W4406346870","https://openalex.org/W4407196790","https://openalex.org/W4408858386","https://openalex.org/W4408888778","https://openalex.org/W4408901326","https://openalex.org/W4408977762","https://openalex.org/W4410068143","https://openalex.org/W4412944812","https://openalex.org/W4416286402"],"related_works":[],"abstract_inverted_index":{"Running":[0],"LLMs":[1],"on":[2,68],"devices":[3],"like":[4],"smartphones":[5],"has":[6],"become":[7],"a":[8,38,59,74],"key":[9,81],"catalyst":[10],"towards":[11],"privacy-preserving":[12],"mobile":[13],"AI.":[14],"In":[15],"state-of-the-art":[16],"frameworks,":[17],"the":[18,24,28,86,90,132],"attention":[19,63],"operator":[20],"falls":[21],"back":[22],"from":[23],"dedicated":[25],"NPU":[26,108],"to":[27,32,35,41,84,96,120,127,146],"public":[29],"CPU/GPU":[30,69,138,144],"due":[31],"its":[33],"sensitivity":[34],"quantization.":[36],"Such":[37],"fallback":[39],"leads":[40],"hampered":[42],"user":[43],"experience":[44],"and":[45,115,124],"extra":[46],"system":[47],"scheduling":[48],"complexity.":[49],"To":[50],"realize":[51],"NPU-centric":[52],"LLM":[53],"inference,":[54],"this":[55],"paper":[56],"presents":[57],"shadowNPU,":[58],"system-algorithm":[60],"codesigned":[61],"sparse":[62],"module":[64],"with":[65,135],"minimal":[66],"reliance":[67],"by":[70,93],"only":[71],"sparsely":[72],"computing":[73],"very":[75],"small":[76],"number":[77],"of":[78,88,100,150],"tokens.":[79],"The":[80],"idea":[82],"is":[83],"hide":[85],"overhead":[87],"estimating":[89],"important":[91],"tokens":[92],"offloading":[94],"it":[95,102,140],"NPU.":[97],"On":[98],"top":[99],"this,":[101],"further":[103],"incorporates":[104],"insightful":[105],"techniques":[106],"including":[107],"compute":[109],"graph":[110],"bucketing,":[111],"head-wise":[112],"NPU-CPU/GPU":[113],"pipeline":[114],"per-head":[116],"fine-grained":[117],"sparsity":[118],"ratio":[119],"achieve":[121,147],"high":[122],"accuracy":[123],"efficiency.":[125],"Compared":[126],"design":[128],"alternatives,":[129],"shadowNPU":[130],"achieves":[131],"best":[133],"performance":[134,149],"strictly":[136],"limited":[137],"resource;":[139],"requires":[141],"much":[142],"less":[143],"resource":[145],"on-par":[148],"SoTA":[151],"frameworks.":[152]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-30T00:00:00"}