{"id":"https://openalex.org/W7123360348","doi":"https://doi.org/10.1145/3788870","title":"Sometimes Painful but Promising: Feasibility and Trade-Offs of On-Device Language Model Inference","display_name":"Sometimes Painful but Promising: Feasibility and Trade-Offs of On-Device Language Model Inference","publication_year":2026,"publication_date":"2026-01-12","ids":{"openalex":"https://openalex.org/W7123360348","doi":"https://doi.org/10.1145/3788870"},"language":"en","primary_location":{"id":"doi:10.1145/3788870","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3788870","pdf_url":null,"source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1145/3788870","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119965467","display_name":"Maximilian Abstreiter","orcid":null},"institutions":[{"id":"https://openalex.org/I133731052","display_name":"University of Helsinki","ror":"https://ror.org/040af2s02","country_code":"FI","type":"education","lineage":["https://openalex.org/I133731052"]}],"countries":["FI"],"is_corresponding":false,"raw_author_name":"Maximilian Abstreiter","raw_affiliation_strings":["Department of Computer Science, University of Helsinki"],"raw_orcid":"https://orcid.org/0009-0002-9535-2393","affiliations":[{"raw_affiliation_string":"Department of Computer Science, University of Helsinki","institution_ids":["https://openalex.org/I133731052"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Sasu Tarkoma","orcid":"https://orcid.org/0000-0003-4220-3650"},"institutions":[{"id":"https://openalex.org/I133731052","display_name":"University of Helsinki","ror":"https://ror.org/040af2s02","country_code":"FI","type":"education","lineage":["https://openalex.org/I133731052"]}],"countries":["FI"],"is_corresponding":false,"raw_author_name":"Sasu Tarkoma","raw_affiliation_strings":["Department of Computer Science, University of Helsinki"],"raw_orcid":"https://orcid.org/0000-0003-4220-3650","affiliations":[{"raw_affiliation_string":"Department of Computer Science, University of Helsinki","institution_ids":["https://openalex.org/I133731052"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5076263478","display_name":"Roberto Morabito","orcid":"https://orcid.org/0000-0002-4240-9934"},"institutions":[{"id":"https://openalex.org/I133731052","display_name":"University of Helsinki","ror":"https://ror.org/040af2s02","country_code":"FI","type":"education","lineage":["https://openalex.org/I133731052"]},{"id":"https://openalex.org/I1902872","display_name":"EURECOM","ror":"https://ror.org/00sse7z02","country_code":"FR","type":"education","lineage":["https://openalex.org/I1902872","https://openalex.org/I205703379"]}],"countries":["FI","FR"],"is_corresponding":false,"raw_author_name":"Roberto Morabito","raw_affiliation_strings":["Communication Systems Department, EURECOM","Department of Computer Science, University of Helsinki"],"raw_orcid":"https://orcid.org/0000-0002-4240-9934","affiliations":[{"raw_affiliation_string":"Communication Systems Department, EURECOM","institution_ids":["https://openalex.org/I1902872"]},{"raw_affiliation_string":"Department of Computer Science, University of Helsinki","institution_ids":["https://openalex.org/I133731052"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":25.9972,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.98030017,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":"25","issue":"2","first_page":"1","last_page":"30"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T14347","display_name":"Big Data and Digital Economy","score":0.23579999804496765,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T14347","display_name":"Big Data and Digital Economy","score":0.23579999804496765,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.12080000340938568,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.06270000338554382,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.8064000010490417},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5388000011444092},{"id":"https://openalex.org/keywords/edge-device","display_name":"Edge device","score":0.48080000281333923},{"id":"https://openalex.org/keywords/enhanced-data-rates-for-gsm-evolution","display_name":"Enhanced Data Rates for GSM Evolution","score":0.47909998893737793},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.45649999380111694},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.43630000948905945},{"id":"https://openalex.org/keywords/generative-model","display_name":"Generative model","score":0.40049999952316284},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.3930000066757202}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.8064000010490417},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7447999715805054},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5388000011444092},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5081999897956848},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.48080000281333923},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.47909998893737793},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.47189998626708984},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.45649999380111694},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.43630000948905945},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.40049999952316284},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.3930000066757202},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.3815000057220459},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.3799999952316284},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3456999957561493},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.32820001244544983},{"id":"https://openalex.org/C45235069","wikidata":"https://www.wikidata.org/wiki/Q278425","display_name":"Table (database)","level":2,"score":0.31779998540878296},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.3061000108718872},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.29170000553131104},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.29100000858306885},{"id":"https://openalex.org/C2776187449","wikidata":"https://www.wikidata.org/wiki/Q1513879","display_name":"Natural language generation","level":3,"score":0.2770000100135803},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.274399995803833},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.2648000121116638}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3788870","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3788870","pdf_url":null,"source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1145/3788870","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3788870","pdf_url":null,"source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3791286706","display_name":null,"funder_award_id":"168043","funder_id":"https://openalex.org/F4320322885","funder_display_name":"NordForsk"},{"id":"https://openalex.org/G6729745602","display_name":"6G Bridge \u2014 Helsingin Yliopisto (8782/31/2022)","funder_award_id":"8782/31/2022","funder_id":"https://openalex.org/F4320328501","funder_display_name":"Business Finland"}],"funders":[{"id":"https://openalex.org/F4320322885","display_name":"NordForsk","ror":"https://ror.org/05bqzfg94"},{"id":"https://openalex.org/F4320328501","display_name":"Business Finland","ror":"https://ror.org/05bgf9v38"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":29,"referenced_works":["https://openalex.org/W2157331557","https://openalex.org/W2794325560","https://openalex.org/W2895432151","https://openalex.org/W2945856475","https://openalex.org/W2946609015","https://openalex.org/W2963341956","https://openalex.org/W2966983573","https://openalex.org/W2996908057","https://openalex.org/W3043571714","https://openalex.org/W3201174429","https://openalex.org/W4241372764","https://openalex.org/W4327810158","https://openalex.org/W4387797345","https://openalex.org/W4388725043","https://openalex.org/W4389672268","https://openalex.org/W4392271821","https://openalex.org/W4392617597","https://openalex.org/W4395474395","https://openalex.org/W4395686786","https://openalex.org/W4395704222","https://openalex.org/W4396686573","https://openalex.org/W4399534536","https://openalex.org/W4400397496","https://openalex.org/W4401211704","https://openalex.org/W4401306886","https://openalex.org/W4404918643","https://openalex.org/W4404990113","https://openalex.org/W4405031232","https://openalex.org/W6890025973"],"related_works":[],"abstract_inverted_index":{"The":[0,216],"rapid":[1],"rise":[2],"of":[3,11,29,31,90,101,127,166,218,242,246,252],"Language":[4],"Models":[5],"(LMs)":[6],"has":[7],"expanded":[8],"the":[9,66,87,97,108,116,190,207,221,240,244,250],"capabilities":[10],"natural":[12],"language":[13],"processing,":[14],"powering":[15],"applications":[16],"from":[17],"text":[18],"generation":[19],"to":[20],"complex":[21],"decision-making.":[22],"While":[23,170],"state-of-the-art":[24],"LMs":[25,69,219],"often":[26],"boast":[27],"hundreds":[28],"billions":[30],"parameters":[32],"and":[33,58,82,134,148,161,192,214,249],"are":[34],"primarily":[35],"deployed":[36],"in":[37,111,225],"data":[38,84],"centers,":[39],"recent":[40],"trends":[41],"show":[42],"a":[43,124,234],"growing":[44],"focus":[45],"on":[46,70,131],"compact":[47],"models\u2014typically":[48],"under":[49],"10":[50],"billion":[51],"parameters\u2013enabled":[52],"by":[53],"techniques":[54],"such":[55,76],"as":[56,77],"quantization":[57,171],"other":[59],"model":[60,168,210],"compression":[61],"techniques.":[62],"This":[63],"shift":[64],"paves":[65],"way":[67],"for":[68,184,199,236],"edge":[71,102,136,222],"devices,":[72],"offering":[73,203],"potential":[74],"benefits":[75],"enhanced":[78],"privacy,":[79],"reduced":[80],"latency,":[81],"improved":[83],"sovereignty.":[85],"However,":[86],"inherent":[88],"complexity":[89],"even":[91],"these":[92,120],"smaller":[93],"models,":[94,243],"combined":[95],"with":[96],"limited":[98],"computing":[99],"resources":[100],"hardware,":[103],"raises":[104],"critical":[105],"questions":[106],"about":[107],"practical":[109,200],"trade-offs":[110,208],"executing":[112],"LM":[113,129],"inference":[114,130,146,212,247],"outside":[115],"cloud.":[117],"To":[118],"address":[119],"challenges,":[121],"we":[122,155],"present":[123],"comprehensive":[125],"evaluation":[126],"generative":[128],"representative":[132],"CPU-based":[133],"GPU-accelerated":[135],"devices.":[137],"Our":[138,187],"study":[139,232],"measures":[140],"key":[141],"performance":[142],"indicators\u2014including":[143],"memory":[144,174,191],"usage,":[145],"speed,":[147],"energy":[149,193],"consumption\u2014across":[150],"various":[151],"device":[152],"configurations.":[153],"Additionally,":[154],"examine":[156],"throughput-energy":[157],"trade-offs,":[158],"cost":[159],"considerations,":[160],"usability,":[162],"alongside":[163],"an":[164],"assessment":[165],"qualitative":[167],"performance.":[169],"helps":[172],"mitigate":[173],"overhead,":[175],"it":[176],"does":[177],"not":[178],"fully":[179],"eliminate":[180],"resource":[181],"bottlenecks,":[182],"especially":[183],"larger":[185],"models.":[186],"findings":[188],"quantify":[189],"constraints":[194],"that":[195],"must":[196],"be":[197],"considered":[198],"real-world":[201],"deployments,":[202],"concrete":[204],"insights":[205],"into":[206],"between":[209],"size,":[211],"performance,":[213],"efficiency.":[215],"exploration":[217],"at":[220],"is":[223],"still":[224],"its":[226],"early":[227],"stages.":[228],"We":[229],"hope":[230],"this":[231],"provides":[233],"foundation":[235],"future":[237],"research,":[238],"guiding":[239],"refinement":[241],"enhancement":[245],"efficiency,":[248],"advancement":[251],"edge-centric":[253],"AI":[254],"systems.":[255]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-13T07:31:44.756512","created_date":"2026-01-14T00:00:00"}