{"id":"https://openalex.org/W7208739326","doi":"https://doi.org/10.1145/3795866.3844470","title":"LeanStream: A Speculate-and-Refine Streaming Framework for Efficient on-Device LLM Inference","display_name":"LeanStream: A Speculate-and-Refine Streaming Framework for Efficient on-Device LLM Inference","publication_year":2026,"publication_date":"2026-10-21","ids":{"openalex":"https://openalex.org/W7208739326","doi":"https://doi.org/10.1145/3795866.3844470"},"language":"en","primary_location":{"id":"doi:10.1145/3795866.3844470","doi":"https://doi.org/10.1145/3795866.3844470","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3795866.3844470","pdf_url":null,"source":{"id":"https://openalex.org/S4363608994","display_name":"ACM Annual International Conference on Mobile Computing and Networking (MobiCom)","issn_l":"1543-5679","issn":["1543-5679"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 32nd Annual International Conference on Mobile Computing and Networking","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1145/3795866.3844470","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5150026295","display_name":"Renyuan Liu","orcid":"https://orcid.org/0000-0001-9710-6116","observed_orcids":["https://orcid.org/0000-0001-9710-6116"]},"institutions":[{"id":"https://openalex.org/I162714631","display_name":"George Mason University","ror":"https://ror.org/02jqj7156","country_code":"US","type":"education","lineage":["https://openalex.org/I162714631"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Renyuan Liu","raw_affiliation_strings":["George Mason University, Fairfax, VA, USA"],"raw_orcid":"https://orcid.org/0000-0001-9710-6116","affiliations":[{"raw_affiliation_string":"George Mason University, Fairfax, VA, USA","institution_ids":["https://openalex.org/I162714631"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048432270","display_name":"Yuyang Leng","orcid":"https://orcid.org/0009-0008-9376-0880","observed_orcids":["https://orcid.org/0009-0008-9376-0880"]},"institutions":[{"id":"https://openalex.org/I162714631","display_name":"George Mason University","ror":"https://ror.org/02jqj7156","country_code":"US","type":"education","lineage":["https://openalex.org/I162714631"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuyang Leng","raw_affiliation_strings":["George Mason University, Fairfax, VA, USA"],"raw_orcid":"https://orcid.org/0009-0008-9376-0880","affiliations":[{"raw_affiliation_string":"George Mason University, Fairfax, VA, USA","institution_ids":["https://openalex.org/I162714631"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016488771","display_name":"Kaiyan Liu","orcid":"https://orcid.org/0009-0009-8663-0698","observed_orcids":["https://orcid.org/0009-0009-8663-0698"]},"institutions":[{"id":"https://openalex.org/I162714631","display_name":"George Mason University","ror":"https://ror.org/02jqj7156","country_code":"US","type":"education","lineage":["https://openalex.org/I162714631"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kaiyan Liu","raw_affiliation_strings":["George Mason University, Fairfax, VA, USA"],"raw_orcid":"https://orcid.org/0009-0009-8663-0698","affiliations":[{"raw_affiliation_string":"George Mason University, Fairfax, VA, USA","institution_ids":["https://openalex.org/I162714631"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5149995670","display_name":"Yuzhou Zhong","orcid":"https://orcid.org/0009-0005-1783-0072","observed_orcids":["https://orcid.org/0009-0005-1783-0072"]},"institutions":[{"id":"https://openalex.org/I162714631","display_name":"George Mason University","ror":"https://ror.org/02jqj7156","country_code":"US","type":"education","lineage":["https://openalex.org/I162714631"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuzhou Zhong","raw_affiliation_strings":["George Mason University, Fairfax, VA, USA"],"raw_orcid":"https://orcid.org/0009-0005-1783-0072","affiliations":[{"raw_affiliation_string":"George Mason University, Fairfax, VA, USA","institution_ids":["https://openalex.org/I162714631"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5032992976","display_name":"Shaohan Hu","orcid":"https://orcid.org/0000-0002-2877-2665","observed_orcids":["https://orcid.org/0000-0002-2877-2665"]},"institutions":[{"id":"https://openalex.org/I1305429384","display_name":"JPMorgan Chase & Co (United States)","ror":"https://ror.org/01x3kkr08","country_code":"US","type":"company","lineage":["https://openalex.org/I1305429384"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shaohan Hu","raw_affiliation_strings":["JPMorganChase, New York, NY, USA"],"raw_orcid":"https://orcid.org/0000-0002-2877-2665","affiliations":[{"raw_affiliation_string":"JPMorganChase, New York, NY, USA","institution_ids":["https://openalex.org/I1305429384"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120452844","display_name":"Chun-Fu","orcid":"https://orcid.org/0000-0002-5912-5620","observed_orcids":["https://orcid.org/0000-0002-5912-5620"]},"institutions":[{"id":"https://openalex.org/I1305429384","display_name":"JPMorgan Chase & Co (United States)","ror":"https://ror.org/01x3kkr08","country_code":"US","type":"company","lineage":["https://openalex.org/I1305429384"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chun-Fu (Richard) Chen","raw_affiliation_strings":["JPMorganChase, New York, NY, USA"],"raw_orcid":"https://orcid.org/0000-0002-5912-5620","affiliations":[{"raw_affiliation_string":"JPMorganChase, New York, NY, USA","institution_ids":["https://openalex.org/I1305429384"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135046610","display_name":"Chen","orcid":"https://orcid.org/0000-0003-2843-6941","observed_orcids":["https://orcid.org/0000-0003-2843-6941"]},"institutions":[{"id":"https://openalex.org/I1305429384","display_name":"JPMorgan Chase & Co (United States)","ror":"https://ror.org/01x3kkr08","country_code":"US","type":"company","lineage":["https://openalex.org/I1305429384"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Peijun Zhao","raw_affiliation_strings":["JPMorganChase, New York, NY, USA"],"raw_orcid":"https://orcid.org/0000-0003-2843-6941","affiliations":[{"raw_affiliation_string":"JPMorganChase, New York, NY, USA","institution_ids":["https://openalex.org/I1305429384"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5064659321","display_name":"Heechul Yun","orcid":"https://orcid.org/0000-0002-8515-2622","observed_orcids":["https://orcid.org/0000-0002-8515-2622"]},"institutions":[{"id":"https://openalex.org/I146416000","display_name":"University of Kansas","ror":"https://ror.org/001tmjg57","country_code":"US","type":"education","lineage":["https://openalex.org/I146416000"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Heechul Yun","raw_affiliation_strings":["University of Kansas, Lawrence, KS, USA"],"raw_orcid":"https://orcid.org/0000-0002-8515-2622","affiliations":[{"raw_affiliation_string":"University of Kansas, Lawrence, KS, USA","institution_ids":["https://openalex.org/I146416000"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5005026261","display_name":"Shuochao Yao","orcid":"https://orcid.org/0000-0001-7446-1430","observed_orcids":["https://orcid.org/0000-0001-7446-1430","https://orcid.org/0000-0002-4070-6345"]},"institutions":[{"id":"https://openalex.org/I162714631","display_name":"George Mason University","ror":"https://ror.org/02jqj7156","country_code":"US","type":"education","lineage":["https://openalex.org/I162714631"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shuochao Yao","raw_affiliation_strings":["George Mason University, Fairfax, VA, USA"],"raw_orcid":"https://orcid.org/0000-0001-7446-1430","affiliations":[{"raw_affiliation_string":"George Mason University, Fairfax, VA, USA","institution_ids":["https://openalex.org/I162714631"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.71922484,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1035","last_page":"1050"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9429000020027161,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9429000020027161,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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":0.021299999207258224,"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.007899999618530273,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/streaming-inference","display_name":"streaming inference","score":0.6629999876022339},{"id":"https://openalex.org/keywords/activation-sparsity","display_name":"activation sparsity","score":0.9359999895095825},{"id":"https://openalex.org/keywords/gpu-memory-usage","display_name":"GPU memory usage","score":0.41999998688697815},{"id":"https://openalex.org/keywords/mobile-and-embedded-devices","display_name":"mobile and embedded devices","score":0.9010000228881836}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8845999836921692},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6531000137329102},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.6306999921798706},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6043000221252441},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.5949000120162964},{"id":"https://openalex.org/C52723943","wikidata":"https://www.wikidata.org/wiki/Q1127410","display_name":"Serialization","level":2,"score":0.552299976348877},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.542900025844574},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.5353000164031982},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3864000141620636},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.351500004529953},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.34790000319480896},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3474999964237213},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.335099995136261},{"id":"https://openalex.org/C126042441","wikidata":"https://www.wikidata.org/wiki/Q1324888","display_name":"Frame (networking)","level":2,"score":0.33070001006126404},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.29589998722076416},{"id":"https://openalex.org/C2989134064","wikidata":"https://www.wikidata.org/wiki/Q288510","display_name":"Execution time","level":2,"score":0.2953000068664551},{"id":"https://openalex.org/C2779191767","wikidata":"https://www.wikidata.org/wiki/Q6887075","display_name":"Mobile cloud computing","level":3,"score":0.28929999470710754},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.27399998903274536},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2703999876976013},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.26420000195503235},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.2581000030040741},{"id":"https://openalex.org/C73602740","wikidata":"https://www.wikidata.org/wiki/Q7795822","display_name":"Thompson sampling","level":3,"score":0.2556000053882599}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1145/3795866.3844470","doi":"https://doi.org/10.1145/3795866.3844470","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3795866.3844470","pdf_url":null,"source":{"id":"https://openalex.org/S4363608994","display_name":"ACM Annual International Conference on Mobile Computing and Networking (MobiCom)","issn_l":"1543-5679","issn":["1543-5679"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 32nd Annual International Conference on Mobile Computing and Networking","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2609.03079","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03079","pdf_url":"https://arxiv.org/pdf/2609.03079","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.03079","doi":"https://doi.org/10.48550/arxiv.2609.03079","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.03079","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.1145/3795866.3844470","doi":"https://doi.org/10.1145/3795866.3844470","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3795866.3844470","pdf_url":null,"source":{"id":"https://openalex.org/S4363608994","display_name":"ACM Annual International Conference on Mobile Computing and Networking (MobiCom)","issn_l":"1543-5679","issn":["1543-5679"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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 32nd Annual International Conference on Mobile Computing and Networking","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G3013630393","display_name":"Collaborative Research: CPS: Medium: Real-time Criticality-Aware Neural Networks for Mission-critical Cyber-Physical Systems","funder_award_id":"2038658","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G5705582075","display_name":"Collaborative Research: CPS: Medium: Real-time Criticality-Aware Neural Networks for Mission-critical Cyber-Physical Systems","funder_award_id":"2038923","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G7323781115","display_name":"III: Medium: Collaborative Research: Principled Uncertainty Quantification in Deep Learning Models for Time Series Analysis","funder_award_id":"2107200","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4412374994","https://openalex.org/W4405717632","https://openalex.org/W4415820843","https://openalex.org/W7169588453","https://openalex.org/W7085633039","https://openalex.org/W7162799881","https://openalex.org/W7161491009","https://openalex.org/W4411486373","https://openalex.org/W7163217967","https://openalex.org/W7161697739"],"abstract_inverted_index":{"On-device":[0],"LLM":[1,87,123],"inference":[2,124],"is":[3],"attractive":[4],"for":[5,84],"privacy":[6],"and":[7,14,29,73,94,107,116],"responsiveness,":[8],"but":[9,37],"remains":[10],"challenging":[11],"on":[12,113],"mobile":[13,115],"embedded":[15,117],"devices":[16],"because":[17],"model":[18],"weights":[19,31],"far":[20],"exceed":[21],"available":[22],"DRAM.":[23],"Prior":[24],"systems":[25,41],"exploit":[26],"activation":[27],"sparsity":[28],"offload":[30],"to":[32],"SSD":[33],"or":[34,66],"flash":[35],"storage,":[36],"face":[38],"a":[39,59,80],"fundamental":[40],"tradeoff:":[42],"accurate":[43],"sparse":[44],"execution":[45,65,106],"decisions":[46],"require":[47],"the":[48,133],"latest":[49],"context,":[50],"whereas":[51],"efficient":[52,85],"computation-I/O":[53],"overlap":[54,103],"requires":[55],"early":[56],"prediction.":[57],"As":[58],"result,":[60],"existing":[61],"designs":[62],"either":[63],"serialize":[64],"incur":[67],"redundant":[68],"weight":[69],"fetches,":[70],"extra":[71],"computation,":[72,92],"large":[74],"cache":[75],"overheads.":[76],"We":[77,110],"present":[78],"LeanStream,":[79],"streaming":[81],"speculate-and-refine":[82],"framework":[83],"on-device":[86,122],"inference.":[88],"LeanStream":[89,112,126],"progressively":[90],"refines":[91],"loading,":[93],"cache-retention":[95],"priorities":[96],"using":[97],"partial":[98],"GPU":[99,105],"results,":[100],"enabling":[101],"fine-grained":[102],"between":[104],"storage":[108],"I/O.":[109],"implement":[111],"both":[114],"platforms.":[118],"Compared":[119],"with":[120],"prior":[121,138],"systems,":[125],"reduces":[127],"memory":[128],"usage":[129],"by":[130,137,146],"4.8\u00d7-7.5\u00d7":[131],"at":[132],"best":[134],"throughput":[135,145],"achieved":[136],"work,":[139],"while":[140],"further":[141],"improving":[142],"token":[143],"generation":[144],"1.6\u00d7-2.1\u00d7.":[147]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-05T00:00:00"}