{"id":"https://openalex.org/W7224078308","doi":"https://doi.org/10.48550/arxiv.2610.35957","title":"Distributed LLM Inference via Communication-Efficient Speculative Decoding","display_name":"Distributed LLM Inference via Communication-Efficient Speculative Decoding","publication_year":2026,"publication_date":"2026-10-25","ids":{"openalex":"https://openalex.org/W7224078308","doi":"https://doi.org/10.48550/arxiv.2610.35957"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.35957","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.35957","pdf_url":"https://arxiv.org/pdf/2610.35957","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.35957","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103905432","display_name":"Joohyun Shin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Shin, Joohyun","raw_affiliation_strings":["School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160034512","display_name":"Byeonghyeon Park","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Park, Byeonghyeon","raw_affiliation_strings":["School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120694063","display_name":"Keumseo Ryum","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Ryum, Keumseo","raw_affiliation_strings":["School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5160037590","display_name":"Joonhyuk Kang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Kang, Joonhyuk","raw_affiliation_strings":["School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical Engineering , Korea Advanced Institute of Science and Technology (KAIST) , Daejeon , South Korea","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I157485424"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.5249000191688538,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.5249000191688538,"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/T10028","display_name":"Topic Modeling","score":0.19670000672340393,"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":0.03460000082850456,"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/large-language-model-inference","display_name":"large language model inference","score":0.9710000157356262},{"id":"https://openalex.org/keywords/communication-efficiency","display_name":"communication efficiency","score":0.9449999928474426},{"id":"https://openalex.org/keywords/rejection-sampling","display_name":"rejection sampling","score":0.9520000219345093},{"id":"https://openalex.org/keywords/decoding-speedup","display_name":"decoding speedup","score":0.7429999709129333}],"concepts":[{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.867900013923645},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8083000183105469},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.7457000017166138},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6753000020980835},{"id":"https://openalex.org/C2775937380","wikidata":"https://www.wikidata.org/wiki/Q1232589","display_name":"Replica","level":2,"score":0.5637000203132629},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.44920000433921814},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.44369998574256897},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.35690000653266907},{"id":"https://openalex.org/C193969084","wikidata":"https://www.wikidata.org/wiki/Q7452500","display_name":"Sequential decoding","level":4,"score":0.3481000065803528},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.34279999136924744},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.33149999380111694},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3221000134944916},{"id":"https://openalex.org/C204397858","wikidata":"https://www.wikidata.org/wiki/Q4437907","display_name":"List decoding","level":5,"score":0.3158999979496002},{"id":"https://openalex.org/C115067241","wikidata":"https://www.wikidata.org/wiki/Q1639854","display_name":"Token passing","level":3,"score":0.3086000084877014},{"id":"https://openalex.org/C130120984","wikidata":"https://www.wikidata.org/wiki/Q2835898","display_name":"Distributed algorithm","level":2,"score":0.2962000072002411},{"id":"https://openalex.org/C110157686","wikidata":"https://www.wikidata.org/wiki/Q922122","display_name":"Broadcasting (networking)","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.27970001101493835},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.2685000002384186},{"id":"https://openalex.org/C153180980","wikidata":"https://www.wikidata.org/wiki/Q19776675","display_name":"Commit","level":2,"score":0.26739999651908875},{"id":"https://openalex.org/C2777317252","wikidata":"https://www.wikidata.org/wiki/Q18393516","display_name":"Rare events","level":2,"score":0.2639999985694885},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.2572000026702881},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C70061542","wikidata":"https://www.wikidata.org/wiki/Q989016","display_name":"Distributed database","level":2,"score":0.2529999911785126}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.35957","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.35957","pdf_url":"https://arxiv.org/pdf/2610.35957","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.35957","doi":"https://doi.org/10.48550/arxiv.2610.35957","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.35957","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.35957","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.35957","pdf_url":"https://arxiv.org/pdf/2610.35957","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7224078308.pdf","grobid_xml":"https://content.openalex.org/works/W7224078308.grobid-xml"},"referenced_works_count":8,"referenced_works":["https://openalex.org/W3177813494","https://openalex.org/W3210904070","https://openalex.org/W4229005866","https://openalex.org/W4327810158","https://openalex.org/W4384918448","https://openalex.org/W4391766565","https://openalex.org/W4392822465","https://openalex.org/W4405655184"],"related_works":["https://openalex.org/W4416350477","https://openalex.org/W7152101543","https://openalex.org/W7212124719","https://openalex.org/W4417293528","https://openalex.org/W4414687626","https://openalex.org/W7080322902","https://openalex.org/W7120232590","https://openalex.org/W7126224786","https://openalex.org/W4416781475","https://openalex.org/W4402683901"],"abstract_inverted_index":{"Distributed":[0],"speculative":[1,32,71,148],"decoding":[2,19,33,143,149],"has":[3],"emerged":[4],"as":[5],"a":[6,66,89,114],"promising":[7],"approach":[8],"for":[9,69,110],"device--server":[10],"collaborative":[11],"large":[12],"language":[13,96],"model":[14,97],"(LLM)":[15],"inference,":[16],"but":[17],"its":[18],"speedup":[20,144],"is":[21],"often":[22],"limited":[23],"by":[24,87],"the":[25,44,48,51,80,93,101,106,127,130],"vocabulary-scale":[26,75],"communication":[27],"overhead.":[28],"In":[29],"existing":[30,146],"distributed":[31,70,147],"methods,":[34,150],"exact":[35,81],"rejection":[36,111],"sampling":[37],"requires":[38],"transmitting":[39],"full":[40],"token":[41,120],"distributions":[42],"over":[43],"LLM":[45],"vocabulary":[46],"between":[47,126],"device":[49,128],"and":[50,122,129],"server,":[52],"which":[53,99],"becomes":[54],"costly":[55],"in":[56],"bandwidth-constrained":[57],"environments.":[58],"To":[59],"address":[60],"this":[61,86],"issue,":[62],"we":[63],"propose":[64],"DISCO,":[65],"communication-efficient":[67],"framework":[68],"decoding,":[72],"that":[73,138],"eliminates":[74],"distribution":[76,108],"transfer":[77],"while":[78],"preserving":[79],"target":[82],"distribution.":[83],"DISCO":[84,116,139],"achieves":[85,141],"introducing":[88],"server-side":[90],"replica":[91],"of":[92],"draft":[94],"small":[95],"(SLM),":[98],"enables":[100],"server":[102],"to":[103],"locally":[104],"obtain":[105],"draft-side":[107],"required":[109],"sampling.":[112],"As":[113],"result,":[115],"exchanges":[117],"only":[118],"drafted":[119],"indices":[121],"scalar":[123],"probability":[124],"values":[125],"server.":[131],"Experimental":[132],"results":[133],"on":[134],"multiple":[135],"datasets":[136],"show":[137],"consistently":[140],"higher":[142],"than":[145],"particularly":[151],"under":[152],"stringent":[153],"bandwidth":[154],"constraints.":[155]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-29T00:00:00"}