{"id":"https://openalex.org/W4416146691","doi":"https://doi.org/10.1145/3712285.3759834","title":"Towards Efficient LLM Inference via Collective and Adaptive Speculative Decoding","display_name":"Towards Efficient LLM Inference via Collective and Adaptive Speculative Decoding","publication_year":2025,"publication_date":"2025-11-12","ids":{"openalex":"https://openalex.org/W4416146691","doi":"https://doi.org/10.1145/3712285.3759834"},"language":null,"primary_location":{"id":"doi:10.1145/3712285.3759834","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3712285.3759834","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101635813","display_name":"Siqi Wang","orcid":"https://orcid.org/0009-0008-0761-3287"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Siqi Wang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-0761-3287","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018705589","display_name":"Hailong Yang","orcid":"https://orcid.org/0000-0003-1101-7927"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hailong Yang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-1101-7927","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113128939","display_name":"Xuezhu Wang","orcid":"https://orcid.org/0009-0001-7293-545X"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xuezhu Wang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0001-7293-545X","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103156062","display_name":"Tongxuan Liu","orcid":"https://orcid.org/0009-0007-2634-2788"},"institutions":[{"id":"https://openalex.org/I126520041","display_name":"University of Science and Technology of China","ror":"https://ror.org/04c4dkn09","country_code":"CN","type":"education","lineage":["https://openalex.org/I126520041","https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tongxuan Liu","raw_affiliation_strings":["University of Science and Technology of China, Hefei, China"],"raw_orcid":"https://orcid.org/0009-0007-2634-2788","affiliations":[{"raw_affiliation_string":"University of Science and Technology of China, Hefei, China","institution_ids":["https://openalex.org/I126520041"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100704040","display_name":"Pengbo Wang","orcid":"https://orcid.org/0009-0001-7197-2873"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Pengbo Wang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0001-7197-2873","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015151039","display_name":"Yufan Xu","orcid":"https://orcid.org/0000-0002-7787-6460"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yufan Xu","raw_affiliation_strings":["Independent Researcher, Cupertino, USA"],"raw_orcid":"https://orcid.org/0000-0002-7787-6460","affiliations":[{"raw_affiliation_string":"Independent Researcher, Cupertino, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066660539","display_name":"Xuning Liang","orcid":"https://orcid.org/0009-0002-7275-201X"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xuning Liang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0002-7275-201X","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100585370","display_name":"Kejie Ma","orcid":"https://orcid.org/0009-0002-0393-7627"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Kejie Ma","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0002-0393-7627","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010232454","display_name":"Tianyu Feng","orcid":"https://orcid.org/0000-0002-4830-5482"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tianyu Feng","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-4830-5482","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100683033","display_name":"Xin You","orcid":"https://orcid.org/0000-0002-5163-4607"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xin You","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-5163-4607","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063549968","display_name":"Ruihao Gong","orcid":"https://orcid.org/0000-0002-0081-5395"},"institutions":[{"id":"https://openalex.org/I4210128910","display_name":"Group Sense (China)","ror":"https://ror.org/036wd5777","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210128910"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ruihao Gong","raw_affiliation_strings":["SenseTime Research, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-0081-5395","affiliations":[{"raw_affiliation_string":"SenseTime Research, Shanghai, China","institution_ids":["https://openalex.org/I4210128910"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100431390","display_name":"Rui Wang","orcid":"https://orcid.org/0000-0003-2741-6033"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Rui Wang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-2741-6033","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074183877","display_name":"Zhongzhi Luan","orcid":"https://orcid.org/0000-0002-7186-0556"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhongzhi Luan","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-7186-0556","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100330624","display_name":"Yi Liu","orcid":"https://orcid.org/0000-0003-1829-2817"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yi Liu","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-1829-2817","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5079362609","display_name":"Depei Qian","orcid":"https://orcid.org/0000-0002-5382-1473"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Depei Qian","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-5382-1473","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"973","last_page":"990"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.1509000062942505,"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/T10028","display_name":"Topic Modeling","score":0.1509000062942505,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.09560000151395798,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.058400001376867294,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7962999939918518},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.7851999998092651},{"id":"https://openalex.org/keywords/speculation","display_name":"Speculation","score":0.6568999886512756},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.6144999861717224},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5436000227928162},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.3425999879837036},{"id":"https://openalex.org/keywords/autoregressive-model","display_name":"Autoregressive model","score":0.33980000019073486}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8382999897003174},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7962999939918518},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7851999998092651},{"id":"https://openalex.org/C47941915","wikidata":"https://www.wikidata.org/wiki/Q107885","display_name":"Speculation","level":2,"score":0.6568999886512756},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.6144999861717224},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5436000227928162},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.3425999879837036},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.33980000019073486},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.3327000141143799},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.2924000024795532},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.2768000066280365},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.2702000141143799},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.26829999685287476},{"id":"https://openalex.org/C2779194965","wikidata":"https://www.wikidata.org/wiki/Q11711607","display_name":"Ixabepilone","level":5,"score":0.26660001277923584},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.26080000400543213},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.25459998846054077},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2531999945640564},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.25099998712539673}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3712285.3759834","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3712285.3759834","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2732425316","display_name":null,"funder_award_id":"JKF-20240598, JKF-2025012343648","funder_id":"https://openalex.org/F4320335787","funder_display_name":"Fundamental Research Funds for the Central Universities"},{"id":"https://openalex.org/G6545491469","display_name":null,"funder_award_id":"62322201, U23B2020","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320335787","display_name":"Fundamental Research Funds for the Central Universities","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":14,"referenced_works":["https://openalex.org/W2093825590","https://openalex.org/W2252136820","https://openalex.org/W2951583236","https://openalex.org/W2963929190","https://openalex.org/W2998617917","https://openalex.org/W3012479151","https://openalex.org/W3012514909","https://openalex.org/W3155584966","https://openalex.org/W4385571011","https://openalex.org/W4387321091","https://openalex.org/W4387494819","https://openalex.org/W4395112660","https://openalex.org/W4400409890","https://openalex.org/W4405756215"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,51],"(LLMs)":[3],"have":[4],"gained":[5],"considerable":[6],"attention":[7],"for":[8],"their":[9],"remarkable":[10],"performance":[11,74,157],"across":[12],"a":[13,36,99,129,142],"wide":[14],"range":[15],"of":[16,25,63,70,136,151,159,171,175],"tasks.":[17],"However,":[18,58],"efficient":[19],"LLM":[20,56,71,83,89,109,125,184],"inference":[21,84,90,176,185],"remains":[22],"challenging":[23],"because":[24],"the":[26,45,59,66,134,147,156,169,182],"autoregressive":[27],"decoding":[28,39],"process,":[29],"which":[30],"generates":[31],"only":[32],"one":[33],"token":[34],"at":[35,153],"time.":[37],"Speculative":[38],"has":[40],"been":[41],"introduced":[42],"to":[43,53,87,106,132,144,181],"address":[44],"limitation":[46],"by":[47],"using":[48],"small":[49],"speculative":[50,95],"(SSMs)":[52],"speed":[54],"up":[55],"inference.":[57],"low":[60],"acceptance":[61],"rate":[62],"SSMs":[64,105],"and":[65,93,127,162,178],"high":[67,116],"verification":[68,117,126,163],"cost":[69],"prohibit":[72],"further":[73],"improvement.":[75],"In":[76],"this":[77],"paper,":[78],"we":[79],"present":[80],"Smurfs,":[81],"an":[82],"system":[85],"designed":[86],"accelerate":[88],"through":[91],"collective":[92],"adaptive":[94],"decoding.":[96],"Smurfs":[97,140,172],"adopts":[98],"majority-voted":[100],"mechanism":[101,143],"that":[102],"harnesses":[103],"multiple":[104],"collaboratively":[107],"predict":[108],"outputs":[110],"in":[111,173],"multi-task":[112],"scenarios,":[113],"while":[114],"avoiding":[115],"cost.":[118,164],"It":[119],"also":[120],"decouples":[121],"SSM":[122,137,152],"speculation":[123,149],"from":[124],"uses":[128],"pipelined":[130],"execution":[131],"hide":[133],"latency":[135,179],"speculation.":[138],"Additionally,":[139],"proposes":[141],"dynamically":[145],"determine":[146],"optimal":[148],"length":[150],"runtime,":[154],"balancing":[155],"impact":[158],"accepted":[160],"tokens":[161],"The":[165],"experimental":[166],"results":[167],"demonstrate":[168],"superiority":[170],"terms":[174],"throughput":[177],"compared":[180],"state-of-the-art":[183],"systems.":[186]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-11-12T00:00:00"}