{"id":"https://openalex.org/W6929240681","doi":"https://doi.org/10.48550/arxiv.2407.16207","title":"Graph-Structured Speculative Decoding","display_name":"Graph-Structured Speculative Decoding","publication_year":2024,"publication_date":"2024-07-23","ids":{"openalex":"https://openalex.org/W6929240681","doi":"https://doi.org/10.48550/arxiv.2407.16207"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2407.16207","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2407.16207","pdf_url":"https://arxiv.org/pdf/2407.16207","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"},"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/2407.16207","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Gong, Zhuocheng","orcid":null,"observed_orcids":[]},"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":"Gong, Zhuocheng","raw_affiliation_strings":["Wangxuan Institute of Computer Technology , Peking University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Wangxuan Institute of Computer Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Liu, Jiahao","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405259410","display_name":"Meituan (China)","ror":"https://ror.org/0272kpg92","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405259410"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liu, Jiahao","raw_affiliation_strings":["Meituan ;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Meituan ;","institution_ids":["https://openalex.org/I4405259410"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Wang, Ziyue","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I162868743","display_name":"Tianjin University","ror":"https://ror.org/012tb2g32","country_code":"CN","type":"education","lineage":["https://openalex.org/I162868743"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Ziyue","raw_affiliation_strings":["Tianjin University ;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tianjin University ;","institution_ids":["https://openalex.org/I162868743"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Wu, Pengfei","orcid":null,"observed_orcids":[]},"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":"Wu, Pengfei","raw_affiliation_strings":["Wangxuan Institute of Computer Technology , Peking University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Wangxuan Institute of Computer Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Wang, Jingang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405259410","display_name":"Meituan (China)","ror":"https://ror.org/0272kpg92","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405259410"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Jingang","raw_affiliation_strings":["Meituan ;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Meituan ;","institution_ids":["https://openalex.org/I4405259410"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Cai, Xunliang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405259410","display_name":"Meituan (China)","ror":"https://ror.org/0272kpg92","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405259410"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Cai, Xunliang","raw_affiliation_strings":["Meituan ;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Meituan ;","institution_ids":["https://openalex.org/I4405259410"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125768494","display_name":"Dongyan Zhao","orcid":null,"observed_orcids":[]},"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":"Zhao, Dongyan","raw_affiliation_strings":["National Key Laboratory of General Artificial Intelligence","Wangxuan Institute of Computer Technology , Peking University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National Key Laboratory of General Artificial Intelligence","institution_ids":[]},{"raw_affiliation_string":"Wangxuan Institute of Computer Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"last","author":{"id":null,"display_name":"Yan, Rui","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yan, Rui","raw_affiliation_strings":["Gaoling School of Artificial Intelligence , Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Gaoling School of Artificial Intelligence , Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":4,"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":true,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.8852999806404114,"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":0.8852999806404114,"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":0.09109999984502792,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.0034000000450760126,"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/speculative-decoding","display_name":"speculative decoding","score":0.9980000257492065},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9210000038146973},{"id":"https://openalex.org/keywords/inference-acceleration","display_name":"inference acceleration","score":0.8429999947547913},{"id":"https://openalex.org/keywords/draft-model","display_name":"draft model","score":0.7049999833106995},{"id":"https://openalex.org/keywords/directed-acyclic-graphs","display_name":"directed acyclic graphs","score":0.9760000109672546},{"id":"https://openalex.org/keywords/llama-2","display_name":"Llama 2","score":0.8600000143051147}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7993999719619751},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7249000072479248},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.6680999994277954},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5939000248908997},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5271000266075134},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4593999981880188},{"id":"https://openalex.org/C74197172","wikidata":"https://www.wikidata.org/wiki/Q1195339","display_name":"Directed acyclic graph","level":2,"score":0.4180999994277954},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.4020000100135803},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3797000050544739},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.3646000027656555},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.359499990940094},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3215000033378601},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.3133000135421753},{"id":"https://openalex.org/C147297375","wikidata":"https://www.wikidata.org/wiki/Q6674930","display_name":"Look-ahead","level":2,"score":0.3084000051021576},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3021000027656555},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.2881999909877777},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.275299996137619},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2711000144481659},{"id":"https://openalex.org/C146380142","wikidata":"https://www.wikidata.org/wiki/Q1137726","display_name":"Directed graph","level":2,"score":0.2590999901294708}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2407.16207","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2407.16207","pdf_url":"https://arxiv.org/pdf/2407.16207","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.2407.16207","doi":"https://doi.org/10.48550/arxiv.2407.16207","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2407.16207","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":"pmh:oai:arXiv.org:2407.16207","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2407.16207","pdf_url":"https://arxiv.org/pdf/2407.16207","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"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W6929240681.pdf","grobid_xml":"https://content.openalex.org/works/W6929240681.grobid-xml"},"referenced_works_count":8,"referenced_works":["https://openalex.org/W2978017171","https://openalex.org/W4229005866","https://openalex.org/W4308760226","https://openalex.org/W4322718191","https://openalex.org/W4378770729","https://openalex.org/W4395112660","https://openalex.org/W4401042997","https://openalex.org/W7133233472"],"related_works":["https://openalex.org/W7214036691","https://openalex.org/W4392120551","https://openalex.org/W7155965484","https://openalex.org/W4403346564","https://openalex.org/W4415274198","https://openalex.org/W7131685159","https://openalex.org/W4415274683","https://openalex.org/W4391590734","https://openalex.org/W7093018523","https://openalex.org/W4393300451"],"abstract_inverted_index":{"Speculative":[0,164],"decoding":[1],"has":[2],"emerged":[3],"as":[4],"a":[5,19,25,114,128,171,176,183],"promising":[6],"technique":[7],"to":[8,23,68,86,133,142,188],"accelerate":[9],"the":[10,33,43,50,60,69,82,91,105,135,152,156],"inference":[11],"of":[12,37,49,62,77,155,173,186],"Large":[13],"Language":[14],"Models":[15],"(LLMs)":[16],"by":[17,32,72,104],"employing":[18],"small":[20],"language":[21],"model":[22,107],"draft":[24,51,63,106,157],"hypothesis":[26],"sequence,":[27],"which":[28],"is":[29],"then":[30],"validated":[31],"LLM.":[34],"The":[35],"effectiveness":[36],"this":[38,120,161],"approach":[39,126,162],"heavily":[40],"relies":[41],"on":[42,58],"balance":[44],"between":[45],"performance":[46],"and":[47,89,145,181],"efficiency":[48],"model.":[52,158],"In":[53],"our":[54],"research,":[55],"we":[56,122],"focus":[57],"enhancing":[59],"proportion":[61],"tokens":[64],"that":[65,94,101],"are":[66],"accepted":[67],"final":[70],"output":[71],"generating":[73],"multiple":[74],"hypotheses":[75,102],"instead":[76],"just":[78],"one.":[79],"This":[80,138],"allows":[81],"LLM":[83],"more":[84],"options":[85],"choose":[87],"from":[88],"select":[90],"longest":[92],"sequence":[93],"meets":[95],"its":[96],"standards.":[97],"Our":[98],"analysis":[99],"reveals":[100],"produced":[103],"share":[108],"many":[109],"common":[110],"token":[111,148],"sequences,":[112,149],"suggesting":[113],"potential":[115],"for":[116],"optimizing":[117],"computation.":[118],"Leveraging":[119],"observation,":[121],"introduce":[123],"an":[124],"innovative":[125],"utilizing":[127],"directed":[129],"acyclic":[130],"graph":[131],"(DAG)":[132],"manage":[134],"drafted":[136],"hypotheses.":[137],"structure":[139],"enables":[140],"us":[141],"efficiently":[143],"predict":[144],"merge":[146],"recurring":[147],"vastly":[150],"reducing":[151],"computational":[153],"demands":[154],"We":[159,167],"term":[160],"Graph-structured":[163],"Decoding":[165],"(GSD).":[166],"apply":[168],"GSD":[169],"across":[170],"range":[172],"LLMs,":[174],"including":[175],"70-billion":[177],"parameter":[178],"LLaMA-2":[179],"model,":[180],"observe":[182],"remarkable":[184],"speedup":[185],"1.73$\\times$":[187],"1.96$\\times$,":[189],"significantly":[190],"surpassing":[191],"standard":[192],"speculative":[193],"decoding.":[194]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":3}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2025-10-10T00:00:00"}