{"id":"https://openalex.org/W7225311827","doi":"https://doi.org/10.1145/3858916","title":"One-Step Limits of Randomized Speculative Decoding","display_name":"One-Step Limits of Randomized Speculative Decoding","publication_year":2026,"publication_date":"2026-11-05","ids":{"openalex":"https://openalex.org/W7225311827","doi":"https://doi.org/10.1145/3858916"},"language":"en","primary_location":{"id":"doi:10.1145/3858916","doi":"https://doi.org/10.1145/3858916","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3858916","pdf_url":null,"source":{"id":"https://openalex.org/S7407056975","display_name":"ACM AI Letters","issn_l":"3068-8590","issn":["3068-8590"],"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":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM AI Letters","raw_type":"journal-article"},"volume":null,"type":"article","is_traditional":true,"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/A5002473893","display_name":"Hossein Javidnia","orcid":"https://orcid.org/0000-0002-5640-4798","observed_orcids":["https://orcid.org/0000-0002-5640-4798"]},"institutions":[{"id":"https://openalex.org/I42934936","display_name":"Dublin City University","ror":"https://ror.org/04a1a1e81","country_code":"IE","type":"education","lineage":["https://openalex.org/I42934936"]}],"countries":["IE"],"is_corresponding":true,"raw_author_name":"Hossein Javidnia","raw_affiliation_strings":["School of Computing, Dublin City University, Collins Ave Ext, Whitehall, Dublin 9, Ireland"],"raw_orcid":"https://orcid.org/0000-0002-5640-4798","affiliations":[{"raw_affiliation_string":"School of Computing, Dublin City University, Collins Ave Ext, Whitehall, Dublin 9, Ireland","institution_ids":["https://openalex.org/I42934936"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5002473893"],"corresponding_institution_ids":["https://openalex.org/I42934936"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.0,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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":0.9028000235557556,"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":0.9028000235557556,"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":0.06210000067949295,"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/T11269","display_name":"Algorithms and Data Compression","score":0.011300000362098217,"subfield":{"id":"https://openalex.org/subfields/2614","display_name":"Theoretical Computer Science"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/speculative-decoding","display_name":"speculative decoding","score":0.9990000128746033},{"id":"https://openalex.org/keywords/language-model-inference","display_name":"language model inference","score":0.7720000147819519},{"id":"https://openalex.org/keywords/parallel-verification","display_name":"parallel verification","score":0.5809999704360962},{"id":"https://openalex.org/keywords/collision-entropy","display_name":"collision entropy","score":0.7850000262260437},{"id":"https://openalex.org/keywords/inference-acceleration","display_name":"inference acceleration","score":0.6309999823570251}],"concepts":[{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.8852999806404114},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.6212999820709229},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6150000095367432},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5641999840736389},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.501800000667572},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.46480000019073486},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.4453999996185303},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.38920000195503235},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.387800008058548},{"id":"https://openalex.org/C2780428219","wikidata":"https://www.wikidata.org/wiki/Q16952335","display_name":"Cover (algebra)","level":2,"score":0.36340001225471497},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.35910001397132874},{"id":"https://openalex.org/C90805587","wikidata":"https://www.wikidata.org/wiki/Q10944557","display_name":"Word (group theory)","level":2,"score":0.3305000066757202},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.3296999931335449},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.299699991941452},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.27129998803138733},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.25769999623298645},{"id":"https://openalex.org/C2779795794","wikidata":"https://www.wikidata.org/wiki/Q7315343","display_name":"Reset (finance)","level":2,"score":0.2572000026702881}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3858916","doi":"https://doi.org/10.1145/3858916","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3858916","pdf_url":null,"source":{"id":"https://openalex.org/S7407056975","display_name":"ACM AI Letters","issn_l":"3068-8590","issn":["3068-8590"],"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":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM AI Letters","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":12,"referenced_works":["https://openalex.org/W4319166707","https://openalex.org/W4385749968","https://openalex.org/W4391124766","https://openalex.org/W4392933336","https://openalex.org/W4394973219","https://openalex.org/W4395112660","https://openalex.org/W4396821494","https://openalex.org/W4401306979","https://openalex.org/W4413360549","https://openalex.org/W4415800630","https://openalex.org/W7128006584","https://openalex.org/W7133563540"],"related_works":["https://openalex.org/W7140096387","https://openalex.org/W7153287649","https://openalex.org/W7166861610","https://openalex.org/W7203671024","https://openalex.org/W7169850532","https://openalex.org/W7172518259","https://openalex.org/W7166839924","https://openalex.org/W4392933336","https://openalex.org/W4407308985","https://openalex.org/W4417300753"],"abstract_inverted_index":{"Speculative":[0],"decoding":[1],"accelerates":[2],"language-model":[3],"inference":[4],"by":[5,70],"proposing":[6],"candidate":[7],"continuations":[8],"and":[9,37,54,119,128,139],"verifying":[10],"them":[11],"in":[12,97],"parallel.":[13],"This":[14],"letter":[15],"studies":[16],"a":[17,23,29],"deliberately":[18],"restricted":[19],"one-step":[20,99],"abstraction:":[21],"under":[22],"parallel":[24],"verification":[25],"budget":[26],"\\(P\\)":[27,35,73,124],",":[28],"non-anticipating":[30],"policy":[31],"proposes":[32],"up":[33],"to":[34],"continuations,":[36],"the":[38,42,60,71,77,94,105,144],"accepted":[39],"length":[40],"is":[41,68],"longest":[43],"verifier-consistent":[44],"prefix.":[45],"We":[46],"show":[47],"that,":[48],"for":[49],"any":[50,55,111],"verifier":[51],"prefix":[52,81,108,113],"law":[53],"deterministic":[56,107,135],"or":[57,133],"randomized":[58],"policy,":[59],"probability":[61],"of":[62,76],"accepting":[63],"at":[64,110],"least":[65],"\\(\\ell\\)":[66,80],"tokens":[67],"bounded":[69],"top-":[72],"mass":[74,92],"\\(M_{P}(\\ell)\\)":[75],"verifier's":[78],"length-":[79],"distribution.":[82],"Consequently,":[83],"\\(\\mathbb{E}[L_{P}]\\leq\\sum_{\\ell\\geq":[84],"1}M_{P}(\\ell)\\)":[85],".":[86],"The":[87],"result":[88],"identifies":[89],"covered":[90],"verifier-prefix":[91],"as":[93],"depthwise":[95],"bottleneck":[96],"this":[98],"model:":[100],"randomization":[101],"cannot":[102],"improve":[103],"on":[104],"best":[106],"cover":[109],"fixed":[112],"length.":[114],"Under":[115],"independent-token":[116],"specializations,":[117],"R\u00e9nyi":[118],"collision-entropy":[120],"bounds":[121],"yield":[122],"logarithmic-in-":[123],"limits,":[125],"while":[126],"uniform":[127],"spread-out":[129],"regimes":[130],"give":[131],"matching":[132],"near-matching":[134],"constructions.":[136],"Stylized":[137],"simulations":[138],"inference-only":[140],"logits":[141],"experiments":[142],"illustrate":[143],"predicted":[145],"diminishing":[146],"returns.":[147]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-11-06T00:00:00"}