{"id":"https://openalex.org/W7130704221","doi":"https://doi.org/10.48550/arxiv.2602.16994","title":"Dynamic Delayed Tree Expansion For Improved Multi-Path Speculative Decoding","display_name":"Dynamic Delayed Tree Expansion For Improved Multi-Path Speculative Decoding","publication_year":2026,"publication_date":"2026-02-19","ids":{"openalex":"https://openalex.org/W7130704221","doi":"https://doi.org/10.48550/arxiv.2602.16994"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2602.16994","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.16994","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2602.16994","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126494371","display_name":"Rahul Thomas","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Thomas, Rahul","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126523494","display_name":"Teo Kitanovski","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kitanovski, Teo","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066564672","display_name":"Micah Goldblum","orcid":"https://orcid.org/0000-0002-8266-2424"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Goldblum, Micah","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5045751957","display_name":"Arka Pal","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pal, Arka","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5126494371"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.3702000081539154,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.3702000081539154,"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":0.33309999108314514,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.05640000104904175,"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/tree-traversal","display_name":"Tree traversal","score":0.848800003528595},{"id":"https://openalex.org/keywords/tree","display_name":"Tree (set theory)","score":0.5360999703407288},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.48809999227523804},{"id":"https://openalex.org/keywords/block","display_name":"Block (permutation group theory)","score":0.4683000147342682},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.41440001130104065},{"id":"https://openalex.org/keywords/lossless-compression","display_name":"Lossless compression","score":0.3846000134944916},{"id":"https://openalex.org/keywords/range","display_name":"Range (aeronautics)","score":0.3779999911785126}],"concepts":[{"id":"https://openalex.org/C140745168","wikidata":"https://www.wikidata.org/wiki/Q1210082","display_name":"Tree traversal","level":2,"score":0.848800003528595},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7027999758720398},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5985999703407288},{"id":"https://openalex.org/C113174947","wikidata":"https://www.wikidata.org/wiki/Q2859736","display_name":"Tree (set theory)","level":2,"score":0.5360999703407288},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.48809999227523804},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.4683000147342682},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.41440001130104065},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.3846000134944916},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.3779999911785126},{"id":"https://openalex.org/C80478641","wikidata":"https://www.wikidata.org/wiki/Q195771","display_name":"Sequential analysis","level":2,"score":0.3476000130176544},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.3427000045776367},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.33889999985694885},{"id":"https://openalex.org/C34736171","wikidata":"https://www.wikidata.org/wiki/Q918333","display_name":"Preprocessor","level":2,"score":0.3303000032901764},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2935999929904938},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.28769999742507935},{"id":"https://openalex.org/C163797641","wikidata":"https://www.wikidata.org/wiki/Q2067937","display_name":"Tree structure","level":3,"score":0.2662000060081482},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2653999924659729},{"id":"https://openalex.org/C207024777","wikidata":"https://www.wikidata.org/wiki/Q621673","display_name":"Search tree","level":3,"score":0.25589999556541443},{"id":"https://openalex.org/C17231256","wikidata":"https://www.wikidata.org/wiki/Q5156540","display_name":"Completeness (order theory)","level":2,"score":0.2547999918460846}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2602.16994","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.16994","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2602.16994","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.16994","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multi-path":[0],"speculative":[1],"decoding":[2],"accelerates":[3],"lossless":[4],"sampling":[5,71,214],"from":[6,176],"a":[7,12,18,26,31,60,134,162,207],"target":[8,120,151,179],"model":[9,15,67],"by":[10],"using":[11],"cheaper":[13],"draft":[14,19,104,115,118,177],"to":[16,193],"generate":[17],"tree":[20,130,147],"of":[21,33,63,102,172,210],"tokens,":[22],"and":[23,70,73,119,153,178,213],"then":[24],"applies":[25],"verification":[27,41,64,174],"algorithm":[28],"that":[29,75,89,145,166],"accepts":[30],"subset":[32],"these.":[34],"While":[35],"prior":[36],"work":[37],"has":[38],"proposed":[39],"various":[40],"algorithms":[42],"for":[43,197],"i.i.d":[44,157],"rollouts,":[45],"their":[46],"relative":[47],"performance":[48],"under":[49],"matched":[50],"settings":[51],"remains":[52],"unclear.":[53],"In":[54],"this":[55,90,125],"work,":[56],"we":[57,127,160],"firstly":[58],"present":[59],"systematic":[61],"evaluation":[62],"strategies":[65],"across":[66,206],"families,":[68],"tasks,":[69],"regimes,":[72],"find":[74],"Traversal":[76,195],"Verification":[77,196],"dominates":[78],"consistently,":[79],"with":[80],"OT-based":[81,93,189],"methods":[82,94,175,190],"lagging":[83],"far":[84],"behind.":[85],"Our":[86,185],"analysis":[87],"uncovers":[88],"occurs":[91],"because":[92],"achieve":[95],"high":[96],"multi-token":[97,107],"acceptance":[98],"near":[99],"the":[100,103,114,139,150,168,198],"root":[101],"tree,":[105,116],"while":[106],"gains":[108],"are":[109],"most":[110],"impactful":[111],"deeper":[112],"in":[113],"where":[117],"distributions":[121],"diverge.":[122],"Based":[123],"on":[124,155],"insight,":[126],"propose":[128],"delayed":[129,146],"expansion,":[131],"which":[132],"drafts":[133],"partial":[135],"single":[136],"path,":[137],"delaying":[138],"i.i.d.":[140],"branching":[141],"point.":[142],"We":[143],"show":[144],"expansion":[148,183],"preserves":[149],"distribution":[152],"improves":[154],"root-node":[156],"rollouts.":[158],"Further,":[159],"develop":[161],"dynamic":[163],"neural":[164,186],"selector":[165,187],"estimates":[167],"expected":[169],"block":[170],"efficiency":[171],"optimal-transport-based":[173],"features,":[180],"enabling":[181],"context-dependent":[182],"decisions.":[184],"allows":[188],"like":[191],"SpecInfer":[192],"outperform":[194],"first":[199],"time,":[200],"achieving":[201],"5%":[202],"higher":[203],"average":[204],"throughput":[205],"wide":[208],"range":[209],"models,":[211],"datasets,":[212],"settings.":[215]},"counts_by_year":[],"updated_date":"2026-02-21T06:16:09.471975","created_date":"2026-02-21T00:00:00"}