{"id":"https://openalex.org/W7204138699","doi":"https://doi.org/10.48550/arxiv.2608.20359","title":"Self-Speculation for Faster Reasoning Models","display_name":"Self-Speculation for Faster Reasoning Models","publication_year":2026,"publication_date":"2026-06-17","ids":{"openalex":"https://openalex.org/W7204138699","doi":"https://doi.org/10.48550/arxiv.2608.20359"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.20359","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.20359","pdf_url":"https://arxiv.org/pdf/2608.20359","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.20359","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5099111481","display_name":"Ravisri Valluri","orcid":"https://orcid.org/0009-0008-5651-877X","observed_orcids":["https://orcid.org/0009-0008-5651-877X"]},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Valluri, Ravisri","raw_affiliation_strings":["University of California , Los Angeles"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Los Angeles","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125413148","display_name":"Tung Nguyen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nguyen, Tung","raw_affiliation_strings":["University of California , Los Angeles"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Los Angeles","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5014409944","display_name":"Aditya Grover","orcid":"https://orcid.org/0000-0002-4217-6624","observed_orcids":["https://orcid.org/0000-0002-4217-6624"]},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Grover, Aditya","raw_affiliation_strings":["University of California , Los Angeles"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Los Angeles","institution_ids":["https://openalex.org/I161318765"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I161318765"],"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":0.6416000127792358,"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":0.6416000127792358,"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":0.3257000148296356,"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":0.002899999963119626,"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/chain-of-thought","display_name":"chain-of-thought","score":0.7839999794960022},{"id":"https://openalex.org/keywords/structured-generation","display_name":"structured generation","score":0.503000020980835},{"id":"https://openalex.org/keywords/long-form-generation","display_name":"long-form generation","score":0.9139999747276306},{"id":"https://openalex.org/keywords/reasoning-traces","display_name":"reasoning traces","score":0.3919999897480011},{"id":"https://openalex.org/keywords/multi-step-decision-making","display_name":"multi-step decision-making","score":0.6740000247955322},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.6349999904632568}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8359000086784363},{"id":"https://openalex.org/C2779804580","wikidata":"https://www.wikidata.org/wiki/Q102047","display_name":"Suffix","level":2,"score":0.7135000228881836},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.532800018787384},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5213000178337097},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.475600004196167},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4496999979019165},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.4368000030517578},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.37040001153945923},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.36039999127388},{"id":"https://openalex.org/C100660578","wikidata":"https://www.wikidata.org/wiki/Q18733","display_name":"Recall","level":2,"score":0.3027999997138977},{"id":"https://openalex.org/C203357204","wikidata":"https://www.wikidata.org/wiki/Q1089605","display_name":"Chunking (psychology)","level":2,"score":0.28949999809265137},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2768000066280365},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C129916263","wikidata":"https://www.wikidata.org/wiki/Q1141183","display_name":"Backward chaining","level":4,"score":0.267300009727478},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2630999982357025},{"id":"https://openalex.org/C17305859","wikidata":"https://www.wikidata.org/wiki/Q382944","display_name":"Soar","level":2,"score":0.2612000107765198},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.259799987077713},{"id":"https://openalex.org/C34736171","wikidata":"https://www.wikidata.org/wiki/Q918333","display_name":"Preprocessor","level":2,"score":0.259799987077713}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.20359","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.20359","pdf_url":"https://arxiv.org/pdf/2608.20359","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.2608.20359","doi":"https://doi.org/10.48550/arxiv.2608.20359","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.20359","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:2608.20359","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.20359","pdf_url":"https://arxiv.org/pdf/2608.20359","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/W7204138699.pdf","grobid_xml":"https://content.openalex.org/works/W7204138699.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416892671","https://openalex.org/W7117299489","https://openalex.org/W7116819995","https://openalex.org/W4416959577","https://openalex.org/W4416033863","https://openalex.org/W4417299515","https://openalex.org/W4392118267","https://openalex.org/W4415336479","https://openalex.org/W7131319044","https://openalex.org/W4400104534"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,239],"(LLMs)":[3],"are":[4],"deployed":[5],"for":[6,33,71,236],"increasingly":[7],"complex":[8],"tasks":[9,21,196,217],"involving":[10],"planning":[11],"and":[12,35,98,128,153,184,204,214,223,243],"multi-step":[13],"decision":[14],"making,":[15],"but":[16],"high-quality":[17],"performance":[18],"on":[19,57,117,151,195,211,232],"these":[20],"often":[22,124],"requires":[23],"generating":[24],"long":[25,142],"reasoning":[26,65,113],"traces.":[27],"This":[28,115],"is":[29,220],"a":[30,74,85,181,225],"poor":[31],"fit":[32],"latency-sensitive":[34],"interactive":[36],"applications":[37],"like":[38],"voice":[39],"assistants":[40],"or":[41],"coding":[42],"agents,":[43],"where":[44,218],"generation":[45,155,216,234],"latency":[46,194,235],"can":[47,140],"strongly":[48],"affect":[49],"user":[50],"experience.":[51],"Existing":[52],"acceleration":[53],"methods":[54],"typically":[55],"focus":[56],"token-level":[58],"generation,":[59],"without":[60],"utilizing":[61],"the":[62,81,91,96,99,103,108,118,132,163,177,189,202,205],"structure":[63],"of":[64,87,228],"workflows.":[66],"We":[67,208],"introduce":[68],"SSR:":[69],"Self-Speculation":[70],"Reasoning":[72],"Models,":[73],"training-free":[75],"self-speculative":[76],"decoding":[77],"method":[78],"that":[79,120],"leverages":[80],"chain-of-thought":[82],"(CoT)":[83],"as":[84,95,102,241],"source":[86],"speculation.":[88],"SSR":[89,139,171,210],"uses":[90],"partial-CoT":[92,122],"answer":[93],"distribution":[94,101],"drafter":[97],"full-CoT":[100],"verifier,":[104],"deriving":[105],"both":[106],"from":[107],"same":[109],"model":[110],"at":[111,145],"different":[112],"budgets.":[114],"builds":[116],"observation":[119],"later":[121],"responses":[123],"exhibit":[125],"greater":[126],"semantic":[127],"lexical":[129,199],"overlap":[130,161,200],"with":[131,197],"full-budget":[133],"response.":[134,207],"Due":[135],"to":[136,148,179,230],"this":[137],"overlap,":[138],"accept":[141],"draft":[143,178,203],"prefixes":[144],"once,":[146],"leading":[147],"large":[149],"speedups":[150],"structured":[152,213],"long-form":[154,215],"tasks.":[156],"To":[157],"further":[158,192],"exploit":[159],"draft-response":[160],"beyond":[162,188],"contiguous":[164],"prefix":[165],"accepted":[166,190],"by":[167],"standard":[168],"speculative":[169],"decoding,":[170,175],"also":[172],"incorporates":[173],"suffix":[174,182],"using":[176],"seed":[180],"cache":[183],"recover":[185],"useful":[186],"spans":[187],"prefix,":[191],"reducing":[193],"high":[198],"between":[201],"final":[206],"evaluate":[209],"multiple":[212],"it":[219],"most":[221],"useful,":[222],"demonstrate":[224],"relative":[226],"improvement":[227],"up":[229],"24.1%":[231],"total":[233],"popular":[237],"open-source":[238],"such":[240],"Qwen3.5":[242],"Gemma-4.":[244]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-25T00:00:00"}