{"id":"https://openalex.org/W7161263451","doi":"https://doi.org/10.48550/arxiv.2605.14978","title":"Performance-Driven Policy Optimization for Speculative Decoding with Adaptive Windowing","display_name":"Performance-Driven Policy Optimization for Speculative Decoding with Adaptive Windowing","publication_year":2026,"publication_date":"2026-05-14","ids":{"openalex":"https://openalex.org/W7161263451","doi":"https://doi.org/10.48550/arxiv.2605.14978"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.14978","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.14978","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,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.14978","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136190889","display_name":"Jie Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Jie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136234470","display_name":"Xing Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Xing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Chen, Ruotian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Ruotian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Su, Jianan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Su, Jianan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Shen, Kaixin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, Kaixin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"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":0.20260000228881836,"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":0.20260000228881836,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1517000049352646,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.055799998342990875,"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/decoding-methods","display_name":"Decoding methods","score":0.7767999768257141},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6177999973297119},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.4781999886035919},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.45350000262260437},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.3912999927997589},{"id":"https://openalex.org/keywords/look-ahead","display_name":"Look-ahead","score":0.36320000886917114},{"id":"https://openalex.org/keywords/prefix","display_name":"Prefix","score":0.3531000018119812},{"id":"https://openalex.org/keywords/deadlock","display_name":"Deadlock","score":0.3375999927520752}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8158000111579895},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.7767999768257141},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6177999973297119},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.4781999886035919},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.45350000262260437},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3912999927997589},{"id":"https://openalex.org/C147297375","wikidata":"https://www.wikidata.org/wiki/Q6674930","display_name":"Look-ahead","level":2,"score":0.36320000886917114},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.3531000018119812},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35089999437332153},{"id":"https://openalex.org/C159023740","wikidata":"https://www.wikidata.org/wiki/Q623276","display_name":"Deadlock","level":2,"score":0.3375999927520752},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3359000086784363},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.30640000104904175},{"id":"https://openalex.org/C126388530","wikidata":"https://www.wikidata.org/wiki/Q1131737","display_name":"Imitation","level":2,"score":0.29120001196861267},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.2890999913215637},{"id":"https://openalex.org/C193969084","wikidata":"https://www.wikidata.org/wiki/Q7452500","display_name":"Sequential decoding","level":4,"score":0.2782999873161316},{"id":"https://openalex.org/C149672232","wikidata":"https://www.wikidata.org/wiki/Q337048","display_name":"Adaptive optimization","level":2,"score":0.2653999924659729},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.2653000056743622},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.2590000033378601},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.25130000710487366}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.14978","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.14978","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,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.14978","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.14978","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,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Speculative":[0],"decoding":[1,136,152],"accelerates":[2],"LLM":[3],"inference":[4],"by":[5,20,32],"having":[6],"a":[7,21,79,95,99,134,146],"lightweight":[8],"draft":[9],"model":[10,129],"propose":[11,71],"speculative":[12,27,49,63,151],"windows":[13,110],"of":[14,47,121,125],"candidate":[15],"tokens":[16],"for":[17],"parallel":[18],"verification":[19],"larger":[22],"target":[23],"model.":[24],"In":[25],"practice,":[26],"efficiency":[28],"is":[29,65,145],"often":[30],"bottlenecked":[31],"hard-to-draft":[33],"positions,":[34],"where":[35],"an":[36],"early":[37],"mismatch":[38],"truncates":[39],"the":[40,45,48],"accepted":[41],"prefix":[42],"and":[43,68,103,123,131],"invalidates":[44],"rest":[46],"window.":[50],"Most":[51],"learning-based":[52],"drafters":[53],"are":[54],"still":[55],"optimized":[56],"with":[57,76,111],"token-level":[58,88],"supervised":[59],"objectives,":[60],"even":[61],"though":[62],"utility":[64],"inherently":[66],"window-level":[67,91,143],"prefix-sensitive.":[69],"We":[70],"PPOW":[72,93,116],"(Performance-Driven":[73],"Policy":[74],"Optimization":[75],"Adaptive":[77,104],"Windowing),":[78],"reinforcement":[80],"learning":[81],"framework":[82],"that":[83,141],"shifts":[84],"drafter":[85],"optimization":[86,144],"from":[87],"imitation":[89],"to":[90,149],"optimization.":[92],"combines":[94],"Cost-Aware":[96],"Speedup":[97],"Reward,":[98,102],"Distribution-Based":[100],"Proximity":[101],"Divergence-Aware":[105],"Windowing,":[106],"which":[107],"prioritizes":[108],"informative":[109],"high":[112],"confidence-weighted":[113],"draft-target":[114],"divergence.":[115],"achieves":[117],"average":[118],"acceptance":[119],"lengths":[120],"6.29-6.52":[122],"speedups":[124],"3.39-4.36$\\times$":[126],"across":[127],"multiple":[128],"families":[130],"benchmarks":[132],"under":[133],"unified":[135],"protocol.":[137],"These":[138],"results":[139],"show":[140],"performance-driven":[142],"practical":[147],"approach":[148],"improving":[150],"efficiency.":[153]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-16T00:00:00"}