{"id":"https://openalex.org/W7154577890","doi":"https://doi.org/10.48550/arxiv.2604.13197","title":"Unleashing Implicit Rewards: Prefix-Value Learning for Distribution-Level Optimization","display_name":"Unleashing Implicit Rewards: Prefix-Value Learning for Distribution-Level Optimization","publication_year":2026,"publication_date":"2026-04-14","ids":{"openalex":"https://openalex.org/W7154577890","doi":"https://doi.org/10.48550/arxiv.2604.13197"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.13197","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13197","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.2604.13197","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5027482212","display_name":"Shang Gao","orcid":"https://orcid.org/0000-0003-3320-8505"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Shiping","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133816921","display_name":"Hongzhan Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Hongzhan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133753065","display_name":"Xiaojun Quan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Quan, Xiaojun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133778534","display_name":"Qifan Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Qifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133737957","display_name":"Lifu Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Lifu","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/T10703","display_name":"Business Process Modeling and Analysis","score":0.19359999895095825,"subfield":{"id":"https://openalex.org/subfields/1404","display_name":"Management Information Systems"},"field":{"id":"https://openalex.org/fields/14","display_name":"Business, Management and Accounting"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10703","display_name":"Business Process Modeling and Analysis","score":0.19359999895095825,"subfield":{"id":"https://openalex.org/subfields/1404","display_name":"Management Information Systems"},"field":{"id":"https://openalex.org/fields/14","display_name":"Business, Management and Accounting"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.1273999959230423,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.08560000360012054,"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/inference","display_name":"Inference","score":0.6621999740600586},{"id":"https://openalex.org/keywords/counterfactual-thinking","display_name":"Counterfactual thinking","score":0.6485999822616577},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.585099995136261},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5794000029563904},{"id":"https://openalex.org/keywords/function","display_name":"Function (biology)","score":0.5498999953269958},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.47040000557899475},{"id":"https://openalex.org/keywords/outcome","display_name":"Outcome (game theory)","score":0.4345000088214874},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.37560001015663147}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7565000057220459},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6621999740600586},{"id":"https://openalex.org/C108650721","wikidata":"https://www.wikidata.org/wiki/Q1783253","display_name":"Counterfactual thinking","level":2,"score":0.6485999822616577},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.585099995136261},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5794000029563904},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5564000010490417},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.5498999953269958},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5175999999046326},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.47040000557899475},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.4345000088214874},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.37560001015663147},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.33899998664855957},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.31060001254081726},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3003000020980835},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.299699991941452},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.2994000017642975},{"id":"https://openalex.org/C2776291640","wikidata":"https://www.wikidata.org/wiki/Q2912517","display_name":"Value (mathematics)","level":2,"score":0.296099990606308},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.28780001401901245},{"id":"https://openalex.org/C63002673","wikidata":"https://www.wikidata.org/wiki/Q2260590","display_name":"Scoring rule","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C159032336","wikidata":"https://www.wikidata.org/wiki/Q2488768","display_name":"Non-monotonic logic","level":2,"score":0.2587999999523163}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.13197","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13197","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.2604.13197","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13197","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":{"Process":[0],"reward":[1],"models":[2],"(PRMs)":[3],"provide":[4],"fine-grained":[5],"supervision":[6],"for":[7,64,97],"reasoning,":[8],"but":[9,174],"reliable":[10],"PRMs":[11,31],"often":[12],"require":[13],"step":[14],"annotations":[15],"or":[16,61],"heavy":[17],"verification":[18],"pipelines,":[19],"making":[20],"them":[21],"costly":[22],"to":[23,148],"scale":[24],"and":[25,152],"refresh":[26],"during":[27,53,135],"online":[28],"RL.":[29],"Implicit":[30,84],"reduce":[32],"this":[33],"cost":[34],"by":[35],"training":[36,118],"log-likelihood-ratio":[37],"rewards":[38],"from":[39,100],"trajectory-level":[40],"outcome":[41,101],"labels.":[42,102],"However,":[43],"the":[44,92,117],"log-ratio":[45],"is":[46,188],"constrained":[47],"only":[48],"as":[49,108],"a":[50],"sequence-level":[51],"aggregate":[52],"training,":[54],"while":[55],"inference":[56],"decomposes":[57],"it":[58],"into":[59],"token-":[60],"step-level":[62],"scores":[63],"partial":[65],"prefixes.":[66],"This":[67],"train-inference":[68],"mismatch":[69],"leaves":[70],"local":[71],"credits":[72],"weakly":[73],"identified,":[74],"so":[75],"distribution-wide":[76],"scoring":[77],"can":[78],"amplify":[79],"misleading":[80],"advantages.":[81],"We":[82],"propose":[83],"Prefix-Value":[85],"Reward":[86],"Model":[87],"(IPVRM),":[88],"which":[89,144],"directly":[90],"learns":[91],"probability":[93],"of":[94,185],"eventual":[95],"correctness":[96],"each":[98],"prefix":[99,114,133],"Step":[103],"signals":[104],"are":[105],"then":[106],"obtained":[107],"temporal-difference":[109],"(TD)":[110],"differences":[111],"between":[112],"consecutive":[113],"values,":[115],"aligning":[116],"target":[119],"with":[120,170,181],"inference-time":[121],"use.":[122],"IPVRM":[123],"markedly":[124],"improves":[125,176],"step-verification":[126],"F1":[127],"on":[128],"ProcessBench.":[129],"To":[130],"exploit":[131],"these":[132],"values":[134],"policy":[136],"optimization,":[137],"we":[138],"further":[139],"introduce":[140],"Distribution-Level":[141],"RL":[142],"(DistRL),":[143],"applies":[145],"TD":[146],"advantages":[147],"both":[149],"sampled":[150],"tokens":[151],"high-probability":[153],"candidate":[154],"tokens,":[155],"providing":[156],"dense":[157],"counterfactual":[158],"updates":[159],"without":[160],"additional":[161],"rollouts.":[162],"Experiments":[163],"show":[164],"that":[165],"DistRL":[166],"brings":[167],"limited":[168],"gains":[169],"unreliable":[171],"implicit":[172],"rewards,":[173],"consistently":[175],"downstream":[177],"reasoning":[178],"when":[179],"paired":[180],"IPVRM.":[182],"The":[183],"implementation":[184],"our":[186],"method":[187],"available":[189],"at":[190],"https://github.com/gaoshiping/IPVRM":[191],".":[192]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-17T00:00:00"}