{"id":"https://openalex.org/W7162318576","doi":"https://doi.org/10.48550/arxiv.2605.23522","title":"Precise: SDE-Consistent Stochastic Sampling for RL Post-Training of Flow-Matching Models","display_name":"Precise: SDE-Consistent Stochastic Sampling for RL Post-Training of Flow-Matching Models","publication_year":2026,"publication_date":"2026-05-22","ids":{"openalex":"https://openalex.org/W7162318576","doi":"https://doi.org/10.48550/arxiv.2605.23522"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.23522","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23522","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":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.2605.23522","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136924964","display_name":"Jade Zou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zou, Jade","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136960433","display_name":"Tao Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136897709","display_name":"Weijie Kong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kong, Weijie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136971771","display_name":"Junzhe Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Junzhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136913954","display_name":"Yue Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yue","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136949933","display_name":"Qi Tian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tian, Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083206979","display_name":"Jiangfeng Xiong","orcid":"https://orcid.org/0000-0001-6840-8262"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong, Jiangfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136904194","display_name":"Jianwei Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jianwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136934767","display_name":"Liefeng Bo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo, Liefeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136960768","display_name":"Zhao Zhong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Zhao","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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.3698999881744385,"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"}},"topics":[{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.3698999881744385,"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/T11206","display_name":"Model Reduction and Neural Networks","score":0.03889999911189079,"subfield":{"id":"https://openalex.org/subfields/3109","display_name":"Statistical and Nonlinear Physics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11309","display_name":"Music and Audio Processing","score":0.027799999341368675,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/discretization","display_name":"Discretization","score":0.6585999727249146},{"id":"https://openalex.org/keywords/stochastic-differential-equation","display_name":"Stochastic differential equation","score":0.6568999886512756},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.5637999773025513},{"id":"https://openalex.org/keywords/convergence","display_name":"Convergence (economics)","score":0.5206999778747559},{"id":"https://openalex.org/keywords/stability","display_name":"Stability (learning theory)","score":0.5156000256538391},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.46790000796318054},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.46160000562667847},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.44620001316070557},{"id":"https://openalex.org/keywords/heuristic","display_name":"Heuristic","score":0.4278999865055084}],"concepts":[{"id":"https://openalex.org/C73000952","wikidata":"https://www.wikidata.org/wiki/Q17007827","display_name":"Discretization","level":2,"score":0.6585999727249146},{"id":"https://openalex.org/C51955184","wikidata":"https://www.wikidata.org/wiki/Q1545585","display_name":"Stochastic differential equation","level":2,"score":0.6568999886512756},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5640000104904175},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.5637999773025513},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.5396999716758728},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.5206999778747559},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.5156000256538391},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.46790000796318054},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.46160000562667847},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.44620001316070557},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.4278999865055084},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.40119999647140503},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.38580000400543213},{"id":"https://openalex.org/C38349280","wikidata":"https://www.wikidata.org/wiki/Q1434290","display_name":"Flow (mathematics)","level":2,"score":0.3813999891281128},{"id":"https://openalex.org/C137631369","wikidata":"https://www.wikidata.org/wiki/Q7617831","display_name":"Stochastic programming","level":2,"score":0.37929999828338623},{"id":"https://openalex.org/C68387754","wikidata":"https://www.wikidata.org/wiki/Q7271585","display_name":"Schedule","level":2,"score":0.3790000081062317},{"id":"https://openalex.org/C194387892","wikidata":"https://www.wikidata.org/wiki/Q1747770","display_name":"Stochastic optimization","level":2,"score":0.3677999973297119},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.36079999804496765},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.34869998693466187},{"id":"https://openalex.org/C127491075","wikidata":"https://www.wikidata.org/wiki/Q7617825","display_name":"Stochastic modelling","level":2,"score":0.3240000009536743},{"id":"https://openalex.org/C74750220","wikidata":"https://www.wikidata.org/wiki/Q2662197","display_name":"Differential evolution","level":2,"score":0.32199999690055847},{"id":"https://openalex.org/C51544822","wikidata":"https://www.wikidata.org/wiki/Q465274","display_name":"Ordinary differential equation","level":3,"score":0.31209999322891235},{"id":"https://openalex.org/C8272713","wikidata":"https://www.wikidata.org/wiki/Q176737","display_name":"Stochastic process","level":2,"score":0.30790001153945923},{"id":"https://openalex.org/C55479107","wikidata":"https://www.wikidata.org/wiki/Q97663916","display_name":"Stochastic approximation","level":3,"score":0.2854999899864197},{"id":"https://openalex.org/C47446073","wikidata":"https://www.wikidata.org/wiki/Q5165890","display_name":"Control theory (sociology)","level":3,"score":0.2793999910354614},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.2705000042915344},{"id":"https://openalex.org/C78045399","wikidata":"https://www.wikidata.org/wiki/Q11214","display_name":"Differential equation","level":2,"score":0.26649999618530273},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2508000135421753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.23522","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23522","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":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.2605.23522","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23522","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":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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1],"(RL)":[2],"has":[3],"become":[4],"an":[5,130],"effective":[6,192],"way":[7],"to":[8,27,138,147,175,228,253],"improve":[9],"prompt":[10],"alignment":[11,241],"and":[12,17,62,71,98,124,128,231],"perceptual":[13],"quality":[14],"in":[15,110,126,218],"diffusion":[16],"flow-matching":[18,156],"generators.":[19],"A":[20],"critical":[21],"step":[22,107],"for":[23],"applying":[24],"online":[25],"RL":[26],"flow":[28],"matching":[29],"is":[30,65],"turning":[31],"the":[32,43,59,69,77,84,92,100,105,114,119,135,139,155,176,199,209,214,255],"deterministic":[33],"sampling":[34],"trajectory":[35,201],"into":[36,87],"a":[37,50,144,186,204],"stochastic":[38,56,96,188],"policy,":[39,70],"typically":[40],"by":[41,159,165],"replacing":[42],"reverse-time":[44],"Ordinary":[45],"Differential":[46,52],"Equation":[47,53],"(ODE)":[48],"with":[49,194],"Stochastic":[51],"(SDE).":[54],"The":[55],"sampler,":[57],"controlling":[58],"exploration":[60,123,193],"behavior":[61],"denoising":[63,127,200],"dynamics,":[64],"thus":[66],"part":[67],"of":[68,95,259],"its":[72],"design":[73,86],"can":[74,152],"significantly":[75,229],"affect":[76],"reward":[78,234],"optimization":[79,235],"performance.":[80],"We":[81],"break":[82],"down":[83],"sampler":[85,189],"two":[88],"interdependent":[89],"components:":[90],"choosing":[91],"right":[93],"amount":[94],"exploration,":[97],"discretizing":[99],"resulting":[101],"SDE":[102,131],"faithfully":[103],"at":[104],"small":[106],"counts":[108],"used":[109],"RL.":[111],"To":[112,179],"address":[113,180],"first":[115],"component,":[116],"we":[117,142,183],"analyze":[118],"inherent":[120],"tension":[121],"between":[122],"stability":[125],"derive":[129],"schedule":[132],"that":[133,149,170,190,207,224],"balances":[134,191],"two.":[136],"Turning":[137],"discretization":[140,162],"challenge,":[141],"use":[143],"toy":[145],"example":[146],"show":[148],"existing":[150],"samplers":[151],"deviate":[153],"from":[154],"process,":[157],"either":[158],"introducing":[160],"excessive":[161],"noise":[163,216],"or":[164],"relying":[166],"on":[167],"heuristic":[168],"rules":[169],"do":[171],"not":[172],"guarantee":[173],"convergence":[174],"data":[177],"distribution.":[178],"these":[181],"issues,":[182],"propose":[184],"Precise,":[185],"new":[187],"stability.":[195],"Crucially,":[196],"Precise":[197],"keeps":[198],"SDE-consistent":[202],"through":[203],"novel":[205],"approximation":[206],"freezes":[208],"clean-latent":[210],"posterior":[211],"mean,":[212],"resolving":[213],"excess":[215],"issue":[217],"standard":[219],"samplers.":[220,261],"Extensive":[221],"experiments":[222],"demonstrate":[223],"this":[225],"formulation":[226],"leads":[227],"faster":[230],"more":[232],"stable":[233],"via":[236],"reinforcement":[237],"learning,":[238],"achieving":[239],"state-of-the-art":[240],"scores":[242],"(e.g.,":[243],"PickScore,":[244],"HPSv2.1)":[245],"while":[246],"requiring":[247],"13.1-53.2%":[248],"less":[249],"wall-clock":[250],"training":[251],"time":[252],"match":[254],"best":[256],"in-domain":[257],"performance":[258],"prior":[260]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-26T00:00:00"}