{"id":"https://openalex.org/W7162459316","doi":"https://doi.org/10.48550/arxiv.2605.26108","title":"Reinforcing Few-step Generators via Reward-Tilted Distribution Matching","display_name":"Reinforcing Few-step Generators via Reward-Tilted Distribution Matching","publication_year":2026,"publication_date":"2026-05-25","ids":{"openalex":"https://openalex.org/W7162459316","doi":"https://doi.org/10.48550/arxiv.2605.26108"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.26108","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.26108","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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.26108","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137011906","display_name":"Yushi Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Yushi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072699828","display_name":"Xiangxin Zhou","orcid":"https://orcid.org/0000-0002-1526-0548"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Xiangxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137001077","display_name":"Ruoyu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ruoyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137016111","display_name":"Chi Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Chi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137082173","display_name":"Jun Zhang","orcid":"https://orcid.org/0000-0003-1706-1611"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137056543","display_name":"Tianyu Pang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pang, Tianyu","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":0.538100004196167,"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":0.538100004196167,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0934000015258789,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.09080000221729279,"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/matching","display_name":"Matching (statistics)","score":0.6195999979972839},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.5629000067710876},{"id":"https://openalex.org/keywords/maximization","display_name":"Maximization","score":0.4672999978065491},{"id":"https://openalex.org/keywords/divergence","display_name":"Divergence (linguistics)","score":0.4625000059604645},{"id":"https://openalex.org/keywords/distribution","display_name":"Distribution (mathematics)","score":0.4284999966621399},{"id":"https://openalex.org/keywords/estimator","display_name":"Estimator","score":0.4259999990463257},{"id":"https://openalex.org/keywords/generator","display_name":"Generator (circuit theory)","score":0.42419999837875366}],"concepts":[{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.6195999979972839},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6039999723434448},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.5629000067710876},{"id":"https://openalex.org/C2776330181","wikidata":"https://www.wikidata.org/wiki/Q18358244","display_name":"Maximization","level":2,"score":0.4672999978065491},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.4625000059604645},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.4284999966621399},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.4259999990463257},{"id":"https://openalex.org/C2780992000","wikidata":"https://www.wikidata.org/wiki/Q17016113","display_name":"Generator (circuit theory)","level":3,"score":0.42419999837875366},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.40959998965263367},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.37389999628067017},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3716000020503998},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.32989999651908875},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.32429999113082886},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.31380000710487366},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3059000074863434},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.29429998993873596},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2892000079154968},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C65778772","wikidata":"https://www.wikidata.org/wiki/Q12345341","display_name":"Asymptotic distribution","level":3,"score":0.2651999890804291},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.26429998874664307},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.25940001010894775}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.26108","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.26108","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.26108","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.26108","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":"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":{"Recent":[0],"advances":[1],"in":[2],"few-step":[3,40,181],"diffusion":[4],"distillation":[5,34],"have":[6],"enabled":[7],"efficient":[8],"image":[9],"generation,":[10],"yet":[11],"aligning":[12],"these":[13],"models":[14,187],"with":[15,35,89,137,174],"human":[16],"preferences":[17],"remains":[18],"challenging.":[19],"We":[20,43],"propose":[21],"Reward-Tilted":[22],"Distribution":[23,74],"Matching":[24,75],"Distillation":[25,76],"(RTDMD),":[26],"a":[27,51,58,63,90,123,129],"two-stage":[28],"framework":[29],"that":[30,45,127,162],"unifies":[31],"distribution":[32,54,59,81,103],"matching":[33,60,82],"reward-guided":[36],"reinforcement":[37],"learning":[38],"for":[39,116,132],"flow":[41],"generators.":[42],"show":[44],"minimizing":[46],"the":[47,68,85,95,100,108,117,133,142],"KL":[48],"divergence":[49],"to":[50,93,152],"reward-tilted":[52],"teacher":[53],"naturally":[55],"decomposes":[56],"into":[57],"term":[61],"and":[62,83,146,159,171,186],"reward":[64,118,139],"maximization":[65,119],"term.":[66],"In":[67,107],"first":[69],"stage,":[70,110],"we":[71,111,121],"introduce":[72,148],"Ambient-Consistent":[73],"(AC-DMD),":[77],"which":[78],"performs":[79],"subinterval-wise":[80],"augments":[84],"fake":[86,96],"score":[87,97],"objective":[88],"consistency":[91],"regularizer":[92],"help":[94],"model":[98],"track":[99],"shifting":[101],"generator":[102],"under":[104],"limited":[105],"updates.":[106],"second":[109],"jointly":[112],"optimize":[113],"both":[114],"terms:":[115],"term,":[120],"derive":[122],"hybrid":[124],"policy":[125],"gradient":[126],"combines":[128],"GRPO-style":[130],"estimator":[131],"stochastic":[134],"intermediate":[135],"transitions":[136],"direct":[138],"backpropagation":[140],"through":[141],"deterministic":[143],"final":[144],"step,":[145],"further":[147],"step-subset":[149],"GRPO":[150],"(SubGRPO)":[151],"reduce":[153],"variance.":[154],"Experiments":[155],"on":[156],"SD3,":[157],"SD3.5,":[158],"FLUX.2":[160],"demonstrate":[161],"RTDMD":[163],"establishes":[164],"new":[165],"state-of-the-art":[166],"results":[167],"across":[168],"preference,":[169],"aesthetic,":[170],"compositional":[172],"metrics":[173],"only":[175],"4":[176],"inference":[177],"steps,":[178],"outperforming":[179],"previous":[180],"text-to-image":[182],"generation":[183],"methods.":[184],"Code":[185],"are":[188],"available":[189],"at":[190],"https://github.com/Harahan/RTDMD.":[191]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-27T00:00:00"}