{"id":"https://openalex.org/W7162126235","doi":"https://doi.org/10.48550/arxiv.2605.21661","title":"Hierarchical Variational Policies for Reward-Guided Diffusion","display_name":"Hierarchical Variational Policies for Reward-Guided Diffusion","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7162126235","doi":"https://doi.org/10.48550/arxiv.2605.21661"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.21661","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21661","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.2605.21661","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103182644","display_name":"Kushagra Pandey","orcid":"https://orcid.org/0009-0009-2017-7878"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pandey, Kushagra","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092919140","display_name":"Farrin Marouf Sofian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sofian, Farrin Marouf","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136783940","display_name":"Jan Niklas Groeneveld","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Groeneveld, Jan Niklas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083537903","display_name":"Felix Draxler","orcid":"https://orcid.org/0000-0003-0978-1539"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Draxler, Felix","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136768198","display_name":"Stephan Mandt","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mandt, Stephan","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":0.47200000286102295,"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":0.47200000286102295,"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":0.19349999725818634,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.07270000129938126,"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":0.7023000121116638},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.5799000263214111},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.569100022315979},{"id":"https://openalex.org/keywords/amortized-analysis","display_name":"Amortized analysis","score":0.5501000285148621},{"id":"https://openalex.org/keywords/inverse","display_name":"Inverse","score":0.536899983882904},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.5012999773025513},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.45559999346733093}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7023000121116638},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.5799000263214111},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.569100022315979},{"id":"https://openalex.org/C142417499","wikidata":"https://www.wikidata.org/wiki/Q331716","display_name":"Amortized analysis","level":3,"score":0.5501000285148621},{"id":"https://openalex.org/C207467116","wikidata":"https://www.wikidata.org/wiki/Q4385666","display_name":"Inverse","level":2,"score":0.536899983882904},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5271999835968018},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.5012999773025513},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.491100013256073},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4609000086784363},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.45559999346733093},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4537999927997589},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.3896999955177307},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.36970001459121704},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3398999869823456},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.3280999958515167},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.32749998569488525},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C135252773","wikidata":"https://www.wikidata.org/wiki/Q1567213","display_name":"Inverse problem","level":2,"score":0.2892000079154968},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.26899999380111694},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.2513999938964844}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.21661","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21661","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.2605.21661","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21661","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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Adapting":[0],"pretrained":[1],"diffusion":[2,59],"models":[3],"to":[4,119,128],"downstream":[5],"objectives":[6],"such":[7],"as":[8,38],"inverse":[9,148],"problems":[10],"often":[11],"requires":[12],"expensive":[13],"test-time":[14,36,93,139],"guidance":[15],"or":[16,90],"optimization.":[17],"We":[18,123],"propose":[19],"a":[20,39,48,85,129],"principled":[21],"framework":[22],"for":[23],"generating":[24],"high-quality":[25],"reward-aligned":[26],"samples":[27],"at":[28],"substantially":[29],"reduced":[30],"inference":[31,117],"cost.":[32],"Our":[33],"approach":[34,127],"formulates":[35],"adaptation":[37],"hierarchical":[40],"variational":[41],"model,":[42],"where":[43],"control":[44],"is":[45],"amortized":[46,82,135],"into":[47],"lightweight":[49],"yet":[50],"expressive":[51],"stochastic":[52],"policy.":[53],"This":[54],"formulation":[55],"naturally":[56],"supports":[57],"few-step":[58],"sampling:":[60],"large":[61],"step":[62],"sizes":[63],"enable":[64],"fast":[65],"inference,":[66],"while":[67,96],"the":[68,120],"learned":[69],"policy":[70],"maintains":[71],"sample":[72],"quality":[73,111,144],"by":[74],"providing":[75],"structured":[76],"per-step":[77],"control.":[78],"The":[79],"resulting":[80],"fully":[81],"sampler":[83],"achieves":[84,108],"strong":[86],"quality--speed":[87],"tradeoff,":[88],"matching":[89],"exceeding":[91],"recent":[92],"scaling":[94],"baselines":[95],"requiring":[97],"significantly":[98],"less":[99],"compute.":[100],"For":[101],"example,":[102],"on":[103],"4x":[104],"super-resolution,":[105],"our":[106,126],"method":[107],"better":[109],"perceptual":[110,143],"with":[112,137],"more":[113],"than":[114],"5x":[115],"faster":[116],"compared":[118],"best-performing":[121],"baseline.":[122],"further":[124],"extend":[125],"semi-amortized":[130],"regime":[131],"that":[132],"combines":[133],"cheap":[134],"proposals":[136],"limited":[138],"optimization,":[140],"achieving":[141],"state-of-the-art":[142],"across":[145],"several":[146],"challenging":[147],"problems.":[149]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-23T00:00:00"}