{"id":"https://openalex.org/W7161586305","doi":"https://doi.org/10.48550/arxiv.2605.15855","title":"Do Less, Achieve More: Do We Need Every-Step Optimization for RL Fine-tuning of Diffusion Models?","display_name":"Do Less, Achieve More: Do We Need Every-Step Optimization for RL Fine-tuning of Diffusion Models?","publication_year":2026,"publication_date":"2026-05-15","ids":{"openalex":"https://openalex.org/W7161586305","doi":"https://doi.org/10.48550/arxiv.2605.15855"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.15855","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15855","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","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.2605.15855","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5089763177","display_name":"Renye Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan, Renye","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041019639","display_name":"Jikang Cheng","orcid":"https://orcid.org/0000-0001-6549-6148"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Jikang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136420036","display_name":"Shikun Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Shikun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136443141","display_name":"Yi Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136417605","display_name":"You Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, You","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136436770","display_name":"Wei Peng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peng, Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136363420","display_name":"Zongwei Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zongwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136396742","display_name":"Ling Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Ling","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136449707","display_name":"Junliang Xing","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xing, Junliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136412184","display_name":"Yimao Cai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cai, Yimao","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":0.590499997138977,"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":0.590499997138977,"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/T11304","display_name":"Advanced Neuroimaging Techniques and Applications","score":0.18080000579357147,"subfield":{"id":"https://openalex.org/subfields/2741","display_name":"Radiology, Nuclear Medicine and Imaging"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.033399999141693115,"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/overfitting","display_name":"Overfitting","score":0.6956999897956848},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.551800012588501},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.5174999833106995},{"id":"https://openalex.org/keywords/noise-reduction","display_name":"Noise reduction","score":0.49140000343322754},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.478300005197525},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.4153999984264374},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.37529999017715454},{"id":"https://openalex.org/keywords/variance","display_name":"Variance (accounting)","score":0.37450000643730164}],"concepts":[{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.6956999897956848},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6288999915122986},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.551800012588501},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.5496000051498413},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.5174999833106995},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.49140000343322754},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.478300005197525},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4268999993801117},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.4153999984264374},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3887999951839447},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.37529999017715454},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.37450000643730164},{"id":"https://openalex.org/C62644790","wikidata":"https://www.wikidata.org/wiki/Q3454689","display_name":"Variance reduction","level":3,"score":0.37439998984336853},{"id":"https://openalex.org/C2983327147","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Image denoising","level":3,"score":0.32710000872612},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.32440000772476196},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.3208000063896179},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.3163999915122986},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3043000102043152},{"id":"https://openalex.org/C94966114","wikidata":"https://www.wikidata.org/wiki/Q29256","display_name":"Black box","level":2,"score":0.3034000098705292},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.29159998893737793},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.2874000072479248},{"id":"https://openalex.org/C112633086","wikidata":"https://www.wikidata.org/wiki/Q381287","display_name":"White noise","level":2,"score":0.26910001039505005},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.2689000070095062},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.2680000066757202},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.26030001044273376},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.2574999928474426}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.15855","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15855","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.15855","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15855","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","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":{"Despite":[0],"strong":[1],"image-generation":[2],"performance,":[3],"diffusion":[4],"models'":[5],"reconstruction":[6],"objectives":[7],"limit":[8],"alignment":[9,16],"with":[10,189],"human":[11],"preferences.":[12],"RL":[13,24,50,73,137],"enables":[14],"such":[15],"through":[17],"explicit":[18],"rewards.":[19],"However,":[20],"most":[21],"studies":[22],"apply":[23],"to":[25,78,103],"the":[26,47,58,68,93,132,140,153,184],"full":[27],"denoising":[28,55,154],"trajectory,":[29],"making":[30],"it":[31,163],"computationally":[32],"costly":[33],"and":[34,65,81,88,99,143,148,156],"weakening":[35],"preference":[36],"alignment,":[37],"i.e.,":[38],"doing":[39],"more":[40],"but":[41],"achieving":[42],"less.":[43],"We":[44,179],"observe":[45],"that":[46,120],"impact":[48],"of":[49,186],"fine-tuning":[51],"varies":[52],"significantly":[53],"across":[54],"stages.":[56],"In":[57],"early":[59],"stage,":[60,95],"image":[61],"structures":[62],"are":[63],"unstable":[64],"distant":[66],"from":[67],"final":[69],"reward":[70,96,108,157],"signal.":[71],"Applying":[72],"at":[74],"this":[75],"stage":[76],"leads":[77],"delayed":[79],"rewards":[80],"action-reward":[82],"mismatching,":[83],"resulting":[84],"in":[85,92,171],"high":[86],"variance":[87],"inefficient":[89],"updates.":[90],"Conversely,":[91],"later":[94],"gains":[97,158],"saturate,":[98],"continued":[100],"training":[101,151],"tends":[102],"overfit":[104],"local":[105],"details,":[106],"intensifying":[107],"hacking.":[109],"To":[110],"tackle":[111],"these":[112],"challenges,":[113],"we":[114],"propose":[115],"AdaScope,":[116],"an":[117],"RL-enhanced":[118],"plug-in":[119],"improves":[121,193],"generation":[122],"quality":[123],"while":[124,197],"reducing":[125],"computational":[126,172,199],"cost.":[127],"Specifically,":[128],"AdaScope":[129,192],"adaptively":[130],"identifies":[131],"optimal":[133],"intervention":[134],"timing":[135],"for":[136,183],"by":[138,195,201],"perceiving":[139],"structural":[141],"evolution":[142],"semantic":[144],"consistency":[145],"during":[146],"denoising,":[147],"dynamically":[149],"terminates":[150],"once":[152],"converges":[155],"saturate.":[159],"As":[160],"a":[161,165,169,175],"result,":[162],"achieves":[164],"rare":[166],"'dual":[167],"benefit':":[168],"reduction":[170],"costs":[173],"alongside":[174],"significant":[176],"performance":[177,194],"improvement.":[178],"offer":[180],"theoretical":[181],"grounds":[182],"design":[185],"AdaScope.":[187],"Compared":[188],"state-of-the-art":[190],"methods,":[191],"66%":[196],"cutting":[198],"cost":[200],"59%.":[202]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-19T00:00:00"}