{"id":"https://openalex.org/W7160558431","doi":"https://doi.org/10.48550/arxiv.2605.05204","title":"D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models","display_name":"D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models","publication_year":2026,"publication_date":"2026-05-06","ids":{"openalex":"https://openalex.org/W7160558431","doi":"https://doi.org/10.48550/arxiv.2605.05204"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.05204","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05204","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.2605.05204","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135583274","display_name":"Dengyang Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Dengyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135624714","display_name":"Xin Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Xin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135588435","display_name":"Dongyang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Dongyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135555988","display_name":"Zanyi Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zanyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135617557","display_name":"Mingzhe Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Mingzhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011876517","display_name":"Ruoyi Du","orcid":"https://orcid.org/0000-0001-8372-5637"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Ruoyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135598482","display_name":"Xiangpeng Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Xiangpeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127822905","display_name":"Qilong Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Qilong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135613123","display_name":"Zhen Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zhen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135606972","display_name":"Peng Gao","orcid":"https://orcid.org/0009-0000-1138-5936"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135580929","display_name":"Harry Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Harry","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5128001271","display_name":"Steven Hoi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hoi, Steven","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":0.8892999887466431,"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":0.8892999887466431,"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":0.021900000050663948,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.01600000075995922,"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/feature","display_name":"Feature (linguistics)","score":0.6486999988555908},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.5575000047683716},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5566999912261963},{"id":"https://openalex.org/keywords/diffusion","display_name":"Diffusion","score":0.550599992275238},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.5501999855041504},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.43459999561309814}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6639000177383423},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.6486999988555908},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5921000242233276},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.5575000047683716},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5566999912261963},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.550599992275238},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.5501999855041504},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.43459999561309814},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34619998931884766},{"id":"https://openalex.org/C136389625","wikidata":"https://www.wikidata.org/wiki/Q334384","display_name":"Supervised learning","level":3,"score":0.33309999108314514},{"id":"https://openalex.org/C83248878","wikidata":"https://www.wikidata.org/wiki/Q344000","display_name":"Active appearance model","level":3,"score":0.3059999942779541},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.2921000123023987},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2703999876976013},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.2703999876976013},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2680000066757202},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.26260000467300415}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.05204","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05204","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.2605.05204","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05204","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":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.47811979055404663}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"landscape":[1],"of":[2,141],"high-performance":[3],"image":[4],"generation":[5],"models":[6,26,63],"is":[7,125,135],"currently":[8],"shifting":[9],"from":[10],"the":[11,16,38,75,80,84,97,109,114,117,123,129,133,138,143,147,152,157,164,175,185],"inefficient":[12],"multi-step":[13],"ones":[14],"to":[15,95,177],"efficient":[17],"few-step":[18,47,187],"counterparts":[19],"(e.g,":[20],"Z-Image-Turbo":[21],"and":[22,116,146,168],"FLUX.2-klein).":[23],"However,":[24],"these":[25],"present":[27],"significant":[28],"challenges":[29],"for":[30,60],"direct":[31],"continuous":[32],"supervised":[33,69],"fine-tuning.":[34,70],"For":[35],"example,":[36],"applying":[37],"commonly":[39],"used":[40],"fine-tuning":[41],"technique":[42],"would":[43],"compromise":[44],"their":[45],"inherent":[46],"inference":[48],"capability.":[49],"To":[50],"address":[51],"this,":[52],"we":[53,107],"propose":[54],"D-OPSD,":[55],"a":[56],"novel":[57],"training":[58,98],"paradigm":[59],"step-distilled":[61],"diffusion":[62,77],"that":[64,74],"enables":[65,93,174],"on-policy":[66,101],"learning":[67],"during":[68,105],"We":[71],"first":[72],"find":[73],"modern":[76],"models,":[78],"where":[79,122],"LLM/VLM":[81],"serves":[82],"as":[83,99,112],"encoder,":[85],"can":[86],"inherit":[87],"its":[88,170],"encoder's":[89],"in-context":[90],"capabilities.":[91],"This":[92],"us":[94],"formulate":[96],"an":[100],"self-distillation":[102],"process.":[103],"Specifically,":[104],"training,":[106],"make":[108],"model":[110,176],"act":[111],"both":[113,142],"teacher":[115,134],"student":[118,124],"with":[119],"different":[120],"contexts,":[121],"conditioned":[126,136],"only":[127],"on":[128,137,163],"text":[130,144],"feature,":[131],"while":[132],"multimodal":[139],"feature":[140],"prompt":[145],"target":[148],"image.":[149],"Training":[150],"minimizes":[151],"two":[153],"predicted":[154],"distributions":[155],"over":[156],"student's":[158],"own":[159,166,171],"roll-outs.":[160],"By":[161],"optimizing":[162],"model's":[165],"trajectory":[167],"under":[169],"supervision,":[172],"D-OPSD":[173],"learn":[178],"new":[179],"concepts,":[180],"styles,":[181],"etc.,":[182],"without":[183],"sacrificing":[184],"original":[186],"capacity.":[188]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-08T00:00:00"}