{"id":"https://openalex.org/W7154712265","doi":"https://doi.org/10.48550/arxiv.2604.15311","title":"LeapAlign: Post-Training Flow Matching Models at Any Generation Step by Building Two-Step Trajectories","display_name":"LeapAlign: Post-Training Flow Matching Models at Any Generation Step by Building Two-Step Trajectories","publication_year":2026,"publication_date":"2026-04-16","ids":{"openalex":"https://openalex.org/W7154712265","doi":"https://doi.org/10.48550/arxiv.2604.15311"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.15311","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.15311","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.2604.15311","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133851037","display_name":"Zhanhao Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Zhanhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017205420","display_name":"Tao Yang","orcid":"https://orcid.org/0000-0002-6451-7127"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133878865","display_name":"Jie Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Jie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101303327","display_name":"Chengjian Feng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng, Chengjian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133839796","display_name":"Liang Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Liang","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":0.4560999870300293,"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":0.4560999870300293,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.38100001215934753,"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":0.029600000008940697,"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/matching","display_name":"Matching (statistics)","score":0.6578999757766724},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.6010000109672546},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5806999802589417},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.48190000653266907},{"id":"https://openalex.org/keywords/flow","display_name":"Flow (mathematics)","score":0.45559999346733093},{"id":"https://openalex.org/keywords/differentiable-function","display_name":"Differentiable function","score":0.43560001254081726},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.4350999891757965},{"id":"https://openalex.org/keywords/image","display_name":"Image (mathematics)","score":0.36719998717308044}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7497000098228455},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.6578999757766724},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.6010000109672546},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5806999802589417},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5267000198364258},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.48190000653266907},{"id":"https://openalex.org/C38349280","wikidata":"https://www.wikidata.org/wiki/Q1434290","display_name":"Flow (mathematics)","level":2,"score":0.45559999346733093},{"id":"https://openalex.org/C202615002","wikidata":"https://www.wikidata.org/wiki/Q783507","display_name":"Differentiable function","level":2,"score":0.43560001254081726},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.4350999891757965},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3952000141143799},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.36719998717308044},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.3614000082015991},{"id":"https://openalex.org/C34862557","wikidata":"https://www.wikidata.org/wiki/Q178985","display_name":"Ode","level":2,"score":0.3562999963760376},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.35190001130104065},{"id":"https://openalex.org/C167879884","wikidata":"https://www.wikidata.org/wiki/Q727568","display_name":"Balanced flow","level":2,"score":0.3479999899864197},{"id":"https://openalex.org/C155542232","wikidata":"https://www.wikidata.org/wiki/Q736111","display_name":"Optical flow","level":3,"score":0.3215000033378601},{"id":"https://openalex.org/C115680565","wikidata":"https://www.wikidata.org/wiki/Q5977448","display_name":"Gradient method","level":2,"score":0.3183000087738037},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.302700012922287},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.29739999771118164},{"id":"https://openalex.org/C155032097","wikidata":"https://www.wikidata.org/wiki/Q798503","display_name":"Backpropagation","level":3,"score":0.29330000281333923},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.28859999775886536},{"id":"https://openalex.org/C55020928","wikidata":"https://www.wikidata.org/wiki/Q3813865","display_name":"Image quality","level":3,"score":0.2635999917984009},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.2597000002861023}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.15311","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.15311","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.2604.15311","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.15311","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":{"This":[0],"paper":[1],"focuses":[2],"on":[3],"the":[4,24,58,62,93,121,127,159,170,190],"alignment":[5],"of":[6,28,61,126,172,179],"flow":[7,29],"matching":[8],"models":[9],"with":[10,158,175],"human":[11],"preferences.":[12],"A":[13],"promising":[14],"way":[15],"is":[16],"fine-tuning":[17,73,189],"by":[18,100],"directly":[19],"backpropagating":[20,32],"reward":[21,85],"gradients":[22],"through":[23,33],"differentiable":[25],"generation":[26,51,88,139,161],"process":[27],"matching.":[30],"However,":[31],"long":[34,94,160],"trajectories":[35],"results":[36],"in":[37,115,185],"prohibitive":[38],"memory":[39],"costs":[40],"and":[41,79,111,123,133,198,208],"gradient":[42,82,166,173],"explosion.":[43],"Therefore,":[44],"direct-gradient":[45,199],"methods":[46,200],"struggle":[47],"to":[48,86,131,152],"update":[49],"early":[50,87],"steps,":[52],"which":[53],"are":[54,155],"crucial":[55],"for":[56],"determining":[57],"global":[59],"structure":[60],"final":[63],"image.":[64],"To":[65,141,163],"address":[66],"this":[67],"issue,":[68],"we":[69,91,147,168],"introduce":[70],"LeapAlign,":[71],"a":[72,116],"method":[74],"that":[75,154],"reduces":[76],"computational":[77],"cost":[78],"enables":[80],"direct":[81],"propagation":[83],"from":[84],"steps.":[89],"Specifically,":[90],"shorten":[92],"trajectory":[95],"into":[96],"only":[97],"two":[98,102],"steps":[99,110],"designing":[101],"consecutive":[103],"leaps,":[104,128],"each":[105],"skipping":[106],"multiple":[107],"ODE":[108],"sampling":[109],"predicting":[112],"future":[113],"latents":[114],"single":[117],"step.":[118,140],"By":[119],"randomizing":[120],"start":[122],"end":[124],"timesteps":[125],"LeapAlign":[129,193],"leads":[130],"efficient":[132],"stable":[134],"model":[135],"updates":[136],"at":[137],"any":[138],"better":[142],"use":[143],"such":[144],"shortened":[145],"trajectories,":[146],"assign":[148],"higher":[149],"training":[150],"weights":[151,171],"those":[153],"more":[156],"consistent":[157],"path.":[162],"further":[164],"enhance":[165],"stability,":[167],"reduce":[169],"terms":[174],"large":[176],"magnitude,":[177],"instead":[178],"completely":[180],"removing":[181],"them":[182],"as":[183],"done":[184],"previous":[186],"works.":[187],"When":[188],"Flux":[191],"model,":[192],"consistently":[194],"outperforms":[195],"state-of-the-art":[196],"GRPO-based":[197],"across":[201],"various":[202],"metrics,":[203],"achieving":[204],"superior":[205],"image":[206],"quality":[207],"image-text":[209],"alignment.":[210]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-18T00:00:00"}