{"id":"https://openalex.org/W7152643856","doi":"https://doi.org/10.48550/arxiv.2604.07026","title":"Not all tokens contribute equally to diffusion learning","display_name":"Not all tokens contribute equally to diffusion learning","publication_year":2026,"publication_date":"2026-04-08","ids":{"openalex":"https://openalex.org/W7152643856","doi":"https://doi.org/10.48550/arxiv.2604.07026"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.07026","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.07026","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.2604.07026","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100419048","display_name":"Guoqing Zhang","orcid":"https://orcid.org/0000-0001-5796-8547"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Guoqing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133253065","display_name":"Lu Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Lu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071889746","display_name":"Wanru Xu","orcid":"https://orcid.org/0000-0003-2206-5051"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Wanru","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133301645","display_name":"Linna Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Linna","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133256296","display_name":"Sen Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Sen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133241664","display_name":"Fangfang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Fangfang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133311350","display_name":"Yigang Cen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cen, Yigang","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":0.4657000005245209,"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":0.4657000005245209,"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":0.27230000495910645,"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":0.10700000077486038,"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/security-token","display_name":"Security token","score":0.7498000264167786},{"id":"https://openalex.org/keywords/overfitting","display_name":"Overfitting","score":0.5787000060081482},{"id":"https://openalex.org/keywords/debiasing","display_name":"Debiasing","score":0.5329999923706055},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.4961000084877014},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.4717000126838684},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.4016999900341034},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.391400009393692},{"id":"https://openalex.org/keywords/fidelity","display_name":"Fidelity","score":0.38850000500679016}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7900000214576721},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7498000264167786},{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.5787000060081482},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5666999816894531},{"id":"https://openalex.org/C2779458634","wikidata":"https://www.wikidata.org/wiki/Q24963715","display_name":"Debiasing","level":2,"score":0.5329999923706055},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.4961000084877014},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.4717000126838684},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.44530001282691956},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.4016999900341034},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.391400009393692},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.38850000500679016},{"id":"https://openalex.org/C2778493491","wikidata":"https://www.wikidata.org/wiki/Q7449072","display_name":"Semantic matching","level":3,"score":0.3752000033855438},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3400000035762787},{"id":"https://openalex.org/C2778180026","wikidata":"https://www.wikidata.org/wiki/Q18378163","display_name":"Semantic heterogeneity","level":4,"score":0.3249000012874603},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.31130000948905945},{"id":"https://openalex.org/C2781122975","wikidata":"https://www.wikidata.org/wiki/Q16928266","display_name":"Semantic feature","level":2,"score":0.2897000014781952},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.28839999437332153},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.2651999890804291},{"id":"https://openalex.org/C67277372","wikidata":"https://www.wikidata.org/wiki/Q7449085","display_name":"Semantic role labeling","level":3,"score":0.26409998536109924},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.26159998774528503},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2590999901294708},{"id":"https://openalex.org/C2775955345","wikidata":"https://www.wikidata.org/wiki/Q7449071","display_name":"Semantic mapping","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.2549000084400177},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.2502000033855438}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.07026","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.07026","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.2604.07026","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.07026","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":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.41029930114746094}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"With":[0],"the":[1,50,95,112,125,143,146,196,201,204],"rapid":[2],"development":[3],"of":[4,97,145,203],"conditional":[5,138],"diffusion":[6,92],"models,":[7],"significant":[8,231],"progress":[9],"has":[10],"been":[11],"made":[12],"in":[13,54,60,91],"text-to-video":[14],"generation.":[15,186],"However,":[16],"we":[17,76,104,157],"observe":[18],"that":[19,87,221],"these":[20,74],"models":[21,93],"often":[22],"neglect":[23],"semantically":[24,63,177],"important":[25,64,178],"tokens":[26,65,119,151,179,193],"during":[27,185],"inference,":[28],"leading":[29],"to":[30,42,127,150,169,180],"biased":[31],"or":[32],"incomplete":[33],"generations":[34],"under":[35],"classifier-free":[36],"guidance.":[37],"We":[38],"attribute":[39],"this":[40],"issue":[41],"two":[43],"key":[44],"factors:":[45],"distributional":[46,98,207],"bias":[47],"caused":[48],"by":[49,68,115,210],"long-tailed":[51],"token":[52,170],"frequency":[53],"training":[55,113],"data,":[56],"and":[57,80,100,133,172,206,227],"spatial":[58,101,183,205],"misalignment":[59],"cross-attention":[61,166],"where":[62],"are":[66],"overshadowed":[67],"less":[69],"informative":[70],"ones.":[71],"To":[72],"address":[73],"issues,":[75],"propose":[77,158],"Distribution-Aware":[78],"Rectification":[79],"Spatial":[81,159],"Ensemble":[82],"(DARE),":[83],"a":[84,135],"unified":[85],"framework":[86],"improves":[88,224],"semantic":[89,122,131,154,228],"guidance":[90,184,208],"from":[94,194],"perspectives":[96],"debiasing":[99],"consistency.":[102],"First,":[103],"introduce":[105],"Distribution-Rectified":[106],"Classifier-Free":[107],"Guidance":[108],"(DR-CFG),":[109],"which":[110,163],"regularizes":[111],"process":[114],"dynamically":[116],"suppressing":[117],"dominant":[118],"with":[120,152],"low":[121,153,191],"density,":[123],"encouraging":[124],"model":[126,147],"better":[128],"capture":[129],"underrepresented":[130],"cues":[132],"learn":[134],"more":[136],"balanced":[137],"distribution.":[139],"This":[140,187],"design":[141],"mitigates":[142],"risk":[144],"distribution":[148],"overfitting":[149],"density.":[155],"Second,":[156],"Representation":[160],"Alignment":[161],"(SRA),":[162],"adaptively":[164],"reweights":[165],"maps":[167],"according":[168],"importance":[171],"enforces":[173],"representation":[174],"consistency,":[175],"enabling":[176],"exert":[181],"stronger":[182],"mechanism":[188],"effectively":[189],"prevents":[190],"semantic-density":[192,212],"dominating":[195],"attention":[197],"allocation,":[198],"thereby":[199],"avoiding":[200],"dilution":[202],"provided":[209],"high":[211],"tokens.":[213],"Extensive":[214],"experiments":[215],"on":[216],"multiple":[217],"benchmark":[218],"datasets":[219],"demonstrate":[220],"DARE":[222],"consistently":[223],"generation":[225],"fidelity":[226],"alignment,":[229],"achieving":[230],"gains":[232],"over":[233],"existing":[234],"approaches.":[235]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-10T00:00:00"}