{"id":"https://openalex.org/W7167043125","doi":"https://doi.org/10.48550/arxiv.2607.00927","title":"Post-Training Pruning for Diffusion Transformers","display_name":"Post-Training Pruning for Diffusion Transformers","publication_year":2026,"publication_date":"2026-07-01","ids":{"openalex":"https://openalex.org/W7167043125","doi":"https://doi.org/10.48550/arxiv.2607.00927"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.00927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.00927","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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.2607.00927","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139905406","display_name":"Chengzhi Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Chengzhi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139920776","display_name":"Xuewen Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Xuewen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139872882","display_name":"Jing Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139928721","display_name":"Mengjuan Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Mengjuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139937588","display_name":"Zhikai Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zhikai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139922369","display_name":"Qingyi Gu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gu, Qingyi","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/T11019","display_name":"Image Enhancement Techniques","score":0.5871000289916992,"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/T11019","display_name":"Image Enhancement Techniques","score":0.5871000289916992,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.1014999970793724,"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/T11996","display_name":"Random lasers and scattering media","score":0.06830000132322311,"subfield":{"id":"https://openalex.org/subfields/3102","display_name":"Acoustics and Ultrasonics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.7354999780654907},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.5220000147819519},{"id":"https://openalex.org/keywords/cluster-analysis","display_name":"Cluster analysis","score":0.5116999745368958},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.435699999332428},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3797000050544739},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.33160001039505005}],"concepts":[{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.7354999780654907},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6248000264167786},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.5220000147819519},{"id":"https://openalex.org/C73555534","wikidata":"https://www.wikidata.org/wiki/Q622825","display_name":"Cluster analysis","level":2,"score":0.5116999745368958},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4546000063419342},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.435699999332428},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4171999990940094},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3797000050544739},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.34869998693466187},{"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/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.33160001039505005},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.3246000111103058},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.29420000314712524},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.27059999108314514},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.2612000107765198},{"id":"https://openalex.org/C2988416141","wikidata":"https://www.wikidata.org/wiki/Q6031139","display_name":"Information loss","level":2,"score":0.25839999318122864}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.00927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.00927","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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.2607.00927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.00927","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"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":"Decent work and economic growth","id":"https://metadata.un.org/sdg/8","score":0.5371038913726807}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"Transformers":[1],"(DiTs)":[2],"have":[3],"demonstrated":[4],"impressive":[5],"performance":[6,43,100],"in":[7,65,72,80,88,139,185],"image":[8,166],"generation":[9],"but":[10],"suffer":[11],"from":[12,122],"substantial":[13],"computational":[14],"overhead":[15],"and":[16,32,106,120],"resource":[17],"consumption.":[18],"Post-training":[19],"pruning":[20,36,84,99,107,149,194],"offers":[21],"a":[22,55,111,147,182],"promising":[23],"solution;":[24],"however,":[25],"due":[26],"to":[27,41],"DiTs'":[28],"unique":[29],"architectural":[30],"design":[31,110],"parameter":[33],"distribution,":[34],"traditional":[35],"methods":[37,47],"are":[38],"inapplicable,":[39],"leading":[40],"significant":[42],"degradation.":[44],"Specifically,":[45],"prior":[46],"developed":[48],"for":[49],"LLMs,":[50],"which":[51,97],"derive":[52],"metrics":[53],"through":[54],"series":[56],"of":[57,63,118,130],"approximations,":[58],"amplify":[59],"the":[60,66,116,140],"relative":[61],"contribution":[62],"weights":[64,71,119],"saliency":[67,104],"metric.":[68],"In":[69,91],"addition,":[70],"DiTs":[73,159],"exhibit":[74],"significantly":[75],"larger":[76],"magnitudes":[77],"than":[78],"those":[79],"LLMs.":[81],"Moreover,":[82],"existing":[83],"granularity":[85],"overlooks":[86],"variations":[87],"model":[89],"structures.":[90],"this":[92],"paper,":[93],"we":[94,134,145],"propose":[95],"DiT-Pruning,":[96],"improves":[98],"by":[101],"introducing":[102],"customized":[103],"criteria":[105],"granularity.":[108],"We":[109],"novel":[112],"metric":[113],"that":[114,161],"balances":[115],"contributions":[117],"activations":[121],"an":[123],"energy-based":[124],"perspective,":[125],"enabling":[126,151],"more":[127],"effective":[128,152],"identification":[129],"important":[131],"elements.":[132],"Furthermore,":[133],"observe":[135],"distinct":[136],"clustering":[137],"patterns":[138],"two-dimensional":[141],"weight":[142],"space.":[143],"Accordingly,":[144],"adopt":[146],"clustering-aware":[148],"granularity,":[150],"sparse":[153],"allocation.":[154],"Extensive":[155],"evaluations":[156],"on":[157,177],"various":[158],"show":[160],"our":[162],"method":[163],"consistently":[164],"preserves":[165],"quality,":[167],"especially":[168],"under":[169],"high":[170],"sparsity.":[171],"For":[172],"FLUX.1-dev":[173],"at":[174,188],"512x512":[175],"resolution":[176],"MJHQ,":[178],"DiT-Pruning":[179],"achieves":[180],"only":[181],"0.001":[183],"loss":[184],"CLIP":[186],"score":[187],"50%":[189],"sparsity,":[190],"dramatically":[191],"outperforming":[192],"recent":[193],"methods.":[195]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-07-03T00:00:00"}