{"id":"https://openalex.org/W7154087423","doi":"https://doi.org/10.48550/arxiv.2604.09088","title":"Memory-Efficient Transfer Learning with Fading Side Networks via Masked Dual Path Distillation","display_name":"Memory-Efficient Transfer Learning with Fading Side Networks via Masked Dual Path Distillation","publication_year":2026,"publication_date":"2026-04-10","ids":{"openalex":"https://openalex.org/W7154087423","doi":"https://doi.org/10.48550/arxiv.2604.09088"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.09088","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09088","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.2604.09088","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133533517","display_name":"Yutong Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yutong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133502681","display_name":"Jiaxin Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Jiaxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133542016","display_name":"Honglin Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Honglin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133497065","display_name":"Kaiqi Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Kaiqi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133524695","display_name":"Shengcai Liao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liao, Shengcai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113402664","display_name":"Hanwen Zhong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Hanwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133526377","display_name":"Weixin Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Weixin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133523554","display_name":"Yunhong Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yunhong","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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.5090000033378601,"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"}},"topics":[{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.5090000033378601,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.2199999988079071,"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":0.16539999842643738,"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/overhead","display_name":"Overhead (engineering)","score":0.70169997215271},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6629999876022339},{"id":"https://openalex.org/keywords/fading","display_name":"Fading","score":0.5821999907493591},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.573199987411499},{"id":"https://openalex.org/keywords/path","display_name":"Path (computing)","score":0.559499979019165},{"id":"https://openalex.org/keywords/dual","display_name":"Dual (grammatical number)","score":0.5460000038146973},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.4812999963760376}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7958999872207642},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.70169997215271},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6629999876022339},{"id":"https://openalex.org/C81978471","wikidata":"https://www.wikidata.org/wiki/Q1196572","display_name":"Fading","level":3,"score":0.5821999907493591},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.573199987411499},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.559499979019165},{"id":"https://openalex.org/C2780980858","wikidata":"https://www.wikidata.org/wiki/Q110022","display_name":"Dual (grammatical number)","level":2,"score":0.5460000038146973},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.4812999963760376},{"id":"https://openalex.org/C2776175482","wikidata":"https://www.wikidata.org/wiki/Q1195816","display_name":"Transfer (computing)","level":2,"score":0.4101000130176544},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.39340001344680786},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35109999775886536},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3321000039577484},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.3183000087738037},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.313400000333786},{"id":"https://openalex.org/C3454156","wikidata":"https://www.wikidata.org/wiki/Q1144241","display_name":"Side effect (computer science)","level":2,"score":0.3122999966144562},{"id":"https://openalex.org/C155032097","wikidata":"https://www.wikidata.org/wiki/Q798503","display_name":"Backpropagation","level":3,"score":0.2872999906539917},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.26930001378059387},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2655999958515167}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.09088","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09088","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.2604.09088","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09088","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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Memory-efficient":[0],"transfer":[1,68],"learning":[2],"(METL)":[3],"approaches":[4],"have":[5],"recently":[6],"achieved":[7],"promising":[8],"performance":[9,109],"in":[10,22,95,120],"adapting":[11],"pre-trained":[12],"models":[13],"to":[14,86,184],"downstream":[15],"tasks.":[16],"They":[17],"avoid":[18],"applying":[19],"gradient":[20],"backpropagation":[21],"large":[23],"backbones,":[24],"thus":[25],"significantly":[26],"reducing":[27],"the":[28,63,72,108,113,124,142,181],"number":[29],"of":[30,66],"trainable":[31],"parameters":[32],"and":[33,46,56,92,116,122,155,173],"high":[34],"memory":[35,55,93,174],"consumption":[36,175],"during":[37,59,127],"fine-tuning.":[38],"However,":[39],"since":[40],"they":[41],"typically":[42],"employ":[43],"a":[44,77,104,135],"lightweight":[45],"learnable":[47,117],"side":[48,99,118,125],"network,":[49],"these":[50],"methods":[51],"inevitably":[52],"introduce":[53],"additional":[54],"time":[57],"overhead":[58],"inference,":[60],"which":[61],"contradicts":[62],"ultimate":[64],"goal":[65],"efficient":[67],"learning.":[69],"To":[70],"address":[71],"above":[73],"issue,":[74],"we":[75,133],"propose":[76],"novel":[78,136],"approach":[79],"dubbed":[80],"Masked":[81],"Dual":[82],"Path":[83],"Distillation":[84],"(MDPD)":[85],"accelerate":[87],"inference":[88,128,165],"while":[89,170],"retaining":[90],"parameter":[91,172],"efficiency":[94],"fine-tuning":[96],"with":[97,145],"fading":[98],"networks.":[100],"Specifically,":[101],"MDPD":[102],"develops":[103],"framework":[105],"that":[106,159],"enhances":[107],"by":[110,166],"mutually":[111],"distilling":[112],"frozen":[114],"backbones":[115,152],"networks":[119],"fine-tuning,":[121],"discard":[123],"network":[126],"without":[129],"sacrificing":[130],"accuracy.":[131],"Moreover,":[132],"design":[134],"feature-based":[137],"knowledge":[138],"distillation":[139],"method":[140,161],"for":[141],"encoder":[143],"structure":[144],"multiple":[146],"layers.":[147],"Extensive":[148],"experiments":[149],"on":[150],"distinct":[151],"across":[153],"vision/language-only":[154],"vision-and-language":[156],"tasks":[157],"demonstrate":[158],"our":[160],"not":[162],"only":[163],"accelerates":[164],"at":[167,192],"least":[168],"25.2\\%":[169],"keeping":[171],"comparable,":[176],"but":[177],"also":[178],"remarkably":[179],"promotes":[180],"accuracy":[182],"compared":[183],"SOTA":[185],"approaches.":[186],"The":[187],"source":[188],"code":[189],"is":[190],"available":[191],"https://github.com/Zhang-VKk/MDPD.":[193]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-14T00:00:00"}