{"id":"https://openalex.org/W7165657213","doi":"https://doi.org/10.48550/arxiv.2606.21947","title":"ScalePredictor: Instance-aware Scale Learning for Accurate Quantization of Vision Transformers","display_name":"ScalePredictor: Instance-aware Scale Learning for Accurate Quantization of Vision Transformers","publication_year":2026,"publication_date":"2026-06-20","ids":{"openalex":"https://openalex.org/W7165657213","doi":"https://doi.org/10.48550/arxiv.2606.21947"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.21947","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21947","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":"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.2606.21947","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101876244","display_name":"Changjun Li","orcid":"https://orcid.org/0000-0002-9942-7690"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Changjun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139153555","display_name":"Runqing Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Runqing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015286750","display_name":"Lei Xu","orcid":"https://orcid.org/0009-0003-1055-9318"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Lian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139174126","display_name":"Ye Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Ye","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036921959","display_name":"Qingyong Hu","orcid":"https://orcid.org/0000-0003-0337-9207"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Qingyong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5101893660","display_name":"Yi Guo","orcid":"https://orcid.org/0009-0006-1890-6726"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Yulan","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/T10036","display_name":"Advanced Neural Network Applications","score":0.5834000110626221,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.5834000110626221,"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/T10531","display_name":"Advanced Vision and Imaging","score":0.093299999833107,"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/T11019","display_name":"Image Enhancement Techniques","score":0.0851999968290329,"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/quantization","display_name":"Quantization (signal processing)","score":0.661300003528595},{"id":"https://openalex.org/keywords/rendering","display_name":"Rendering (computer graphics)","score":0.4413999915122986},{"id":"https://openalex.org/keywords/high-dynamic-range","display_name":"High dynamic range","score":0.4404999911785126},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.3734000027179718},{"id":"https://openalex.org/keywords/edge-device","display_name":"Edge device","score":0.3589000105857849},{"id":"https://openalex.org/keywords/computational-complexity-theory","display_name":"Computational complexity theory","score":0.33390000462532043}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.661300003528595},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5771999955177307},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.4413999915122986},{"id":"https://openalex.org/C2780056265","wikidata":"https://www.wikidata.org/wiki/Q106239881","display_name":"High dynamic range","level":3,"score":0.4404999911785126},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43149998784065247},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3849000036716461},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.3734000027179718},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.3589000105857849},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.33660000562667847},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.33390000462532043},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3156999945640564},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.3050000071525574},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2822999954223633},{"id":"https://openalex.org/C90119067","wikidata":"https://www.wikidata.org/wiki/Q43260","display_name":"Polynomial","level":2,"score":0.2802000045776367},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.2718000113964081},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.2637999951839447},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.26190000772476196}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.21947","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21947","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":"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.2606.21947","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21947","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":"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":[{"display_name":"Decent work and economic growth","id":"https://metadata.un.org/sdg/8","score":0.4286429286003113}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Vision":[0],"Transformers":[1],"have":[2],"achieved":[3],"remarkable":[4],"success":[5],"in":[6,199],"many":[7],"fields,":[8],"yet":[9],"their":[10,19],"deployment":[11],"on":[12,118,177],"edge":[13],"devices":[14],"remains":[15],"challenging":[16],"due":[17],"to":[18,56,133,153],"substantial":[20,61],"computational":[21,168],"demands.":[22],"Post-Training":[23],"Quantization":[24],"(PTQ)":[25],"offers":[26],"an":[27,128],"attractive":[28],"solution":[29],"by":[30],"compressing":[31],"models":[32],"using":[33],"a":[34,48,84,100,122,147,188],"small":[35],"calibration":[36],"set":[37],"with":[38],"minimal":[39],"training":[40],"overhead.":[41],"However,":[42],"most":[43],"existing":[44],"PTQ":[45,185],"works":[46],"adopt":[47],"static":[49],"quantization":[50,86,92,156],"paradigm":[51],"that":[52,126,180],"is":[53],"uniformly":[54],"applied":[55],"all":[57,155],"instances.":[58],"Given":[59],"the":[60,66,104,111,139,160,200],"diversity":[62],"of":[63,95,107,114,162],"natural":[64],"images,":[65],"activation":[67],"distributions":[68],"vary":[69],"significantly":[70],"across":[71],"samples,":[72],"making":[73],"these":[74],"methods":[75],"inherently":[76],"suboptimal.":[77],"In":[78],"this":[79],"paper,":[80],"we":[81,120],"propose":[82],"ScalePredictor,":[83],"dynamic":[85],"framework":[87],"for":[88],"accurate":[89],"and":[90,110,194],"efficient":[91,129],"scale":[93,123,150],"learning":[94,124],"ViTs.":[96],"We":[97],"first":[98],"reveal":[99],"hidden":[101],"correlation":[102],"between":[103],"distribution":[105],"range":[106,130,136],"shallow-layer":[108],"activations":[109],"optimal":[112],"scales":[113,157],"deeper":[115],"layers.":[116],"Based":[117],"this,":[119],"develop":[121],"mechanism":[125],"integrates":[127],"extraction":[131],"approach":[132],"capture":[134],"robust":[135],"statistics":[137],"at":[138],"shallow":[140],"stage,":[141],"which":[142],"are":[143,197],"then":[144],"fed":[145],"into":[146],"Taylor-motivated":[148],"polynomial":[149,163],"projection":[151],"module":[152],"generate":[154],"simultaneously.":[158],"With":[159],"efficiency":[161],"approximation,":[164],"ScalePredictor":[165,181],"introduces":[166],"insignificant":[167],"overhead":[169],"while":[170],"avoiding":[171],"costly":[172],"just-in-time":[173],"calibration.":[174],"Extensive":[175],"experiments":[176],"ImageNet":[178],"demonstrate":[179],"consistently":[182],"outperforms":[183],"prior":[184],"methods,":[186],"achieving":[187],"more":[189],"favorable":[190],"accuracy-efficiency":[191],"trade-off.":[192],"Code":[193],"additional":[195],"results":[196],"shown":[198],"supplementary":[201],"materials.":[202]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-06-24T00:00:00"}