{"id":"https://openalex.org/W7138884604","doi":"https://doi.org/10.48550/arxiv.2603.17354","title":"Beyond Outliers: A Data-Free Layer-wise Mixed-Precision Quantization Approach Driven by Numerical and Structural Dual-Sensitivity","display_name":"Beyond Outliers: A Data-Free Layer-wise Mixed-Precision Quantization Approach Driven by Numerical and Structural Dual-Sensitivity","publication_year":2026,"publication_date":"2026-03-18","ids":{"openalex":"https://openalex.org/W7138884604","doi":"https://doi.org/10.48550/arxiv.2603.17354"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.17354","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.17354","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.2603.17354","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129834530","display_name":"Hengyuan Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Hengyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129994549","display_name":"Xinrong Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Xinrong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122136707","display_name":"Zunhai Su","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Su, Zunhai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129946115","display_name":"Xiao Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Xiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100567751","display_name":"Jing Xiong","orcid":"https://orcid.org/0009-0000-2263-124X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong, Jing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130211562","display_name":"Wendong Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Wendong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129798934","display_name":"He Xiao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao, He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129834444","display_name":"Chaofan Tao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tao, Chaofan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129762218","display_name":"Wei Zhang (405)","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101577090","display_name":"Ruobing Xie","orcid":"https://orcid.org/0000-0003-3170-5647"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Ruobing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129762191","display_name":"Lei Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Lei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129907888","display_name":"Hayden Kwok-Hay So","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"So, Hayden Kwok-Hay","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130047835","display_name":"Ngai Wong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wong, Ngai","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/T10901","display_name":"Advanced Data Compression Techniques","score":0.7228999733924866,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.7228999733924866,"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/T11165","display_name":"Image and Video Quality Assessment","score":0.043699998408555984,"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/T12131","display_name":"Wireless Signal Modulation Classification","score":0.03500000014901161,"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/quantization","display_name":"Quantization (signal processing)","score":0.613099992275238},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.5115000009536743},{"id":"https://openalex.org/keywords/compression","display_name":"Compression (physics)","score":0.4399999976158142},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.3986000120639801},{"id":"https://openalex.org/keywords/numerical-models","display_name":"Numerical models","score":0.36820000410079956},{"id":"https://openalex.org/keywords/performance-metric","display_name":"Performance metric","score":0.3637000024318695},{"id":"https://openalex.org/keywords/numerical-analysis","display_name":"Numerical analysis","score":0.35670000314712524},{"id":"https://openalex.org/keywords/scheme","display_name":"Scheme (mathematics)","score":0.3529999852180481}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.613099992275238},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5515999794006348},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.5115000009536743},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.508899986743927},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.4399999976158142},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.3986000120639801},{"id":"https://openalex.org/C2986605239","wikidata":"https://www.wikidata.org/wiki/Q925667","display_name":"Numerical models","level":3,"score":0.36820000410079956},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.3637000024318695},{"id":"https://openalex.org/C48753275","wikidata":"https://www.wikidata.org/wiki/Q11216","display_name":"Numerical analysis","level":2,"score":0.35670000314712524},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.3529999852180481},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.33899998664855957},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.32760000228881836},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.32409998774528503},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.32350000739097595},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.3124000132083893},{"id":"https://openalex.org/C176321772","wikidata":"https://www.wikidata.org/wiki/Q1430640","display_name":"Numerical stability","level":3,"score":0.3102000057697296},{"id":"https://openalex.org/C500300565","wikidata":"https://www.wikidata.org/wiki/Q925667","display_name":"Computer simulation","level":2,"score":0.30660000443458557},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.3028999865055084},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.29260000586509705},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.28459998965263367},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.28299999237060547},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2791000008583069},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.25429999828338623}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.17354","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.17354","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.2603.17354","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.17354","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Layer-wise":[0],"mixed-precision":[1],"quantization":[2],"(LMPQ)":[3],"enables":[4],"effective":[5],"compression":[6],"under":[7],"extreme":[8],"low-bit":[9],"settings":[10],"by":[11,58],"allocating":[12],"higher":[13],"precision":[14],"to":[15,105,119],"sensitive":[16],"layers.":[17],"However,":[18],"existing":[19],"methods":[20],"typically":[21],"treat":[22],"all":[23],"intra-layer":[24],"weight":[25],"modules":[26],"uniformly":[27],"and":[28,43,60,74,81,103,125],"rely":[29],"on":[30,101,130],"a":[31,52,91,96],"single":[32],"numerical":[33,80],"property":[34],"when":[35],"estimating":[36],"sensitivity,":[37],"overlooking":[38],"their":[39,76],"distinct":[40,71],"operational":[41,72],"roles":[42,73],"structural":[44,82],"characteristics.":[45],"To":[46],"address":[47],"this,":[48],"we":[49],"propose":[50],"NSDS,":[51],"novel":[53],"calibration-free":[54],"LMPQ":[55],"framework":[56],"driven":[57],"Numerical":[59],"Structural":[61],"Dual-Sensitivity.":[62],"Specifically,":[63],"it":[64],"first":[65],"mechanistically":[66],"decomposes":[67],"each":[68],"layer":[69],"into":[70,90],"quantifies":[75],"sensitivity":[77],"from":[78],"both":[79],"perspectives.":[83],"These":[84],"dual-aspect":[85],"scores":[86],"are":[87],"then":[88],"aggregated":[89],"unified":[92],"layer-wise":[93],"metric":[94],"through":[95],"robust":[97],"aggregation":[98],"scheme":[99],"based":[100],"MAD-Sigmoid":[102],"Soft-OR":[104],"guide":[106],"bit":[107],"allocation.":[108],"Extensive":[109],"experiments":[110],"demonstrate":[111],"that":[112],"NSDS":[113],"consistently":[114],"achieves":[115],"superior":[116],"performance":[117],"compared":[118],"various":[120],"baselines":[121],"across":[122],"diverse":[123],"models":[124],"downstream":[126],"tasks,":[127],"without":[128],"relying":[129],"any":[131],"calibration":[132],"data.":[133]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-20T00:00:00"}