{"id":"https://openalex.org/W7160299616","doi":"https://doi.org/10.48550/arxiv.2605.01330","title":"Colinearity Decay: Training Quantization-Friendly ViTs with Outlier Decay","display_name":"Colinearity Decay: Training Quantization-Friendly ViTs with Outlier Decay","publication_year":2026,"publication_date":"2026-05-02","ids":{"openalex":"https://openalex.org/W7160299616","doi":"https://doi.org/10.48550/arxiv.2605.01330"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.01330","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01330","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.01330","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135341029","display_name":"Jin Tong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tong, Jin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110853420","display_name":"Guang Liang","orcid":"https://orcid.org/0009-0003-1933-5783"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Guang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133835478","display_name":"Peilin Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Peilin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135360581","display_name":"Jianxin Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Jianxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"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":0.5733000040054321,"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":0.5733000040054321,"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/T10502","display_name":"Advanced Memory and Neural Computing","score":0.09960000216960907,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.058800000697374344,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.7512999773025513},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5756000280380249},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4625000059604645},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.45089998841285706},{"id":"https://openalex.org/keywords/anomaly-detection","display_name":"Anomaly detection","score":0.41280001401901245},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3555000126361847},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.34389999508857727},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.32359999418258667}],"concepts":[{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.7512999773025513},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6290000081062317},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5778999924659729},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5756000280380249},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4625000059604645},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.45089998841285706},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4410000145435333},{"id":"https://openalex.org/C739882","wikidata":"https://www.wikidata.org/wiki/Q3560506","display_name":"Anomaly detection","level":2,"score":0.41280001401901245},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3555000126361847},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.34389999508857727},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.32359999418258667},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.29660001397132874},{"id":"https://openalex.org/C2780150128","wikidata":"https://www.wikidata.org/wiki/Q21948731","display_name":"Extreme learning machine","level":3,"score":0.2957000136375427},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2948000133037567},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2946000099182129},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.2793000042438507},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.27230000495910645},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.2632000148296356},{"id":"https://openalex.org/C67226441","wikidata":"https://www.wikidata.org/wiki/Q1665389","display_name":"Robust statistics","level":3,"score":0.251800000667572}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.01330","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01330","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.01330","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.01330","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[{"display_name":"No poverty","id":"https://metadata.un.org/sdg/1","score":0.5218650698661804}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Low-bit":[0],"quantization":[1,22],"is":[2,107],"a":[3,40,75,103],"practical":[4],"route":[5],"for":[6,78,148],"efficiently":[7],"deploying":[8],"vision":[9,35,146],"Transformers,":[10],"yet":[11],"activation":[12],"outliers":[13,33],"complicate":[14],"fully":[15],"quantized":[16,46,125],"deployment.":[17],"Existing":[18],"methods":[19],"either":[20],"handle":[21],"post-training":[23],"or":[24,98,132],"suppress":[25],"large":[26],"activations":[27,93],"during":[28],"training;":[29],"however,":[30],"aggressively":[31],"restricting":[32],"in":[34],"models":[36],"can":[37],"lead":[38],"to":[39],"poorer":[41],"trade-off":[42],"between":[43],"full-precision":[44,135],"and":[45,90,109,119],"accuracy.":[47],"We":[48],"argue":[49],"that":[50,64,141],"rather":[51],"than":[52],"simply":[53],"suppressing":[54],"outliers,":[55],"the":[56,61,96],"training":[57,112],"objective":[58],"should":[59],"control":[60],"structural":[62,76,142],"amplification":[63],"makes":[65],"them":[66],"harmful.":[67],"To":[68],"this":[69],"end,":[70],"we":[71],"introduce":[72],"Colinearity-Decay":[73],"(CD),":[74],"regularizer":[77],"ordered":[79],"matrix":[80],"pairs":[81],"within":[82],"Transformer":[83],"blocks.":[84],"CD":[85,106,122],"penalizes":[86],"detrimental":[87],"cross-matrix":[88],"alignment":[89],"mitigates":[91],"extreme":[92],"without":[94],"altering":[95],"architecture":[97],"task":[99],"loss.":[100],"Applied":[101],"as":[102],"decoupled":[104],"update,":[105],"non-invasive":[108],"introduces":[110],"minimal":[111],"overhead.":[113,154],"Across":[114],"ImageNet-1K":[115],"pre-training,":[116],"COCO":[117],"detection,":[118],"downstream":[120],"fine-tuning,":[121],"consistently":[123],"boosts":[124],"accuracy":[126],"across":[127],"multiple":[128],"pipelines":[129],"while":[130],"preserving,":[131],"even":[133],"improving,":[134],"performance.":[136],"Ultimately,":[137],"our":[138],"results":[139],"demonstrate":[140],"regularization":[143],"effectively":[144],"prepares":[145],"Transformers":[147],"low-bit":[149],"deployment":[150],"with":[151],"zero":[152],"inference-time":[153]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-06T00:00:00"}