{"id":"https://openalex.org/W4412939730","doi":"https://doi.org/10.1109/tpami.2025.3593987","title":"NUPES: Non-Uniform Post-Training Quantization via Power Exponent Search","display_name":"NUPES: Non-Uniform Post-Training Quantization via Power Exponent Search","publication_year":2025,"publication_date":"2025-08-04","ids":{"openalex":"https://openalex.org/W4412939730","doi":"https://doi.org/10.1109/tpami.2025.3593987","pmid":"https://pubmed.ncbi.nlm.nih.gov/40758517"},"language":"en","primary_location":{"id":"doi:10.1109/tpami.2025.3593987","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tpami.2025.3593987","pdf_url":null,"source":{"id":"https://openalex.org/S199944782","display_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","issn_l":"0162-8828","issn":["0162-8828","1939-3539","2160-9292"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320439","host_organization_name":"IEEE Computer Society","host_organization_lineage":["https://openalex.org/P4310320439","https://openalex.org/P4310319808"],"host_organization_lineage_names":["IEEE Computer Society","Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","pubmed"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5085696448","display_name":"Edouard Yvinec","orcid":"https://orcid.org/0000-0002-4318-612X"},"institutions":[{"id":"https://openalex.org/I1338034494","display_name":"XLAB (Slovenia)","ror":"https://ror.org/04xa5qt51","country_code":"SI","type":"company","lineage":["https://openalex.org/I1338034494"]}],"countries":["SI"],"is_corresponding":false,"raw_author_name":"Edouard Yvinec","raw_affiliation_strings":["Datakalab, Paris, France","Datakalab, 114 boulevard Malesherbes, Paris, France"],"raw_orcid":"https://orcid.org/0000-0002-4318-612X","affiliations":[{"raw_affiliation_string":"Datakalab, Paris, France","institution_ids":[]},{"raw_affiliation_string":"Datakalab, 114 boulevard Malesherbes, Paris, France","institution_ids":["https://openalex.org/I1338034494"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037619048","display_name":"Arnaud Dapogny","orcid":"https://orcid.org/0000-0002-0074-8719"},"institutions":[{"id":"https://openalex.org/I1338034494","display_name":"XLAB (Slovenia)","ror":"https://ror.org/04xa5qt51","country_code":"SI","type":"company","lineage":["https://openalex.org/I1338034494"]}],"countries":["SI"],"is_corresponding":false,"raw_author_name":"Arnaud Dapogny","raw_affiliation_strings":["Datakalab, Paris, France","Datakalab, France"],"raw_orcid":"https://orcid.org/0000-0002-0074-8719","affiliations":[{"raw_affiliation_string":"Datakalab, Paris, France","institution_ids":[]},{"raw_affiliation_string":"Datakalab, France","institution_ids":["https://openalex.org/I1338034494"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5022871131","display_name":"K\u00e9vin Bailly","orcid":"https://orcid.org/0000-0001-7802-3673"},"institutions":[{"id":"https://openalex.org/I1338034494","display_name":"XLAB (Slovenia)","ror":"https://ror.org/04xa5qt51","country_code":"SI","type":"company","lineage":["https://openalex.org/I1338034494"]}],"countries":["SI"],"is_corresponding":false,"raw_author_name":"Kevin Bailly","raw_affiliation_strings":["Datakalab, Paris, France","Datakalab, France"],"raw_orcid":"https://orcid.org/0000-0001-7802-3673","affiliations":[{"raw_affiliation_string":"Datakalab, Paris, France","institution_ids":[]},{"raw_affiliation_string":"Datakalab, France","institution_ids":["https://openalex.org/I1338034494"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.08682274,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"47","issue":"11","first_page":"10012","last_page":"10021"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9347000122070312,"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/T10028","display_name":"Topic Modeling","score":0.9347000122070312,"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":0.6462883949279785},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5945054888725281},{"id":"https://openalex.org/keywords/exponent","display_name":"Exponent","score":0.5828467607498169},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5436711311340332},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.46798086166381836},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.4678904116153717},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.41897037625312805},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.34263211488723755},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.3262873888015747},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.32528775930404663},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.13307541608810425}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6462883949279785},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5945054888725281},{"id":"https://openalex.org/C2780388253","wikidata":"https://www.wikidata.org/wiki/Q5421508","display_name":"Exponent","level":2,"score":0.5828467607498169},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5436711311340332},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.46798086166381836},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.4678904116153717},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.41897037625312805},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.34263211488723755},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3262873888015747},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.32528775930404663},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.13307541608810425},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.0},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/tpami.2025.3593987","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tpami.2025.3593987","pdf_url":null,"source":{"id":"https://openalex.org/S199944782","display_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","issn_l":"0162-8828","issn":["0162-8828","1939-3539","2160-9292"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320439","host_organization_name":"IEEE Computer Society","host_organization_lineage":["https://openalex.org/P4310320439","https://openalex.org/P4310319808"],"host_organization_lineage_names":["IEEE Computer Society","Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Pattern Analysis and Machine Intelligence","raw_type":"journal-article"},{"id":"pmid:40758517","is_oa":false,"landing_page_url":"https://pubmed.ncbi.nlm.nih.gov/40758517","pdf_url":null,"source":{"id":"https://openalex.org/S4306525036","display_name":"PubMed","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1299303238","host_organization_name":"National Institutes of Health","host_organization_lineage":["https://openalex.org/I1299303238"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE transactions on pattern analysis and machine intelligence","raw_type":null}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Climate action","id":"https://metadata.un.org/sdg/13","score":0.4699999988079071}],"awards":[{"id":"https://openalex.org/G3584957902","display_name":"Face Interpretation with deep and ensemble Learning","funder_award_id":"ANR-17-CE33-0002","funder_id":"https://openalex.org/F4320320883","funder_display_name":"Agence Nationale de la Recherche"}],"funders":[{"id":"https://openalex.org/F4320320883","display_name":"Agence Nationale de la Recherche","ror":"https://ror.org/00rbzpz17"},{"id":"https://openalex.org/F4320321663","display_name":"Association Nationale de la Recherche et de la Technologie","ror":"https://ror.org/00ht2ab73"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":20,"referenced_works":["https://openalex.org/W1861492603","https://openalex.org/W2108598243","https://openalex.org/W2171074980","https://openalex.org/W2194775991","https://openalex.org/W2783538964","https://openalex.org/W2890894339","https://openalex.org/W2946609015","https://openalex.org/W2963163009","https://openalex.org/W2963351448","https://openalex.org/W2981751377","https://openalex.org/W2998617917","https://openalex.org/W3004061291","https://openalex.org/W3097297321","https://openalex.org/W3131984434","https://openalex.org/W3153887752","https://openalex.org/W3194676777","https://openalex.org/W4226419870","https://openalex.org/W4312705808","https://openalex.org/W4317934857","https://openalex.org/W4386076113"],"related_works":["https://openalex.org/W230091440","https://openalex.org/W2233261550","https://openalex.org/W2467119940","https://openalex.org/W2810751659","https://openalex.org/W258997015","https://openalex.org/W2997094352","https://openalex.org/W4412817058","https://openalex.org/W3216976533","https://openalex.org/W2110652681","https://openalex.org/W3209251257"],"abstract_inverted_index":{"Deep":[0],"neural":[1],"network":[2],"(DNN)":[3],"deployment":[4],"has":[5,20],"been":[6],"confined":[7],"to":[8,13,35,55,74,106,126,141,183,192,256,275],"larger":[9],"hardware":[10],"devices":[11],"due":[12],"their":[14,38],"expensive":[15],"computational":[16],"requirements.":[17],"This":[18,71,95],"challenge":[19],"recently":[21],"reached":[22],"another":[23],"scale":[24],"with":[25,174,201,245],"the":[26,76,121,143,154,157,194,210,217,220,224,227,236,251,254,271,277],"emergence":[27],"of":[28,156,219,226,253,273,279],"large":[29,287],"language":[30,288],"models":[31,133,289],"(LLMs).":[32],"In":[33,113],"order":[34,191],"reduce":[36],"both":[37],"memory":[39],"footprint":[40],"and":[41,89,160,166,264,286],"latency,":[42],"a":[43,64,68,92,164,202],"promising":[44],"technique":[45],"is":[46,96,243],"quantization.":[47,137],"It":[48],"consists":[49],"in":[50,75,130,190,261,290],"converting":[51],"floating":[52],"point":[53,59],"representations":[54],"low":[56],"bit-width":[57],"fixed":[58],"representations,":[60],"usually":[61],"by":[62,233],"assuming":[63],"uniform":[65,79],"mapping":[66],"onto":[67],"regular":[69],"grid.":[70],"process,":[72],"referred":[73],"literature":[77],"as":[78,85],"quantization,":[80,134],"may":[81],"however":[82],"be":[83,172],"ill-suited":[84],"most":[86,122],"DNN":[87],"weights":[88,208],"activations":[90],"follow":[91],"bell-shaped":[93],"distribution.":[94],"even":[97],"worse":[98],"on":[99,284],"LLMs":[100],"whose":[101],"weight":[102,161,188],"distributions":[103],"are":[104,148],"known":[105],"exhibit":[107],"large,":[108],"high":[109],"impact,":[110],"outlier":[111],"values.":[112],"this":[114,128,199],"work,":[115],"we":[116,215],"propose":[117],"an":[118],"improvement":[119],"over":[120,209],"commonly":[123],"adopted":[124],"way":[125],"tackle":[127],"limitation":[129,200],"deep":[131],"learning":[132,205],"namely,":[135],"non-uniform":[136],"NUPES":[138,274],"leverages":[139],"automorphisms":[140],"preserve":[142,193],"scalar":[144],"multiplications.":[145],"Such":[146],"transformations":[147],"derived":[149],"from":[150],"power":[151,221],"functions.":[152],"However,":[153],"optimization":[155,178,218,225],"exponent":[158],"parameter":[159],"values":[162,189],"remains":[163],"challenging":[165],"novel":[167],"problem":[168],"which":[169,180],"could":[170],"not":[171],"solved":[173],"previous":[175,280],"post":[176],"training":[177,232],"techniques":[179,283],"only":[181],"learn":[182],"round":[184],"up":[185],"or":[186],"down":[187],"predictive":[195,241],"function.":[196],"We":[197,249],"circumvent":[198,276],"new":[203,206],"paradigm:":[204],"quantized":[207,212],"entire":[211],"space.":[213],"Similarly,":[214],"enable":[216],"exponent,":[222],"i.e.":[223],"quantization":[228,282],"operator":[229],"itself":[230],"during":[231],"alleviating":[234],"all":[235],"numerical":[237],"instabilities.":[238],"The":[239],"resulting":[240],"function":[242],"compatible":[244],"integer-only":[246],"low-bit":[247],"inference.":[248],"show":[250],"ability":[252,272],"method":[255],"achieve":[257],"state-of-the-art":[258],"compression":[259],"rates":[260],"both,":[262],"data-free":[263],"data-driven":[265],"configurations.":[266],"Our":[267],"empirical":[268],"benchmarks":[269],"highlight":[270],"limitations":[278],"post-training":[281],"transformers":[285],"particular.":[291]},"counts_by_year":[],"updated_date":"2026-06-22T08:00:12.763002","created_date":"2025-10-10T00:00:00"}