{"id":"https://openalex.org/W7148489039","doi":"https://doi.org/10.1109/access.2026.3680045","title":"UDP: Up-or-Down Precision Quantization for Large Language Models via Evolutionary Search","display_name":"UDP: Up-or-Down Precision Quantization for Large Language Models via Evolutionary Search","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7148489039","doi":"https://doi.org/10.1109/access.2026.3680045"},"language":"en","primary_location":{"id":"doi:10.1109/access.2026.3680045","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2026.3680045","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1109/access.2026.3680045","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124928636","display_name":"Anh-Tuan Mai","orcid":null},"institutions":[{"id":"https://openalex.org/I109689652","display_name":"FPT University","ror":"https://ror.org/03esj4g97","country_code":"VN","type":"education","lineage":["https://openalex.org/I109689652"]}],"countries":["VN"],"is_corresponding":false,"raw_author_name":"Anh-Tuan Mai","raw_affiliation_strings":["FPT Software AI Center, Hanoi, Vietnam"],"raw_orcid":"https://orcid.org/0009-0002-7211-2637","affiliations":[{"raw_affiliation_string":"FPT Software AI Center, Hanoi, Vietnam","institution_ids":["https://openalex.org/I109689652"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132789614","display_name":"Tien-Son Pham","orcid":null},"institutions":[{"id":"https://openalex.org/I177233841","display_name":"Vietnam National University, Hanoi","ror":"https://ror.org/02jmfj006","country_code":"VN","type":"education","lineage":["https://openalex.org/I177233841"]}],"countries":["VN"],"is_corresponding":false,"raw_author_name":"Tien-Son Pham","raw_affiliation_strings":["Efficient Computing Research Group, Vietnam National University, Hanoi, Vietnam"],"raw_orcid":"https://orcid.org/0009-0008-3944-131X","affiliations":[{"raw_affiliation_string":"Efficient Computing Research Group, Vietnam National University, Hanoi, Vietnam","institution_ids":["https://openalex.org/I177233841"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014900695","display_name":"Xuan Truong Nguyen","orcid":"https://orcid.org/0000-0002-7527-6971"},"institutions":[{"id":"https://openalex.org/I124125349","display_name":"Korea National Open University","ror":"https://ror.org/016ebag96","country_code":"KR","type":"education","lineage":["https://openalex.org/I124125349"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Xuan Truong Nguyen","raw_affiliation_strings":["Department of Next-Generation Semiconductor Convergence and Open Sharing System (COSS), Seoul National University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0000-0002-7527-6971","affiliations":[{"raw_affiliation_string":"Department of Next-Generation Semiconductor Convergence and Open Sharing System (COSS), Seoul National University, Seoul, South Korea","institution_ids":["https://openalex.org/I124125349"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5124951464","display_name":"Thanh-Tuan Dao","orcid":null},"institutions":[{"id":"https://openalex.org/I177233841","display_name":"Vietnam National University, Hanoi","ror":"https://ror.org/02jmfj006","country_code":"VN","type":"education","lineage":["https://openalex.org/I177233841"]}],"countries":["VN"],"is_corresponding":false,"raw_author_name":"Thanh-Tuan Dao","raw_affiliation_strings":["Efficient Computing Research Group, Vietnam National University, Hanoi, Vietnam"],"raw_orcid":"https://orcid.org/0000-0002-9897-2769","affiliations":[{"raw_affiliation_string":"Efficient Computing Research Group, Vietnam National University, Hanoi, Vietnam","institution_ids":["https://openalex.org/I177233841"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":{"value":1850,"currency":"USD","value_usd":1850},"apc_paid":{"value":1850,"currency":"USD","value_usd":1850},"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.37152844,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"14","issue":null,"first_page":"57033","last_page":"57047"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.18279999494552612,"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":0.18279999494552612,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1599999964237213,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.0925000011920929,"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":0.49230000376701355},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4546000063419342},{"id":"https://openalex.org/keywords/evolutionary-algorithm","display_name":"Evolutionary algorithm","score":0.43220001459121704},{"id":"https://openalex.org/keywords/evolutionary-computation","display_name":"Evolutionary computation","score":0.3871000111103058},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.35519999265670776},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.320499986410141}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7347000241279602},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5353999733924866},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.49230000376701355},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4546000063419342},{"id":"https://openalex.org/C159149176","wikidata":"https://www.wikidata.org/wiki/Q14489129","display_name":"Evolutionary algorithm","level":2,"score":0.43220001459121704},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3970000147819519},{"id":"https://openalex.org/C105902424","wikidata":"https://www.wikidata.org/wiki/Q1197129","display_name":"Evolutionary computation","level":2,"score":0.3871000111103058},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.37529999017715454},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.35519999265670776},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.320499986410141},{"id":"https://openalex.org/C129792486","wikidata":"https://www.wikidata.org/wiki/Q1050419","display_name":"Language identification","level":3,"score":0.2992999851703644},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.289900004863739},{"id":"https://openalex.org/C121835503","wikidata":"https://www.wikidata.org/wiki/Q2596288","display_name":"Evolutionary programming","level":3,"score":0.28110000491142273},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2734000086784363},{"id":"https://openalex.org/C106516650","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm design","level":2,"score":0.25870001316070557},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.25780001282691956}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/access.2026.3680045","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2026.3680045","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},{"id":"pmh:oai:doaj.org/article:5b1134b1df644dbd862caf9de7ad2d36","is_oa":true,"landing_page_url":"https://doaj.org/article/5b1134b1df644dbd862caf9de7ad2d36","pdf_url":null,"source":{"id":"https://openalex.org/S4306401280","display_name":"DOAJ (DOAJ: Directory of Open Access Journals)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-sa","license_id":"https://openalex.org/licenses/cc-by-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"IEEE Access, Vol 14, Pp 57033-57047 (2026)","raw_type":"article"}],"best_oa_location":{"id":"doi:10.1109/access.2026.3680045","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2026.3680045","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},"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":{"Large":[0],"language":[1,10],"models":[2,132],"(LLMs)":[3],"have":[4,30],"shown":[5],"outstanding":[6],"capabilities":[7],"across":[8],"complex":[9],"tasks,":[11],"but":[12],"come":[13],"with":[14,207],"massive":[15],"computational":[16],"cost":[17],"and":[18,28,42,45,49,70,99,117,130,133,156,190],"memory":[19,195],"footprint.":[20],"LLM":[21],"post-training":[22],"quantizers":[23,55],"such":[24],"as":[25],"AWQ,":[26],"SmoothQuant,":[27,138],"QoQ":[29],"demonstrated":[31],"excellent":[32],"performance":[33],"under":[34,105,168],"specific":[35],"configurations,":[36],"for":[37],"example,":[38],"W4A16":[39],"(four-bit":[40],"weight":[41,48],"sixteen-bit":[43],"activation)":[44],"W8A8":[46],"(eight-bit":[47],"eight-bit":[50],"activation).":[51],"However,":[52],"leveraging":[53],"these":[54],"in":[56,74],"a":[57,61,86,95,118],"mixed-precision":[58,88,159,204],"problem":[59],"is":[60],"nontrivial":[62],"task":[63],"due":[64],"to":[65,94,103,113,122,145,166],"the":[66,124,152],"colossal":[67],"design":[68],"space":[69],"prolonged":[71],"inference":[72],"time":[73],"LLMs.":[75],"To":[76],"address":[77],"this":[78,80],"problem,":[79],"work":[81],"presents":[82],"UDP":[83,109,140,177,201],"(Up-or-Down":[84],"Precision),":[85],"quantizer-agnostic":[87],"framework":[89],"that":[90],"restricts":[91],"each":[92],"layer":[93,125],"low/high":[96],"precision":[97],"pair":[98],"selects":[100],"which":[101],"layers":[102],"promote":[104],"deployment-driven":[106],"memory/latency":[107],"budgets.":[108,170],"uses":[110],"KLD-guided":[111],"profiling":[112],"seed":[114],"high-quality":[115],"candidates":[116],"budget-aware":[119],"genetic":[120],"search":[121],"optimize":[123],"promotion":[126],"set.":[127],"Across":[128],"Llama":[129],"Mistral":[131],"three":[134],"PTQ":[135],"backends":[136],"(AWQ,":[137],"QoQ),":[139],"improvesWikiText-2":[141],"perplexity":[142],"by":[143,164,187],"up":[144,165],"0.69":[146],"over":[147],"uniform":[148],"low-bit":[149],"baselines":[150],"at":[151],"same":[153],"low-precision":[154],"setting":[155],"outperforms":[157],"strong":[158],"heuristics":[160],"(FlattenQuant,":[161],"QFeM,":[162],"AsymKV)":[163],"0.11":[167],"identical":[169],"In":[171],"OmniServe":[172],"on":[173],"Llama-2-7B":[174],"(1024/512":[175],"tokens),":[176],"preserves":[178],"throughput":[179],"(\u22640.5%":[180],"vs.":[181],"W4A8)":[182],"while":[183],"reducing":[184],"p95":[185],"time-to-first-token":[186],"\u224860":[188],"ms":[189],"using":[191],"\u224810%":[192],"less":[193],"GPU":[194],"than":[196],"W8A8.":[197],"These":[198],"results":[199],"show":[200],"produces":[202],"practical":[203],"policies":[205],"aligned":[206],"real":[208],"serving":[209],"constraints.":[210]},"counts_by_year":[],"updated_date":"2026-04-21T06:05:55.347736","created_date":"2026-04-03T00:00:00"}