{"id":"https://openalex.org/W4389523971","doi":"https://doi.org/10.18653/v1/2023.emnlp-main.910","title":"Enhancing Computation Efficiency in Large Language Models through Weight and Activation Quantization","display_name":"Enhancing Computation Efficiency in Large Language Models through Weight and Activation Quantization","publication_year":2023,"publication_date":"2023-01-01","ids":{"openalex":"https://openalex.org/W4389523971","doi":"https://doi.org/10.18653/v1/2023.emnlp-main.910"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2023.emnlp-main.910","doi":"https://doi.org/10.18653/v1/2023.emnlp-main.910","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2023.emnlp-main.910","pdf_url":"https://aclanthology.org/2023.emnlp-main.910.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2023.emnlp-main.910.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5036363860","display_name":"Janghwan Lee","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4575257","display_name":"Hanyang University","ror":"https://ror.org/046865y68","country_code":"KR","type":"education","lineage":["https://openalex.org/I4575257"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Janghwan Lee","raw_affiliation_strings":["Hanyang University,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hanyang University,","institution_ids":["https://openalex.org/I4575257"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100362650","display_name":"Minsoo Kim","orcid":"https://orcid.org/0000-0001-7056-1952","observed_orcids":["https://orcid.org/0000-0001-7056-1952"]},"institutions":[{"id":"https://openalex.org/I4575257","display_name":"Hanyang University","ror":"https://ror.org/046865y68","country_code":"KR","type":"education","lineage":["https://openalex.org/I4575257"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Minsoo Kim","raw_affiliation_strings":["Hanyang University,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hanyang University,","institution_ids":["https://openalex.org/I4575257"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101606377","display_name":"Seung-Cheol Baek","orcid":"https://orcid.org/0009-0007-1624-4586","observed_orcids":["https://orcid.org/0009-0007-1624-4586"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Seungcheol Baek","raw_affiliation_strings":["SAPEON Korea Inc.,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SAPEON Korea Inc.,","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111629569","display_name":"Seok Yoon Hwang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Seok Hwang","raw_affiliation_strings":["SAPEON Korea Inc.,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SAPEON Korea Inc.,","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113491293","display_name":"Wonyong Sung","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Wonyong Sung","raw_affiliation_strings":["Seoul National University Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Seoul National University Republic of Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5089720016","display_name":"Jungwook Choi","orcid":"https://orcid.org/0000-0001-5916-9714","observed_orcids":["https://orcid.org/0000-0001-5916-9714","https://orcid.org/0009-0007-8208-3784"]},"institutions":[{"id":"https://openalex.org/I4575257","display_name":"Hanyang University","ror":"https://ror.org/046865y68","country_code":"KR","type":"education","lineage":["https://openalex.org/I4575257"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jungwook Choi","raw_affiliation_strings":["Hanyang University,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hanyang University,","institution_ids":["https://openalex.org/I4575257"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.26,"has_fulltext":true,"cited_by_count":14,"citation_normalized_percentile":{"value":0.88327701,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"14726","last_page":"14739"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9384999871253967,"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":0.9384999871253967,"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/T10363","display_name":"Low-power high-performance VLSI design","score":0.014399999752640724,"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.01209999993443489,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/large-language-models","display_name":"large language models","score":0.9589999914169312},{"id":"https://openalex.org/keywords/post-training-quantization","display_name":"post-training quantization","score":0.9919999837875366},{"id":"https://openalex.org/keywords/weight-and-activation-quantization","display_name":"weight and activation quantization","score":0.8619999885559082},{"id":"https://openalex.org/keywords/underflow","display_name":"underflow","score":0.5440000295639038},{"id":"https://openalex.org/keywords/arithmetic-units","display_name":"arithmetic units","score":0.5799999833106995},{"id":"https://openalex.org/keywords/hardware-efficiency","display_name":"hardware efficiency","score":0.7549999952316284},{"id":"https://openalex.org/keywords/computational-efficiency","display_name":"computational efficiency","score":0.7239999771118164}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8686478137969971},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6477718353271484},{"id":"https://openalex.org/C3265923","wikidata":"https://www.wikidata.org/wiki/Q669129","display_name":"Arithmetic underflow","level":2,"score":0.5640478134155273},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.5469451546669006},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5030414462089539},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.4870010018348694},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.47065725922584534},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.39698055386543274},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.38486406207084656},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.33430033922195435},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.31217706203460693},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.23203936219215393},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.20880606770515442}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2023.emnlp-main.910","doi":"https://doi.org/10.18653/v1/2023.emnlp-main.910","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2023.emnlp-main.910","pdf_url":"https://aclanthology.org/2023.emnlp-main.910.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2311.05161","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2311.05161","pdf_url":"https://arxiv.org/pdf/2311.05161","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2311.05161","doi":"https://doi.org/10.48550/arxiv.2311.05161","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2311.05161","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,"listed_in":[],"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"}],"best_oa_location":{"id":"doi:10.18653/v1/2023.emnlp-main.910","doi":"https://doi.org/10.18653/v1/2023.emnlp-main.910","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2023.emnlp-main.910","pdf_url":"https://aclanthology.org/2023.emnlp-main.910.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2320816490","display_name":null,"funder_award_id":"1415178807","funder_id":"https://openalex.org/F4320321681","funder_display_name":"Ministry of Trade, Industry and Energy"},{"id":"https://openalex.org/G3200376425","display_name":null,"funder_award_id":"2020-0-01305","funder_id":"https://openalex.org/F4320335489","funder_display_name":"Institute for Information and Communications Technology Promotion"},{"id":"https://openalex.org/G7079512917","display_name":null,"funder_award_id":"2020-0-01297","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"},{"id":"https://openalex.org/G8878611666","display_name":null,"funder_award_id":"2020-0-01305","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"}],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"},{"id":"https://openalex.org/F4320321681","display_name":"Ministry of Trade, Industry and Energy","ror":"https://ror.org/008nkqk13"},{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"},{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"},{"id":"https://openalex.org/F4320335489","display_name":"Institute for Information and Communications Technology Promotion","ror":"https://ror.org/01g0hqq23"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4389523971.pdf","grobid_xml":"https://content.openalex.org/works/W4389523971.grobid-xml"},"referenced_works_count":33,"referenced_works":["https://openalex.org/W2525332836","https://openalex.org/W2794325560","https://openalex.org/W2970821029","https://openalex.org/W2996908057","https://openalex.org/W2998617917","https://openalex.org/W3020212829","https://openalex.org/W3030163527","https://openalex.org/W3100710793","https://openalex.org/W3101100041","https://openalex.org/W3118781290","https://openalex.org/W3121904249","https://openalex.org/W3202028501","https://openalex.org/W3204625459","https://openalex.org/W3210580311","https://openalex.org/W4221148635","https://openalex.org/W4224308101","https://openalex.org/W4226539834","https://openalex.org/W4229005866","https://openalex.org/W4283313765","https://openalex.org/W4303443398","https://openalex.org/W4307934016","https://openalex.org/W4309591680","https://openalex.org/W4322718191","https://openalex.org/W4327810158","https://openalex.org/W4375868853","https://openalex.org/W4378770729","https://openalex.org/W4379260375","https://openalex.org/W4379548477","https://openalex.org/W4385574133","https://openalex.org/W4387321091","https://openalex.org/W4393147284","https://openalex.org/W7133198810","https://openalex.org/W7133231030"],"related_works":["https://openalex.org/W4386231419","https://openalex.org/W4386357215","https://openalex.org/W4309591680","https://openalex.org/W4412888573","https://openalex.org/W4404390875","https://openalex.org/W4389500673","https://openalex.org/W4404783494","https://openalex.org/W7166810475","https://openalex.org/W4403754519","https://openalex.org/W4396913356"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"are":[4,105],"proficient":[5],"in":[6,30,99],"natural":[7],"language":[8],"processing":[9],"tasks,":[10],"but":[11],"their":[12],"deployment":[13],"is":[14],"often":[15],"restricted":[16],"by":[17,65],"extensive":[18],"parameter":[19],"sizes":[20],"and":[21,35,58,72,74,91,115],"computational":[22,42],"demands.This":[23],"paper":[24],"focuses":[25],"on":[26,70],"post-training":[27],"quantization":[28],"(PTQ)":[29],"LLMs,":[31,112],"specifically":[32],"4-bit":[33],"weight":[34],"8-bit":[36,152],"activation":[37],"(W4A8)":[38],"quantization,":[39,101],"to":[40,48,62,79,94,107,126,151],"enhance":[41,63],"efficiency-a":[43],"topic":[44],"less":[45],"explored":[46],"compared":[47,150],"weight-only":[49],"quantization.We":[50],"present":[51],"two":[52],"innovative":[53],"techniques:":[54],"activation-quantization-aware":[55],"scaling":[56],"(AQAS)":[57],"sequence-length-aware":[59],"calibration":[60,76],"(SLAC)":[61],"PTQ":[64],"considering":[66],"the":[67,96],"combined":[68],"effects":[69],"weights":[71],"activations":[73],"aligning":[75],"sequence":[77],"lengths":[78],"target":[80],"tasks.Moreover,":[81],"we":[82,117,138],"introduce":[83],"dINT,":[84,137],"a":[85,145],"hybrid":[86],"data":[87],"format":[88],"combining":[89],"integer":[90,153],"denormal":[92],"representations,":[93],"address":[95],"underflow":[97],"issue":[98],"W4A8":[100],"where":[102],"small":[103],"values":[104],"rounded":[106],"zero.Through":[108],"rigorous":[109],"evaluations":[110],"of":[111],"including":[113],"OPT":[114],"LLaMA,":[116],"demonstrate":[118],"that":[119,141],"our":[120,142],"techniques":[121],"significantly":[122],"boost":[123],"task":[124],"accuracies":[125],"levels":[127],"comparable":[128],"with":[129,136],"full-precision":[130],"models.By":[131],"developing":[132],"arithmetic":[133],"units":[134],"compatible":[135],"further":[139],"confirm":[140],"methods":[143],"yield":[144],"2\u00d7":[146],"hardware":[147],"efficiency":[148],"improvement":[149],"MAC":[154],"unit.":[155]},"counts_by_year":[{"year":2026,"cited_by_count":5},{"year":2025,"cited_by_count":6},{"year":2024,"cited_by_count":3}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2023-12-11T00:00:00"}