{"id":"https://openalex.org/W4411725788","doi":"https://doi.org/10.1109/iscas56072.2025.11043714","title":"A 4.21 TFLOPS/W Memory-Efficient LLM Inference Accelerator with Bit-Layered Non-Uniform Quantization","display_name":"A 4.21 TFLOPS/W Memory-Efficient LLM Inference Accelerator with Bit-Layered Non-Uniform Quantization","publication_year":2025,"publication_date":"2025-05-25","ids":{"openalex":"https://openalex.org/W4411725788","doi":"https://doi.org/10.1109/iscas56072.2025.11043714"},"language":"en","primary_location":{"id":"doi:10.1109/iscas56072.2025.11043714","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iscas56072.2025.11043714","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Symposium on Circuits and Systems (ISCAS)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"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/A5091616745","display_name":"Byeongcheol Kim","orcid":"https://orcid.org/0000-0003-1079-7220"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Byeongcheol Kim","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101426371","display_name":"Sang-Jin Kim","orcid":"https://orcid.org/0000-0001-9317-0479"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sangjin Kim","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083733804","display_name":"Sangwoo Ha","orcid":"https://orcid.org/0000-0002-9191-6326"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sangwoo Ha","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074016364","display_name":"Soyeon Um","orcid":"https://orcid.org/0000-0002-8526-2047"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Soyeon Um","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050861404","display_name":"Kyomin Sohn","orcid":"https://orcid.org/0000-0002-8094-9843"},"institutions":[{"id":"https://openalex.org/I2250650973","display_name":"Samsung (South Korea)","ror":"https://ror.org/04w3jy968","country_code":"KR","type":"company","lineage":["https://openalex.org/I2250650973"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Kyomin Sohn","raw_affiliation_strings":["Samsung Electronics,Memory Business Division,Hwaseong,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Samsung Electronics,Memory Business Division,Hwaseong,Republic of Korea","institution_ids":["https://openalex.org/I2250650973"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5108365621","display_name":"Hoi-Jun Yoo","orcid":null},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hoi-Jun Yoo","raw_affiliation_strings":["Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Korea Advanced Institute of Science and Technology (KAIST),Daejeon,Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"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":0.14923777,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"5"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10232","display_name":"Optical Network Technologies","score":0.9717000126838684,"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"}},"topics":[{"id":"https://openalex.org/T10232","display_name":"Optical Network Technologies","score":0.9717000126838684,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.9702000021934509,"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/T10299","display_name":"Photonic and Optical Devices","score":0.9589999914169312,"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/quantization","display_name":"Quantization (signal processing)","score":0.6203954815864563},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6162980198860168},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5913928151130676},{"id":"https://openalex.org/keywords/bit","display_name":"Bit (key)","score":0.5653336644172668},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.489420086145401},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.19940981268882751},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.18978992104530334}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6203954815864563},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6162980198860168},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5913928151130676},{"id":"https://openalex.org/C117011727","wikidata":"https://www.wikidata.org/wiki/Q1278488","display_name":"Bit (key)","level":2,"score":0.5653336644172668},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.489420086145401},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.19940981268882751},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.18978992104530334},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iscas56072.2025.11043714","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iscas56072.2025.11043714","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Symposium on Circuits and Systems (ISCAS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy","score":0.46000000834465027}],"awards":[],"funders":[{"id":"https://openalex.org/F4320322202","display_name":"IC Design Education Center","ror":"https://ror.org/005v57z85"},{"id":"https://openalex.org/F4320332195","display_name":"Samsung","ror":"https://ror.org/04w3jy968"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":14,"referenced_works":["https://openalex.org/W2044373032","https://openalex.org/W2060108852","https://openalex.org/W2921918777","https://openalex.org/W3100985894","https://openalex.org/W3134750800","https://openalex.org/W4308089490","https://openalex.org/W4313377427","https://openalex.org/W4360606017","https://openalex.org/W4391892592","https://openalex.org/W4393407021","https://openalex.org/W4400230909","https://openalex.org/W4401568853","https://openalex.org/W4406650295","https://openalex.org/W4411403346"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W4327546585","https://openalex.org/W2411923897","https://openalex.org/W4394546135","https://openalex.org/W4285347720","https://openalex.org/W4200259850","https://openalex.org/W2333831899","https://openalex.org/W2484894494"],"abstract_inverted_index":{"Non-uniform":[0],"Quantization":[1],"(NUQ)":[2],"is":[3],"widely":[4],"used":[5],"in":[6,83,98],"LLM":[7,149],"accelerators":[8],"due":[9],"to":[10,125],"its":[11],"high":[12],"accuracy.":[13],"However,":[14],"employing":[15],"NUQ":[16,35,85],"with":[17],"models":[18,132],"of":[19,68,103,119,133],"varying":[20],"sizes":[21],"can":[22],"substantially":[23],"increase":[24],"storage":[25,117],"requirements":[26],"on":[27],"mobile":[28,148],"devices.":[29],"This":[30],"paper":[31],"presents":[32],"a":[33,73],"bit-layered":[34],"accelerator":[36],"architecture":[37],"that":[38],"supports":[39,131],"multiple":[40,69],"bit-width":[41],"configurations":[42],"while":[43],"minimizing":[44],"memory":[45,141],"usage.":[46],"Key":[47],"features":[48],"include":[49],"Reconfigurable":[50],"Condensed":[51],"Look-up":[52],"Accumulator":[53],"(RCLA),":[54],"Dual-Sign":[55],"Path":[56],"Accumulation":[57],"(DSPA),":[58],"and":[59,139],"MSB-Sparse":[60],"Encoding":[61],"Compression":[62],"(MSEC).":[63],"RCLA":[64],"enables":[65],"the":[66,115],"use":[67],"weight":[70,108,112],"precisions":[71],"within":[72],"uniform":[74],"PE":[75,81],"array.":[76],"In":[77],"particular,":[78],"it":[79,145],"optimizes":[80],"utilization":[82],"high-bit-width":[84],"modes,":[86],"reducing":[87,114],"accumulation":[88],"cycles":[89],"by":[90,123],"63.2":[91],"%.":[92,127],"DSPA":[93],"facilitates":[94],"energy-efficient":[95],"computation,":[96],"resulting":[97],"an":[99],"average":[100],"power":[101],"reduction":[102],"40.7":[104],"%":[105],"across":[106],"various":[107],"modes.":[109],"MSEC":[110],"enhances":[111],"compression,":[113],"data":[116],"size":[118],"each":[120],"bit":[121],"plane":[122],"up":[124],"47.7":[126],"The":[128],"proposed":[129],"design":[130],"different":[134],"sizes,":[135],"improves":[136],"energy":[137],"efficiency,":[138],"reduces":[140],"capacity":[142],"requirements,":[143],"rendering":[144],"ideal":[146],"for":[147],"inference.":[150]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}
