{"id":"https://openalex.org/W4416429667","doi":"https://doi.org/10.1109/iccad66269.2025.11240926","title":"OA-LAMA: An Outlier-Adaptive LLM Inference Accelerator with Memory-Aligned Mixed-Precision Group Quantization","display_name":"OA-LAMA: An Outlier-Adaptive LLM Inference Accelerator with Memory-Aligned Mixed-Precision Group Quantization","publication_year":2025,"publication_date":"2025-10-26","ids":{"openalex":"https://openalex.org/W4416429667","doi":"https://doi.org/10.1109/iccad66269.2025.11240926"},"language":null,"primary_location":{"id":"doi:10.1109/iccad66269.2025.11240926","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240926","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","raw_type":"proceedings-article"},"type":"article","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/A5081375882","display_name":"Huangxu Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Huangxu Chen","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou)"],"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou)","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100732976","display_name":"Yanpeng Hao","orcid":"https://orcid.org/0000-0002-7938-2835"},"institutions":[{"id":"https://openalex.org/I90610280","display_name":"South China University of Technology","ror":"https://ror.org/0530pts50","country_code":"CN","type":"education","lineage":["https://openalex.org/I90610280"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yingbo Hao","raw_affiliation_strings":["South China University of Technology"],"affiliations":[{"raw_affiliation_string":"South China University of Technology","institution_ids":["https://openalex.org/I90610280"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100458990","display_name":"Yi Zou","orcid":"https://orcid.org/0000-0002-4382-4670"},"institutions":[{"id":"https://openalex.org/I90610280","display_name":"South China University of Technology","ror":"https://ror.org/0530pts50","country_code":"CN","type":"education","lineage":["https://openalex.org/I90610280"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yi Zou","raw_affiliation_strings":["South China University of Technology"],"affiliations":[{"raw_affiliation_string":"South China University of Technology","institution_ids":["https://openalex.org/I90610280"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5107928490","display_name":"Xinyu Chen","orcid":"https://orcid.org/0000-0002-4589-4111"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinyu Chen","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou)"],"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou)","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5081375882"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.3568,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.91896459,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"9"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.2574000060558319,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.2574000060558319,"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":0.2207999974489212,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.10429999977350235,"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":0.7419000267982483},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.6614000201225281},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5228000283241272},{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.5004000067710876},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.4577000141143799},{"id":"https://openalex.org/keywords/architecture","display_name":"Architecture","score":0.38670000433921814},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.3637999892234802}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7419000267982483},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7378000020980835},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.6614000201225281},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.5435000061988831},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5228000283241272},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.5004000067710876},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.4577000141143799},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3952000141143799},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.38670000433921814},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.3637999892234802},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.3626999855041504},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.35670000314712524},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.353300005197525},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.34540000557899475},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.32659998536109924},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3244999945163727},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.2985999882221222},{"id":"https://openalex.org/C137635306","wikidata":"https://www.wikidata.org/wiki/Q182667","display_name":"Pareto principle","level":2,"score":0.2985000014305115},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.2809999883174896},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.2750999927520752},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.2671000063419342},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2531999945640564},{"id":"https://openalex.org/C65232700","wikidata":"https://www.wikidata.org/wiki/Q5656403","display_name":"Hardware architecture","level":3,"score":0.25099998712539673}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iccad66269.2025.11240926","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240926","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":17,"referenced_works":["https://openalex.org/W2606722458","https://openalex.org/W2883920103","https://openalex.org/W3100985894","https://openalex.org/W3133253223","https://openalex.org/W4242118294","https://openalex.org/W4242577057","https://openalex.org/W4308083739","https://openalex.org/W4366341968","https://openalex.org/W4385245566","https://openalex.org/W4387321091","https://openalex.org/W4393147284","https://openalex.org/W4401211807","https://openalex.org/W4404133534","https://openalex.org/W4404133703","https://openalex.org/W4404783683","https://openalex.org/W4414281281","https://openalex.org/W4415799217"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"face":[4],"significant":[5],"deployment":[6],"challenges":[7],"due":[8],"to":[9,103,119,141],"their":[10],"substantial":[11],"memory":[12,40,109],"and":[13,57,137,164],"computational":[14],"demands.":[15],"While":[16],"low-precision":[17],"quantization":[18,44,157],"offers":[19],"a":[20,63,96,112,133],"promising":[21],"solution,":[22],"the":[23,128,143],"presence":[24],"of":[25],"activation":[26],"outliers":[27],"severely":[28],"degrades":[29],"model":[30],"accuracy.":[31],"Existing":[32],"approaches":[33],"either":[34],"compromise":[35],"hardware":[36,130],"efficiency":[37,167],"through":[38,46],"misaligned":[39],"access":[41],"or":[42],"sacrifice":[43],"granularity":[45],"rigid":[47],"bit-width":[48],"allocation,":[49],"particularly":[50],"when":[51],"handling":[52],"non-uniform":[53],"tensor":[54],"distributions":[55],"across":[56],"within":[58],"layers.":[59],"This":[60],"paper":[61],"presents":[62],"hardware-software":[64],"co-designed":[65],"framework":[66,82],"resulting":[67],"in":[68,178],"an":[69,88],"outlier-adaptive":[70,89],"LLM":[71,171,182],"inference":[72],"accelerator":[73],"with":[74,95,132],"memory-aligned":[75,90],"mixed-precision":[76,91],"group":[77,92,114],"quantization,":[78],"named":[79],"OA-LAMA.":[80],"The":[81],"comprises":[83],"three":[84],"key":[85],"innovations:":[86],"First,":[87],"(OAMAG)":[93],"format":[94,145],"novel":[97],"outlier":[98,122],"reordering":[99],"technique":[100],"is":[101,117,185],"proposed":[102,118],"preserve":[104],"accuracy":[105,153],"while":[106,159],"maintaining":[107],"DRAM-aligned":[108],"access.":[110],"Second,":[111],"distribution-aware":[113],"allocation":[115],"strategy":[116],"address":[120],"inter-layer":[121],"ratio":[123],"variance.":[124],"Finally,":[125],"we":[126],"design":[127],"OA-LAMA":[129,150,173,184],"architecture":[131,136],"three-level":[134],"accumulation":[135],"timing-balanced":[138],"processing":[139],"elements":[140],"support":[142],"OAMAG":[144],"efficiently.":[146],"Evaluations":[147],"demonstrate":[148],"that":[149],"achieves":[151],"better":[152],"than":[154],"state-of-the-art":[155],"4-bit":[156],"methods":[158],"delivering":[160],"1.21\u20133.09\u00d7":[161],"performance":[162],"improvement":[163],"1.35\u20132.47\u00d7":[165],"energy":[166],"gains":[168],"over":[169],"leading":[170],"accelerators.":[172],"establishes":[174],"new":[175],"Pareto":[176],"frontiers":[177],"accuracy-efficiency":[179],"co-optimization":[180],"for":[181],"inference.":[183],"open-sourced":[186],"at":[187],"https://github.com/CLab-HKUST-GZ/ICCAD25_OA-LAMA.git.":[188]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-04-09T08:11:56.329763","created_date":"2025-11-20T00:00:00"}