{"id":"https://openalex.org/W7202132339","doi":"https://doi.org/10.1145/3816440.3818527","title":"L 2 Mersit: A Scaling-Free Sub-8-bit Data Format for On-Device Reliable Large Language Model Serving","display_name":"L 2 Mersit: A Scaling-Free Sub-8-bit Data Format for On-Device Reliable Large Language Model Serving","publication_year":2026,"publication_date":"2026-08-05","ids":{"openalex":"https://openalex.org/W7202132339","doi":"https://doi.org/10.1145/3816440.3818527"},"language":null,"primary_location":{"id":"doi:10.1145/3816440.3818527","doi":"https://doi.org/10.1145/3816440.3818527","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3816440.3818527","pdf_url":null,"source":{"id":"https://openalex.org/S7407086783","display_name":"ACM/IEEE International Symposium on Low Power Electronics and Design (ISLPED)","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-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM/IEEE International Symposium on Low Power Electronics and Design","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1145/3816440.3818527","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5146397701","display_name":"Myeongjin Kim","orcid":"https://orcid.org/0009-0004-2816-3386","observed_orcids":["https://orcid.org/0009-0004-2816-3386"]},"institutions":[{"id":"https://openalex.org/I35928602","display_name":"Kyung Hee University","ror":"https://ror.org/01zqcg218","country_code":"KR","type":"education","lineage":["https://openalex.org/I35928602"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Myeongjin Kim","raw_affiliation_strings":["Kyung Hee University, Yongin-si, Gyeonggi-do, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0004-2816-3386","affiliations":[{"raw_affiliation_string":"Kyung Hee University, Yongin-si, Gyeonggi-do, Republic of Korea","institution_ids":["https://openalex.org/I35928602"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103126454","display_name":"Hyeonseong Kim","orcid":"https://orcid.org/0000-0002-9160-8876","observed_orcids":["https://orcid.org/0000-0002-9160-8876","https://orcid.org/0009-0003-9218-2795"]},"institutions":[{"id":"https://openalex.org/I35928602","display_name":"Kyung Hee University","ror":"https://ror.org/01zqcg218","country_code":"KR","type":"education","lineage":["https://openalex.org/I35928602"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hyeonseong Kim","raw_affiliation_strings":["Kyung Hee University, Yongin-si, Gyeonggi-do, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0003-9218-2795","affiliations":[{"raw_affiliation_string":"Kyung Hee University, Yongin-si, Gyeonggi-do, Republic of Korea","institution_ids":["https://openalex.org/I35928602"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045028723","display_name":"Ik\u2010Joon Chang","orcid":"https://orcid.org/0000-0002-8871-8695","observed_orcids":["https://orcid.org/0000-0002-8871-8695"]},"institutions":[{"id":"https://openalex.org/I35928602","display_name":"Kyung Hee University","ror":"https://ror.org/01zqcg218","country_code":"KR","type":"education","lineage":["https://openalex.org/I35928602"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Ik Joon Chang","raw_affiliation_strings":["Kyung Hee University, Yongin-si, Gyeonggi-do, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0002-8871-8695","affiliations":[{"raw_affiliation_string":"Kyung Hee University, Yongin-si, Gyeonggi-do, Republic of Korea","institution_ids":["https://openalex.org/I35928602"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5034558336","display_name":"Seungkyu Choi","orcid":"https://orcid.org/0000-0002-3125-9707","observed_orcids":["https://orcid.org/0000-0002-3125-9707","https://orcid.org/0009-0007-3006-1487"]},"institutions":[{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Seungkyu Choi","raw_affiliation_strings":["Yonsei University, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0002-3125-9707","affiliations":[{"raw_affiliation_string":"Yonsei University, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I193775966"]}]}],"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":0.75267142,"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":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11697","display_name":"Numerical Methods and Algorithms","score":0.9887999892234802,"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"}},"topics":[{"id":"https://openalex.org/T11697","display_name":"Numerical Methods and Algorithms","score":0.9887999892234802,"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"}},{"id":"https://openalex.org/T10363","display_name":"Low-power high-performance VLSI design","score":0.008899999782443047,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.0003000000142492354,"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/dual-mode-operation","display_name":"dual-mode operation","score":0.6359999775886536},{"id":"https://openalex.org/keywords/perplexity","display_name":"perplexity","score":0.5540000200271606},{"id":"https://openalex.org/keywords/hardware-area-reduction","display_name":"hardware area reduction","score":0.6539999842643738},{"id":"https://openalex.org/keywords/activation-distribution","display_name":"activation distribution","score":0.49300000071525574}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6802999973297119},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3336000144481659},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32839998602867126},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.31279999017715454},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.30889999866485596},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.30239999294281006},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.2957000136375427},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.2921999990940094},{"id":"https://openalex.org/C2985331491","wikidata":"https://www.wikidata.org/wiki/Q5227298","display_name":"Data format","level":2,"score":0.29190000891685486},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.28610000014305115},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.2791999876499176},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.27140000462532043},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2596000134944916},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.25189998745918274},{"id":"https://openalex.org/C45235069","wikidata":"https://www.wikidata.org/wiki/Q278425","display_name":"Table (database)","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3816440.3818527","doi":"https://doi.org/10.1145/3816440.3818527","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3816440.3818527","pdf_url":null,"source":{"id":"https://openalex.org/S7407086783","display_name":"ACM/IEEE International Symposium on Low Power Electronics and Design (ISLPED)","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-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM/IEEE International Symposium on Low Power Electronics and Design","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3816440.3818527","doi":"https://doi.org/10.1145/3816440.3818527","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3816440.3818527","pdf_url":null,"source":{"id":"https://openalex.org/S7407086783","display_name":"ACM/IEEE International Symposium on Low Power Electronics and Design (ISLPED)","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-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM/IEEE International Symposium on Low Power Electronics and Design","raw_type":"proceedings-article"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":3,"referenced_works":["https://openalex.org/W4233996382","https://openalex.org/W4404133534","https://openalex.org/W4404133682"],"related_works":[],"abstract_inverted_index":{"On-device":[0],"large":[1],"language":[2],"model":[3],"(LLM)":[4],"serving":[5],"drives":[6],"low-precision":[7,101],"computing":[8],"to":[9,48],"address":[10],"memory":[11],"and":[12,42,61,68,75],"compute":[13],"limits.":[14],"This":[15],"paper":[16],"presents":[17],"L2Mersit,":[18],"a":[19,105],"scaling-free,":[20],"range-adjustable":[21],"exponent-encoded":[22,90],"data":[23],"format":[24],"tailored":[25],"for":[26,100],"sub-8-bit":[27],"LLM":[28],"quantization.":[29],"Building":[30],"upon":[31],"the":[32,84,94],"Mersit":[33],"framework,":[34],"L2Mersit":[35,82],"employs":[36],"dual":[37],"mode":[38],"operation,":[39],"comprising":[40],"range-expanded":[41],"precision-enhanced":[43],"modes":[44],"that":[45,81],"dynamically":[46],"adapt":[47],"activation":[49],"distributions":[50],"with":[51],"minimal":[52],"control":[53],"overhead.":[54],"The":[55],"proposed":[56],"design":[57],"eliminates":[58],"on-the-fly":[59],"scaling":[60],"auxiliary":[62,98],"computations":[63],"while":[64,92],"effectively":[65],"preserving":[66],"range":[67],"precision,":[69],"thereby":[70],"achieving":[71],"both":[72],"superior":[73],"perplexity":[74],"hardware":[76,95],"efficiency.":[77],"Experimental":[78],"results":[79],"demonstrate":[80],"achieves":[83],"highest":[85],"accuracy":[86],"among":[87],"all":[88],"6-bit":[89],"formats":[91],"reducing":[93],"complexity":[96],"of":[97],"units":[99],"computing,":[102],"resulting":[103],"in":[104],"62.7%":[106],"area":[107],"reduction.":[108]},"counts_by_year":[],"updated_date":"2026-10-05T09:50:58.205359","created_date":"2026-08-11T00:00:00"}