{"id":"https://openalex.org/W4417452953","doi":"https://doi.org/10.1109/iscas66217.2026.11562300","title":"SeVeDo: A Heterogeneous Transformer Accelerator for Low-Bit Inference via Hierarchical Group Quantization and SVD-Guided Mixed Precision","display_name":"SeVeDo: A Heterogeneous Transformer Accelerator for Low-Bit Inference via Hierarchical Group Quantization and SVD-Guided Mixed Precision","publication_year":2026,"publication_date":"2026-05-24","ids":{"openalex":"https://openalex.org/W4417452953","doi":"https://doi.org/10.1109/iscas66217.2026.11562300"},"language":null,"primary_location":{"id":"doi:10.1109/iscas66217.2026.11562300","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iscas66217.2026.11562300","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE International Symposium on Circuits and Systems (ISCAS)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2512.12930","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5033586441","display_name":"Yuseon Choi","orcid":"https://orcid.org/0009-0008-0698-8116"},"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":"Yuseon Choi","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/A5100748581","display_name":"Sangjin Kim","orcid":"https://orcid.org/0000-0001-6665-9973"},"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/A5055502318","display_name":"Jun-Ho Oh","orcid":"https://orcid.org/0000-0002-2609-6712"},"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":"Jungjun Oh","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/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":"last","author":{"id":"https://openalex.org/A5077896259","display_name":"Hoi\u2010Jun Yoo","orcid":"https://orcid.org/0000-0002-6661-4879"},"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":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I157485424"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"539","last_page":"543"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.1573999971151352,"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":0.1573999971151352,"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.10809999704360962,"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/T10500","display_name":"Sparse and Compressive Sensing Techniques","score":0.052799999713897705,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"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":0.7271000146865845},{"id":"https://openalex.org/keywords/residual","display_name":"Residual","score":0.6055999994277954},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6029000282287598},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5963000059127808},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5742999911308289},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.5386000275611877},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.4584999978542328}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7271000146865845},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.621999979019165},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.6055999994277954},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6029000282287598},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5963000059127808},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5742999911308289},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5530999898910522},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.5386000275611877},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.4584999978542328},{"id":"https://openalex.org/C2781198647","wikidata":"https://www.wikidata.org/wiki/Q1633673","display_name":"Datapath","level":2,"score":0.4375},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.32190001010894775},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3138999938964844},{"id":"https://openalex.org/C2781311116","wikidata":"https://www.wikidata.org/wiki/Q83306","display_name":"Group (periodic table)","level":2,"score":0.2912999987602234},{"id":"https://openalex.org/C73301696","wikidata":"https://www.wikidata.org/wiki/Q5469984","display_name":"Formalism (music)","level":3,"score":0.28780001401901245},{"id":"https://openalex.org/C135402231","wikidata":"https://www.wikidata.org/wiki/Q898440","display_name":"Dissipation","level":2,"score":0.2865999937057495},{"id":"https://openalex.org/C134835016","wikidata":"https://www.wikidata.org/wiki/Q690265","display_name":"Lookup table","level":2,"score":0.2612000107765198},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2526000142097473}],"mesh":[],"locations_count":4,"locations":[{"id":"doi:10.1109/iscas66217.2026.11562300","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iscas66217.2026.11562300","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE International Symposium on Circuits and Systems (ISCAS)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2512.12930","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.12930","pdf_url":"https://arxiv.org/pdf/2512.12930","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"pmh:oai:arXiv.org:2512.12930","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2512.12930","pdf_url":"https://arxiv.org/pdf/2512.12930","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,"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.2512.12930","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2512.12930","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2512.12930","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.12930","pdf_url":"https://arxiv.org/pdf/2512.12930","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4417452953.pdf","grobid_xml":"https://content.openalex.org/works/W4417452953.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Low-bit":[0],"quantization":[1],"is":[2],"a":[3,63,74,127],"promising":[4],"technique":[5],"for":[6],"efficient":[7],"transformer":[8],"inference":[9],"by":[10],"reducing":[11,97],"computational":[12],"and":[13,35,140],"memory":[14],"overhead.":[15],"However,":[16],"aggressive":[17],"bitwidth":[18],"reduction":[19],"remains":[20],"challenging":[21],"due":[22],"to":[23,27,109],"activation":[24],"outliers,":[25],"leading":[26],"accuracy":[28,40],"degradation.":[29],"Existing":[30],"methods,":[31],"such":[32],"as":[33],"outlier-handling":[34],"group":[36,79],"quantization,":[37],"achieve":[38],"high":[39],"but":[41],"incur":[42],"substantial":[43],"energy":[44,129],"consumption.":[45],"To":[46,81],"address":[47],"this,":[48],"we":[49],"propose":[50],"SeVeDo,":[51],"an":[52],"energy-efficient":[53],"SVD-based":[54],"heterogeneous":[55],"accelerator":[56],"that":[57,124],"structurally":[58],"separates":[59],"outlier-sensitive":[60],"components":[61,111],"into":[62],"high-precision":[64],"low-rank":[65,114],"path,":[66],"while":[67],"the":[68],"remaining":[69],"computations":[70],"are":[71],"executed":[72],"in":[73],"low-bit":[75],"residual":[76],"datapath":[77],"with":[78,93,136],"quantization.":[80],"further":[82],"enhance":[83],"efficiency,":[84],"Hierarchical":[85],"Group":[86],"Quantization":[87],"(HGQ)":[88],"combines":[89],"coarse-grained":[90],"floating-point":[91,118],"scaling":[92],"fine-grained":[94],"shifting,":[95],"effectively":[96],"dequantization":[98],"cost.":[99,120],"Also,":[100],"SVD-guided":[101],"mixed":[102],"precision":[103],"(SVD-MP)":[104],"statically":[105],"allocates":[106],"higher":[107],"bitwidths":[108],"precision-sensitive":[110],"identified":[112],"through":[113],"decomposition,":[115],"thereby":[116],"minimizing":[117],"operation":[119],"Experimental":[121],"results":[122],"show":[123],"SeVeDo":[125],"achieves":[126],"peak":[128],"efficiency":[130],"of":[131],"13.8TOPS/W,":[132],"surpassing":[133],"conventional":[134],"designs,":[135],"12.7TOPS/W":[137],"on":[138,142],"ViT-Base":[139],"13.4TOPS/W":[141],"Llama2-7B":[143],"benchmarks.":[144]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2025-12-17T00:00:00"}