{"id":"https://openalex.org/W7161147056","doi":"https://doi.org/10.1109/dcc66757.2026.00082","title":"Pushing Bit-Width Limits in LLM Quantization with Saliency-Guided Mix-Precision Allocation and Learnable Affine Transformation","display_name":"Pushing Bit-Width Limits in LLM Quantization with Saliency-Guided Mix-Precision Allocation and Learnable Affine Transformation","publication_year":2026,"publication_date":"2026-03-24","ids":{"openalex":"https://openalex.org/W7161147056","doi":"https://doi.org/10.1109/dcc66757.2026.00082"},"language":null,"primary_location":{"id":"doi:10.1109/dcc66757.2026.00082","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dcc66757.2026.00082","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 Data Compression Conference (DCC)","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/A5100581025","display_name":"Shuoyu Ma","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Shuoyu Ma","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Electronic Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Electronic Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136148866","display_name":"Wenrui Dai","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wenrui Dai","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Electronic Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Electronic Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5086886949","display_name":"Maida Cao","orcid":"https://orcid.org/0000-0002-9275-2095"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Maida Cao","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Computer Science & Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Computer Science & Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033205427","display_name":"S B Li","orcid":null},"institutions":[{"id":"https://openalex.org/I76130692","display_name":"Zhejiang University","ror":"https://ror.org/00a2xv884","country_code":"CN","type":"education","lineage":["https://openalex.org/I76130692"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shaohui Li","raw_affiliation_strings":["College of Information Science and Electronic Engineering, Zhejiang University,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"College of Information Science and Electronic Engineering, Zhejiang University,China","institution_ids":["https://openalex.org/I76130692"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5073652872","display_name":"Ziyang Zheng","orcid":"https://orcid.org/0000-0001-9923-8016"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ziyang Zheng","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Electronic Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Electronic Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019072920","display_name":"C. Li","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chenglin Li","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Electronic Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Electronic Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016006337","display_name":"Junni Zou","orcid":"https://orcid.org/0000-0002-9694-9880"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Junni Zou","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Computer Science & Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Computer Science & Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5136114727","display_name":"Hongkai Xiong","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hongkai Xiong","raw_affiliation_strings":["Shanghai Jiao Tong University,Department of Electronic Engineering,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,Department of Electronic Engineering,China","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5100581025"],"corresponding_institution_ids":["https://openalex.org/I183067930"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.94456665,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"455","last_page":"455"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.11879999935626984,"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":0.11879999935626984,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.1120000034570694,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.0617000013589859,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/affine-transformation","display_name":"Affine transformation","score":0.5242999792098999},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.3874000012874603},{"id":"https://openalex.org/keywords/transformation","display_name":"Transformation (genetics)","score":0.3149999976158142},{"id":"https://openalex.org/keywords/affine-combination","display_name":"Affine combination","score":0.3070000112056732},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.2953999936580658},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.2822999954223633}],"concepts":[{"id":"https://openalex.org/C92757383","wikidata":"https://www.wikidata.org/wiki/Q382497","display_name":"Affine transformation","level":2,"score":0.5242999792098999},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.49889999628067017},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.3874000012874603},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.38679999113082886},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3837999999523163},{"id":"https://openalex.org/C204241405","wikidata":"https://www.wikidata.org/wiki/Q461499","display_name":"Transformation (genetics)","level":3,"score":0.3149999976158142},{"id":"https://openalex.org/C207221997","wikidata":"https://www.wikidata.org/wiki/Q938614","display_name":"Affine combination","level":3,"score":0.3070000112056732},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.2953999936580658},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.2822999954223633},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2791000008583069},{"id":"https://openalex.org/C118615104","wikidata":"https://www.wikidata.org/wiki/Q121416","display_name":"Discrete mathematics","level":1,"score":0.2784999907016754},{"id":"https://openalex.org/C136119220","wikidata":"https://www.wikidata.org/wiki/Q1000660","display_name":"Algebra over a field","level":2,"score":0.2734000086784363},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2542000114917755},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.25209999084472656}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/dcc66757.2026.00082","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dcc66757.2026.00082","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 Data Compression Conference (DCC)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2845643761","display_name":null,"funder_award_id":"62371288,62431017,62301299","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Despite":[0],"strong":[1],"generalization":[2],"ability,":[3],"large":[4],"language":[5],"models":[6],"(LLMs)":[7],"are":[8],"resource-intensive":[9],"to":[10,94,102],"deploy.":[11],"Post-training":[12],"quantization":[13,96],"(PTQ)":[14],"effectively":[15],"compresses":[16],"pretrained":[17],"LLMs":[18],"without":[19],"retraining.":[20],"Although":[21],"recent":[22],"studies":[23],"target":[24],"extremely":[25],"low-bit":[26,113],"quantization,":[27],"most":[28],"methods":[29],"rely":[30],"on":[31],"extra":[32],"parameters":[33],"and":[34,92],"the":[35,108,118],"effective":[36],"bit-width":[37,53,120],"usually":[38],"exceeds":[39],"2":[40],"bits.":[41],"We":[42],"introduce":[43],"a":[44,70,76,84],"saliency-guided":[45],"mixed-precision":[46],"PTQ":[47,114],"framework":[48],"that":[49,107],"achieves":[50],"an":[51],"average":[52,119],"of":[54,89],"1.70":[55],"bits":[56,124],"while":[57,116],"maintaining":[58],"competitive":[59],"performance.":[60],"The":[61,80],"proposed":[62,109],"method":[63,110],"allocates":[64],"bit-widths":[65],"across":[66],"weight":[67,100],"groups":[68],"under":[69],"strict":[71],"total":[72],"budget,":[73],"guided":[74],"by":[75,121],"cross-precision":[77],"saliency":[78],"metric.":[79],"key":[81],"innovation":[82],"is":[83],"learnable":[85],"affine":[86],"transformation":[87],"consisting":[88],"rotation,":[90],"scaling,":[91],"shifting":[93],"align":[95],"grids":[97],"with":[98],"local":[99],"distributions":[101],"minimize":[103],"error.":[104],"Experiments":[105],"show":[106],"matches":[111],"state-of-the-art":[112],"performance,":[115],"reducing":[117],"approximately":[122],"0.55":[123],"in":[125],"Table":[126],"1.":[127]},"counts_by_year":[],"updated_date":"2026-05-16T06:04:12.930555","created_date":"2026-05-15T00:00:00"}