{"id":"https://openalex.org/W4414360254","doi":"https://doi.org/10.24963/ijcai.2025/844","title":"FBQuant: FeedBack Quantization for Large Language Models","display_name":"FBQuant: FeedBack Quantization for Large Language Models","publication_year":2025,"publication_date":"2025-09-01","ids":{"openalex":"https://openalex.org/W4414360254","doi":"https://doi.org/10.24963/ijcai.2025/844"},"language":"en","primary_location":{"id":"doi:10.24963/ijcai.2025/844","is_oa":false,"landing_page_url":"https://doi.org/10.24963/ijcai.2025/844","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence","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/A5101671393","display_name":"Yijiang Liu","orcid":"https://orcid.org/0000-0003-2800-4508"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yijiang Liu","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Hengyu Fang","orcid":null},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hengyu Fang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055327514","display_name":"Liulu He","orcid":null},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liulu He","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059182665","display_name":"Rongyu Zhang","orcid":"https://orcid.org/0000-0001-5699-3943"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Rongyu Zhang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080870367","display_name":"Yichuan Bai","orcid":"https://orcid.org/0000-0001-8895-1780"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yichuan Bai","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084816122","display_name":"Yuan Du","orcid":"https://orcid.org/0000-0002-5316-619X"},"institutions":[{"id":"https://openalex.org/I4210153682","display_name":"Intelligent Health (United Kingdom)","ror":"https://ror.org/0576zak10","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210153682"]},{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN","GB"],"is_corresponding":false,"raw_author_name":"Yuan Du","raw_affiliation_strings":["Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Nanjing University, Suzhou","School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Nanjing University, Suzhou","institution_ids":["https://openalex.org/I4210153682"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5039665206","display_name":"Li Du","orcid":"https://orcid.org/0000-0003-2687-6978"},"institutions":[{"id":"https://openalex.org/I4210153682","display_name":"Intelligent Health (United Kingdom)","ror":"https://ror.org/0576zak10","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210153682"]},{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN","GB"],"is_corresponding":false,"raw_author_name":"Li Du","raw_affiliation_strings":["Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Nanjing University, Suzhou","School of Electronic Science and Engineering, Nanjing University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Interdisciplinary Research Center for Future Intelligent Chips (Chip-X), Nanjing University, Suzhou","institution_ids":["https://openalex.org/I4210153682"]},{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University","institution_ids":["https://openalex.org/I881766915"]}]}],"institutions":[],"countries_distinct_count":2,"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":0.21276221,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"7589","last_page":"7597"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9635999798774719,"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/T10028","display_name":"Topic Modeling","score":0.9635999798774719,"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":0.9480000138282776,"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/quantization","display_name":"Quantization (signal processing)","score":0.7605000138282776},{"id":"https://openalex.org/keywords/edge-device","display_name":"Edge device","score":0.6434999704360962},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.632099986076355},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6039000153541565},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.45210000872612},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.4424000084400177},{"id":"https://openalex.org/keywords/offset","display_name":"Offset (computer science)","score":0.41940000653266907}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7605000138282776},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7425000071525574},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.6434999704360962},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.632099986076355},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6039000153541565},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.45210000872612},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.4424000084400177},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.41940000653266907},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.3646000027656555},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3578000068664551},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.35269999504089355},{"id":"https://openalex.org/C2781357197","wikidata":"https://www.wikidata.org/wiki/Q5757597","display_name":"High memory","level":2,"score":0.3497999906539917},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.34139999747276306},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.33469998836517334},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.32690000534057617},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3102000057697296},{"id":"https://openalex.org/C188045654","wikidata":"https://www.wikidata.org/wiki/Q17148339","display_name":"Memory bandwidth","level":2,"score":0.30410000681877136},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.29789999127388},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.26159998774528503},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2558000087738037}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.24963/ijcai.2025/844","is_oa":false,"landing_page_url":"https://doi.org/10.24963/ijcai.2025/844","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Deploying":[0],"Large":[1],"Language":[2],"Models":[3],"(LLMs)":[4],"on":[5,15,88],"edge":[6,38],"devices":[7],"is":[8,29],"increasingly":[9],"important,":[10],"as":[11],"it":[12],"eliminates":[13],"reliance":[14],"network":[16],"connections,":[17],"reduces":[18,57],"expensive":[19],"API":[20],"calls,":[21],"and":[22,157],"enhances":[23],"user":[24],"privacy.":[25],"However,":[26],"on-device":[27],"deployment":[28],"challenging":[30],"due":[31],"to":[32,51,68],"the":[33,42,115,121,126,133,155],"limited":[34],"computational":[35],"resources":[36],"of":[37,128,148,159],"devices.":[39],"In":[40],"particular,":[41],"key":[43],"bottleneck":[44],"stems":[45],"from":[46],"memory":[47,58],"bandwidth":[48],"constraints":[49],"related":[50],"weight":[52],"loading.":[53],"Weight-only":[54],"quantization":[55,76,122],"effectively":[56],"access,":[59],"yet":[60],"often":[61],"induces":[62],"significant":[63],"accuracy":[64,171],"degradation.":[65],"Recent":[66],"efforts":[67],"incorporate":[69],"sub-branches":[70],"have":[71],"shown":[72],"promise":[73],"for":[74,165],"mitigating":[75],"errors,":[77],"but":[78],"these":[79,93],"methods":[80],"either":[81],"lack":[82],"robust":[83],"optimization":[84],"strategies":[85],"or":[86],"rely":[87],"suboptimal":[89],"objectives.":[90],"To":[91,130],"address":[92],"gaps,":[94],"we":[95,139],"propose":[96],"FeedBack":[97],"Quantization":[98],"(FBQuant),":[99],"a":[100],"novel":[101],"approach":[102],"inspired":[103],"by":[104,120,137,172],"negative":[105],"feedback":[106],"mechanisms":[107],"in":[108],"automatic":[109],"control.":[110],"FBQuant":[111,160,168],"inherently":[112],"ensures":[113],"that":[114,145],"reconstructed":[116],"weights":[117],"remain":[118],"bounded":[119],"process,":[123],"thereby":[124],"reducing":[125],"risk":[127],"overfitting.":[129],"further":[131],"offset":[132],"additional":[134],"latency":[135],"introduced":[136],"sub-branches,":[138],"develop":[140],"an":[141],"efficient":[142],"CUDA":[143],"kernel":[144],"decreases":[146],"60%":[147],"extra":[149],"inference":[150],"time.":[151],"Comprehensive":[152],"experiments":[153],"demonstrate":[154],"efficiency":[156],"effectiveness":[158],"across":[161],"various":[162],"LLMs.":[163],"Notably,":[164],"3-bit":[166],"Llama2-7B,":[167],"improves":[169],"zero-shot":[170],"1.2%.":[173]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}