{"id":"https://openalex.org/W7204475239","doi":"https://doi.org/10.18653/2026.emnlp-main.1524","title":"FAMPWQ: Fisher Information-based Adaptive Mixed Precision Weight Quantization for Effective LLM Inference","display_name":"FAMPWQ: Fisher Information-based Adaptive Mixed Precision Weight Quantization for Effective LLM Inference","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7204475239","doi":"https://doi.org/10.18653/2026.emnlp-main.1524"},"language":"en","primary_location":{"id":"doi:10.18653/2026.emnlp-main.1524","doi":"https://doi.org/10.18653/2026.emnlp-main.1524","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.1524","pdf_url":"https://aclanthology.org/2026.emnlp-main.1524.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","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":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2026.emnlp-main.1524.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5086708262","display_name":"Gongwei Lee","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I3923682","display_name":"Soochow University","ror":"https://ror.org/05t8y2r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I3923682"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"GongWei Lee","raw_affiliation_strings":["Hithink Research , Hangzhou , China","School of Computer Science and Technology , Soochow University , Suzhou , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hithink Research , Hangzhou , China","institution_ids":[]},{"raw_affiliation_string":"School of Computer Science and Technology , Soochow University , Suzhou , China","institution_ids":["https://openalex.org/I3923682"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127563082","display_name":"Ji Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ji Liu","raw_affiliation_strings":["Electronic Engineering , Tsinghua University , Beijing , China","Hithink Research , Hangzhou , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Electronic Engineering , Tsinghua University , Beijing , China","institution_ids":["https://openalex.org/I99065089"]},{"raw_affiliation_string":"Hithink Research , Hangzhou , China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031306843","display_name":"Juncheng Jia","orcid":"https://orcid.org/0000-0001-8276-7640","observed_orcids":["https://orcid.org/0000-0001-8276-7640"]},"institutions":[{"id":"https://openalex.org/I3923682","display_name":"Soochow University","ror":"https://ror.org/05t8y2r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I3923682"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Juncheng Jia","raw_affiliation_strings":["School of Computer Science and Technology , Soochow University , Suzhou , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Technology , Soochow University , Suzhou , China","institution_ids":["https://openalex.org/I3923682"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5123549912","display_name":"Ji Wu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ji Wu","raw_affiliation_strings":["Electronic Engineering , Tsinghua University , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Electronic Engineering , Tsinghua University , Beijing , China","institution_ids":["https://openalex.org/I99065089"]}]}],"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":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.69829442,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"34781","last_page":"34801"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9735999703407288,"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":0.9735999703407288,"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/T10028","display_name":"Topic Modeling","score":0.003000000026077032,"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.00279999990016222,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7360000014305115},{"id":"https://openalex.org/keywords/weight-quantization","display_name":"weight quantization","score":0.925000011920929},{"id":"https://openalex.org/keywords/mixed-precision-quantization","display_name":"mixed-precision quantization","score":0.9279999732971191},{"id":"https://openalex.org/keywords/fisher-information","display_name":"Fisher information","score":0.8759999871253967},{"id":"https://openalex.org/keywords/bit-width-allocation","display_name":"bit-width allocation","score":0.9490000009536743},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"reinforcement learning","score":0.847000002861023},{"id":"https://openalex.org/keywords/large-language-model-inference","display_name":"large language model inference","score":0.7770000100135803},{"id":"https://openalex.org/keywords/resource-constrained-devices","display_name":"resource-constrained devices","score":0.8019999861717224},{"id":"https://openalex.org/keywords/gpu-deployment","display_name":"GPU deployment","score":0.6899999976158142}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7756999731063843},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6010000109672546},{"id":"https://openalex.org/C29406490","wikidata":"https://www.wikidata.org/wiki/Q1420659","display_name":"Fisher information","level":2,"score":0.5605000257492065},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5425999760627747},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.4830999970436096},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4453999996185303},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.42669999599456787},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3935000002384186},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34470000863075256},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3370000123977661},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.3273000121116638},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.3116999864578247},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3098999857902527},{"id":"https://openalex.org/C162262903","wikidata":"https://www.wikidata.org/wiki/Q343527","display_name":"Allocator","level":2,"score":0.3098999857902527},{"id":"https://openalex.org/C2780069185","wikidata":"https://www.wikidata.org/wiki/Q7977945","display_name":"Equivalence (formal languages)","level":2,"score":0.3028999865055084},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2971000075340271},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.2888999879360199},{"id":"https://openalex.org/C2983203078","wikidata":"https://www.wikidata.org/wiki/Q255166","display_name":"Information gain","level":2,"score":0.2711000144481659},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.26440000534057617},{"id":"https://openalex.org/C207798031","wikidata":"https://www.wikidata.org/wiki/Q8563425","display_name":"Fisher kernel","level":5,"score":0.26269999146461487}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/2026.emnlp-main.1524","doi":"https://doi.org/10.18653/2026.emnlp-main.1524","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.1524","pdf_url":"https://aclanthology.org/2026.emnlp-main.1524.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","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":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2608.24945","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24945","pdf_url":"https://arxiv.org/pdf/2608.24945","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.24945","doi":"https://doi.org/10.48550/arxiv.2608.24945","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.24945","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/2026.emnlp-main.1524","doi":"https://doi.org/10.18653/2026.emnlp-main.1524","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.1524","pdf_url":"https://aclanthology.org/2026.emnlp-main.1524.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","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":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320327518","display_name":"Priority Academic Program Development of Jiangsu Higher Education Institutions","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204475239.pdf","grobid_xml":"https://content.openalex.org/works/W7204475239.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W6948447698","https://openalex.org/W4386080707","https://openalex.org/W4415948623","https://openalex.org/W4393147284","https://openalex.org/W4416035109","https://openalex.org/W7162503198","https://openalex.org/W7138927646","https://openalex.org/W4409282445","https://openalex.org/W4392181806","https://openalex.org/W4403579085"],"abstract_inverted_index":{"Recent":[0],"years":[1],"have":[2],"witnessed":[3],"remarkable":[4],"achievements":[5],"of":[6,19,138],"Large":[7],"Language":[8],"Models":[9],"(LLMs)":[10],"in":[11,105,136],"multiple":[12],"domains,":[13],"while":[14],"the":[15,22,93,116],"excessive":[16],"resource":[17],"requirements":[18],"LLMs":[20],"hinder":[21],"deployment":[23],"on":[24,77,115,122],"resource-constrained":[25],"devices.Although":[26],"model":[27,84],"quantization":[28,36,72],"stands":[29],"out":[30],"as":[31],"an":[32,109],"effective":[33,74],"approach,":[34,65],"conventional":[35],"approaches":[37,135],"typically":[38],"incur":[39],"severe":[40],"performance":[41],"degradation":[42],"due":[43],"to":[44,91,96,141,146,153],"uniform":[45],"bit-width":[46,103,111],"or":[47],"simple":[48],"heuristic":[49],"sensitivity":[50,95,119],"evaluation.In":[51],"this":[52],"paper,":[53],"we":[54,80,98],"propose":[55,81,99],"a":[56,82,86,100],"novel":[57,87],"Fisher":[58,88,117],"information-based":[59],"Adaptive":[60],"Mixed":[61],"Precision":[62],"Weight":[63],"Quantization":[64],"i.e.,":[66],"FAMPWQ,":[67,106],"which":[68,107],"performs":[69],"layer-adaptive":[70],"weight":[71],"for":[73],"LLM":[75],"inference":[76],"commodity":[78],"GPUs.First,":[79],"system":[83],"with":[85],"information":[89,118],"metric":[90],"measure":[92],"layer-wise":[94],"quantization.Second,":[97],"reinforcement":[101],"learning-based":[102],"allocator":[104],"generates":[108],"adaptive":[110],"allocation":[112],"strategy":[113],"based":[114],"metric.Extensive":[120],"experiments":[121],"7":[123,133],"models":[124],"and":[125,149],"5":[126],"benchmarks":[127],"demonstrate":[128],"that":[129],"FAMPWQ":[130],"significantly":[131],"outperforms":[132],"baseline":[134],"terms":[137],"PPL":[139],"(up":[140,145,152],"3.39":[142],"smaller),":[143],"accuracy":[144],"6.87%":[147],"higher),":[148],"LLM-as-a-judge":[150],"comparison":[151],"76%":[154],"win":[155],"rate).":[156]},"counts_by_year":[],"updated_date":"2026-10-31T12:09:11.096373","created_date":"2026-08-28T00:00:00"}