{"id":"https://openalex.org/W4415428134","doi":"https://doi.org/10.3233/faia251320","title":"CG-FedLLM: How to Compress Gradients in Federated Fine-Tuning for Large Language Models","display_name":"CG-FedLLM: How to Compress Gradients in Federated Fine-Tuning for Large Language Models","publication_year":2025,"publication_date":"2025-10-21","ids":{"openalex":"https://openalex.org/W4415428134","doi":"https://doi.org/10.3233/faia251320"},"language":null,"primary_location":{"id":"doi:10.3233/faia251320","is_oa":true,"landing_page_url":"https://doi.org/10.3233/faia251320","pdf_url":null,"source":{"id":"https://openalex.org/S4210201731","display_name":"Frontiers in artificial intelligence and applications","issn_l":"0922-6389","issn":["0922-6389","1879-8314"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Frontiers in Artificial Intelligence and Applications","raw_type":"book-chapter"},"type":"book-chapter","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.3233/faia251320","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101835178","display_name":"Huiwen Wu","orcid":"https://orcid.org/0000-0001-8471-4219"},"institutions":[{"id":"https://openalex.org/I4210123185","display_name":"Zhejiang Lab","ror":"https://ror.org/02m2h7991","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210123185"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Huiwen Wu","raw_affiliation_strings":["Zhejiang Laboratory"],"affiliations":[{"raw_affiliation_string":"Zhejiang Laboratory","institution_ids":["https://openalex.org/I4210123185"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035085515","display_name":"Xiaogang Xu","orcid":"https://orcid.org/0000-0002-2991-6077"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiaogang Xu","raw_affiliation_strings":["The Chinese University of Hong Kong"],"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120092142","display_name":"Deyi Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I4210123185","display_name":"Zhejiang Lab","ror":"https://ror.org/02m2h7991","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210123185"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Deyi Zhang","raw_affiliation_strings":["Zhejiang Laboratory"],"affiliations":[{"raw_affiliation_string":"Zhejiang Laboratory","institution_ids":["https://openalex.org/I4210123185"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100644758","display_name":"Xiaohan Li","orcid":"https://orcid.org/0000-0003-3156-1989"},"institutions":[{"id":"https://openalex.org/I4210123185","display_name":"Zhejiang Lab","ror":"https://ror.org/02m2h7991","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210123185"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiaohan Li","raw_affiliation_strings":["Zhejiang Laboratory"],"affiliations":[{"raw_affiliation_string":"Zhejiang Laboratory","institution_ids":["https://openalex.org/I4210123185"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101366364","display_name":"Jiafei Wu","orcid":null},"institutions":[{"id":"https://openalex.org/I4210123185","display_name":"Zhejiang Lab","ror":"https://ror.org/02m2h7991","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210123185"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiafei Wu","raw_affiliation_strings":["Zhejiang Laboratory"],"affiliations":[{"raw_affiliation_string":"Zhejiang Laboratory","institution_ids":["https://openalex.org/I4210123185"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100356609","display_name":"Zhe Liu","orcid":"https://orcid.org/0000-0001-7528-1730"},"institutions":[{"id":"https://openalex.org/I4210123185","display_name":"Zhejiang Lab","ror":"https://ror.org/02m2h7991","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210123185"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhe Liu","raw_affiliation_strings":["Zhejiang Laboratory"],"affiliations":[{"raw_affiliation_string":"Zhejiang Laboratory","institution_ids":["https://openalex.org/I4210123185"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5101835178"],"corresponding_institution_ids":["https://openalex.org/I4210123185"],"apc_list":null,"apc_paid":null,"fwci":5.2022,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.96092488,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.8662999868392944,"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":0.8662999868392944,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.8400999903678894,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.8321999907493591,"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/pipeline","display_name":"Pipeline (software)","score":0.6944000124931335},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.6251999735832214},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.560699999332428},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.5577999949455261},{"id":"https://openalex.org/keywords/filter","display_name":"Filter (signal processing)","score":0.5019000172615051},{"id":"https://openalex.org/keywords/resource","display_name":"Resource (disambiguation)","score":0.4375},{"id":"https://openalex.org/keywords/federated-learning","display_name":"Federated learning","score":0.40849998593330383},{"id":"https://openalex.org/keywords/raw-data","display_name":"Raw data","score":0.37070000171661377}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8062000274658203},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6944000124931335},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.6251999735832214},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.560699999332428},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.5577999949455261},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.5019000172615051},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.4375},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4334999918937683},{"id":"https://openalex.org/C2992525071","wikidata":"https://www.wikidata.org/wiki/Q50818671","display_name":"Federated learning","level":2,"score":0.40849998593330383},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.37070000171661377},{"id":"https://openalex.org/C29202148","wikidata":"https://www.wikidata.org/wiki/Q287260","display_name":"Resource allocation","level":2,"score":0.33379998803138733},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3294000029563904},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3224000036716461},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.30649998784065247},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2906999886035919},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.2863999903202057},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.27410000562667847},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.2639999985694885},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.2540000081062317},{"id":"https://openalex.org/C158156997","wikidata":"https://www.wikidata.org/wiki/Q1416645","display_name":"Models of communication","level":2,"score":0.25270000100135803}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.3233/faia251320","is_oa":true,"landing_page_url":"https://doi.org/10.3233/faia251320","pdf_url":null,"source":{"id":"https://openalex.org/S4210201731","display_name":"Frontiers in artificial intelligence and applications","issn_l":"0922-6389","issn":["0922-6389","1879-8314"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Frontiers in Artificial Intelligence and Applications","raw_type":"book-chapter"}],"best_oa_location":{"id":"doi:10.3233/faia251320","is_oa":true,"landing_page_url":"https://doi.org/10.3233/faia251320","pdf_url":null,"source":{"id":"https://openalex.org/S4210201731","display_name":"Frontiers in artificial intelligence and applications","issn_l":"0922-6389","issn":["0922-6389","1879-8314"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Frontiers in Artificial Intelligence and Applications","raw_type":"book-chapter"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"success":[1],"of":[2,130,200,222],"current":[3],"Large-Language":[4],"Models":[5],"(LLMs)":[6],"hinges":[7],"on":[8,92,105,171,205],"extensive":[9],"training":[10,118],"data":[11],"that":[12,120,146,203],"are":[13],"collected":[14],"and":[15,31,102,134,152,164,175,185,210,225],"stored":[16],"centrally,":[17],"called":[18],"Centralized":[19],"Learning":[20,37],"(CL).":[21],"However,":[22],"such":[23],"a":[24,28,103,116,198],"collection":[25],"manner":[26],"poses":[27],"privacy":[29],"threat,":[30],"one":[32],"potential":[33],"solution":[34],"is":[35,178],"Federated":[36,135],"(FL),":[38],"which":[39],"transfers":[40],"gradients,":[41],"not":[42],"raw":[43],"data,":[44],"among":[45],"clients.":[46],"Unlike":[47],"traditional":[48,162],"networks,":[49],"FL":[50,83],"for":[51],"LLMs":[52,227],"incurs":[53],"significant":[54],"communication":[55,75,150,207],"costs":[56,151],"due":[57],"to":[58,70,73,96,109,126,139],"their":[59],"tremendous":[60],"parameters.":[61],"In":[62],"this":[63,214],"study,":[64],"we":[65,196],"introduce":[66],"an":[67,90],"innovative":[68],"approach":[69,88,148],"compress":[71,140],"gradients":[72,129,141,189],"improve":[74],"efficiency":[76],"during":[77],"LLM":[78],"FL,":[79],"formulating":[80],"the":[81,93,98,106,111,131,206,220],"new":[82],"pipeline":[84],"named":[85],"CG-FedLLM.":[86],"This":[87,177],"integrates":[89],"encoder":[91],"client":[94],"side":[95,108],"acquire":[97],"compressed":[99],"gradient":[100],"features":[101],"decoder":[104],"server":[107],"reconstruct":[110],"gradients.":[112],"We":[113],"also":[114],"develop":[115],"novel":[117],"strategy":[119],"comprises":[121],"Temporal-ensemble":[122],"Gradient-Aware":[123],"Pre-training":[124],"(TGAP)":[125],"identify":[127],"characteristic":[128],"target":[132],"model":[133],"AutoEncoder-Involved":[136],"Fine-tuning":[137],"(FAF)":[138],"adaptively.":[142],"Extensive":[143],"experiments":[144],"confirm":[145],"our":[147,180],"reduces":[149],"improves":[153],"performance":[154],"(e.g.,":[155],"average":[156],"3":[157],"points":[158],"increment":[159],"compared":[160],"with":[161,167],"CL-":[163],"FL-based":[165],"fine-tuning":[166],"several":[168],"foundation":[169],"models":[170],"well-recognized":[172],"benchmarks,":[173],"MMLU":[174],"C-Eval).":[176],"because":[179],"encoder-decoder,":[181],"trained":[182],"via":[183],"TGAP":[184],"FAF,":[186],"can":[187],"filter":[188],"while":[190],"selectively":[191],"preserving":[192],"critical":[193],"features.":[194],"Furthermore,":[195],"present":[197],"series":[199],"experimental":[201],"analyses":[202],"focus":[204],"efficiency,":[208],"accuracy,":[209],"generalization":[211],"ability":[212],"within":[213],"privacy-centric":[215],"framework,":[216],"providing":[217],"insights":[218],"into":[219],"development":[221],"more":[223],"efficient":[224],"private":[226],"fine-tuning.":[228]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2025-10-24T00:00:00"}