{"id":"https://openalex.org/W4416036650","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.432","title":"MUZO: Leveraging Multiple Queries and Momentum for Zeroth-Order Fine-Tuning of Large Language Models","display_name":"MUZO: Leveraging Multiple Queries and Momentum for Zeroth-Order Fine-Tuning of Large Language Models","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416036650","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.432"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.432","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.432","pdf_url":"https://aclanthology.org/2025.emnlp-main.432.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.432.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5120794135","display_name":"Yuezhang Peng","orcid":"https://orcid.org/0000-0001-6358-2333"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuezhang Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036501342","display_name":"Yu-xin Liu","orcid":"https://orcid.org/0000-0002-9054-2510"},"institutions":[{"id":"https://openalex.org/I190752583","display_name":"ParisTech","ror":"https://ror.org/05c2qg481","country_code":"FR","type":"education","lineage":["https://openalex.org/I190752583"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Yuxin Liu","raw_affiliation_strings":["SJTU Paris Elite Institute of Technology ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SJTU Paris Elite Institute of Technology ,","institution_ids":["https://openalex.org/I190752583"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034251794","display_name":"Fei Wen","orcid":"https://orcid.org/0000-0002-3083-9611"},"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":"Fei Wen","raw_affiliation_strings":["School of Information Science and Electronic Engineering / School of Integrated Circuits , Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Information Science and Electronic Engineering / School of Integrated Circuits , Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100329117","display_name":"Xie Chen","orcid":"https://orcid.org/0000-0001-5801-2571"},"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":"Xie Chen","raw_affiliation_strings":["X-LANCE Lab , School of Computer Science , MoE Key Lab of Artificial Intelligence , Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"X-LANCE Lab , School of Computer Science , MoE Key Lab of Artificial Intelligence , Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]}],"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":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.29004997,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"8577","last_page":"8595"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.33090001344680786,"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":0.33090001344680786,"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":0.20000000298023224,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.045099999755620956,"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/momentum","display_name":"Momentum (technical analysis)","score":0.3675000071525574},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.3084999918937683},{"id":"https://openalex.org/keywords/field","display_name":"Field (mathematics)","score":0.28139999508857727},{"id":"https://openalex.org/keywords/information-system","display_name":"Information system","score":0.26759999990463257}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6389999985694885},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3725999891757965},{"id":"https://openalex.org/C60718061","wikidata":"https://www.wikidata.org/wiki/Q1414747","display_name":"Momentum (technical analysis)","level":2,"score":0.3675000071525574},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3237000107765198},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3084999918937683},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2863999903202057},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.28139999508857727},{"id":"https://openalex.org/C180198813","wikidata":"https://www.wikidata.org/wiki/Q121182","display_name":"Information system","level":2,"score":0.26759999990463257},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2667999863624573},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.25870001316070557}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.432","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.432","pdf_url":"https://aclanthology.org/2025.emnlp-main.432.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.432","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.432","pdf_url":"https://aclanthology.org/2025.emnlp-main.432.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3278014984","display_name":null,"funder_award_id":"U23B2018","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8721642152","display_name":null,"funder_award_id":"2021SHZDZX0102","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":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416036650.pdf","grobid_xml":"https://content.openalex.org/works/W4416036650.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Fine-tuning":[0],"pre-trained":[1],"large":[2,65],"language":[3,66],"models":[4],"(LLMs)":[5],"on":[6,80,129],"downstream":[7,130],"tasks":[8],"has":[9,56],"achieved":[10],"significant":[11],"success":[12],"across":[13],"various":[14,101],"domains.However,":[15],"as":[16,59],"model":[17],"sizes":[18],"grow,":[19],"traditional":[20],"firstorder":[21],"fine-tuning":[22,64,100],"algorithms":[23],"incur":[24],"substantial":[25],"memory":[26],"overhead":[27],"due":[28],"to":[29,84],"the":[30,50,70,86,108,111],"need":[31],"for":[32,35,63,107],"activation":[33,53],"storage":[34,51],"back-propagation":[36],"(BP).The":[37],"BP-free":[38],"Memory-Efficient":[39],"Zeroth-Order":[40,74],"Optimization":[41],"(MeZO)":[42],"method":[43],"estimates":[44],"gradients":[45],"through":[46],"finite":[47],"differences,":[48],"avoiding":[49],"of":[52,88,110],"values,":[54],"and":[55,123,133],"been":[57],"demonstrated":[58],"a":[60],"viable":[61],"approach":[62],"models.This":[67],"work":[68],"proposes":[69],"MUltiple-query":[71],"Memory":[72],"Efficient":[73],"(MUZO)":[75],"method,":[76],"which":[77],"is":[78],"based":[79],"variance-reduced":[81],"multiple":[82],"queries":[83],"obtain":[85],"average":[87],"gradient":[89],"estimates.When":[90],"combined":[91],"with":[92],"Adam":[93],"optimizer,":[94],"MUZO-Adam":[95,112,118],"demonstrates":[96],"superior":[97],"performance":[98,128],"in":[99],"LLMs.Furthermore,":[102],"we":[103],"provide":[104],"theoretical":[105],"guarantees":[106],"convergence":[109],"optimizer.Extensive":[113],"experiments":[114],"empirically":[115],"demonstrate":[116],"that":[117],"converges":[119],"better":[120],"than":[121],"MeZO-SGD":[122],"achieves":[124],"near":[125],"first-order":[126],"optimizer":[127],"classification,":[131],"multiple-choice,":[132],"generation":[134],"tasks.":[135]},"counts_by_year":[],"updated_date":"2026-08-24T07:32:12.397491","created_date":"2025-11-08T00:00:00"}