{"id":"https://openalex.org/W4384807847","doi":"https://doi.org/10.48550/arxiv.2307.08941","title":"MLP Fusion: Towards Efficient Fine-tuning of Dense and Mixture-of-Experts Language Models","display_name":"MLP Fusion: Towards Efficient Fine-tuning of Dense and Mixture-of-Experts Language Models","publication_year":2023,"publication_date":"2023-07-18","ids":{"openalex":"https://openalex.org/W4384807847","doi":"https://doi.org/10.48550/arxiv.2307.08941"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2307.08941","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2307.08941","pdf_url":"https://arxiv.org/pdf/2307.08941","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2307.08941","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Ai, Mengting","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ai, Mengting","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024869027","display_name":"Tianxin Wei","orcid":"https://orcid.org/0000-0003-4450-2005"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Tianxin","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100405119","display_name":"Yifan Chen","orcid":"https://orcid.org/0000-0001-5494-4435"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Yifan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101225164","display_name":"Zeming Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Zeming","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5073158087","display_name":"Jingrui He","orcid":"https://orcid.org/0000-0002-6429-6272"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Jingrui","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.9926999807357788,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.9926999807357788,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9850999712944031,"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/T11609","display_name":"Geophysical Methods and Applications","score":0.9735000133514404,"subfield":{"id":"https://openalex.org/subfields/2212","display_name":"Ocean Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7370898127555847},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.48303279280662537},{"id":"https://openalex.org/keywords/multilayer-perceptron","display_name":"Multilayer perceptron","score":0.47267183661460876},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.44217047095298767},{"id":"https://openalex.org/keywords/fusion","display_name":"Fusion","score":0.4215090274810791}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7370898127555847},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.48303279280662537},{"id":"https://openalex.org/C179717631","wikidata":"https://www.wikidata.org/wiki/Q2991667","display_name":"Multilayer perceptron","level":3,"score":0.47267183661460876},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44217047095298767},{"id":"https://openalex.org/C158525013","wikidata":"https://www.wikidata.org/wiki/Q2593739","display_name":"Fusion","level":2,"score":0.4215090274810791},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2307.08941","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2307.08941","pdf_url":"https://arxiv.org/pdf/2307.08941","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},{"id":"doi:10.48550/arxiv.2307.08941","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2307.08941","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2307.08941","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2307.08941","pdf_url":"https://arxiv.org/pdf/2307.08941","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.46000000834465027}],"awards":[{"id":"https://openalex.org/G1366408413","display_name":null,"funder_award_id":"211790","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G360821608","display_name":"EAGER: Weakly Supervised Graph Neural Networks","funder_award_id":"2137468","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G7286171126","display_name":null,"funder_award_id":"1947203","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G7374246537","display_name":"III: Small: RareXplain: A Computational Framework for Explainable Rare Category Analysis","funder_award_id":"2117902","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4384807847.pdf"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W2358668433","https://openalex.org/W2376932109","https://openalex.org/W3158157485","https://openalex.org/W2789124470","https://openalex.org/W3000407446","https://openalex.org/W2116531472","https://openalex.org/W2103550798"],"abstract_inverted_index":{"Fine-tuning":[0],"a":[1,85,91,128,136,147],"pre-trained":[2],"language":[3,14,202],"model":[4],"(PLM)":[5],"emerges":[6],"as":[7,127,146],"the":[8,46,66,72,79,102,110,125,154,157,188,192,211],"predominant":[9],"strategy":[10],"in":[11,84,173],"many":[12],"natural":[13,201],"processing":[15],"applications.":[16],"However,":[17],"this":[18,62],"process":[19],"is":[20,162,218],"known":[21],"to":[22,44,89,164,185,209],"be":[23,144],"expensive,":[24],"especially":[25],"on":[26,199],"edge":[27],"devices":[28],"with":[29],"low":[30],"computing":[31],"power.":[32],"While":[33],"general":[34],"approaches":[35],"(e.g.":[36],"quantization":[37],"and":[38,87,132,150,169,178,204],"distillation)":[39],"have":[40],"been":[41],"widely":[42],"studied":[43],"reduce":[45],"compute/memory":[47],"of":[48,76,130,139,156,196,213],"PLM":[49,86,93,197],"fine-tuning,":[50],"one-shot":[51],"compression":[52,103,190],"techniques":[53],"specifically":[54],"designed":[55],"for":[56],"fine-tuning":[57,198],"remain":[58],"largely":[59],"unexplored.":[60],"In":[61],"paper,":[63],"we":[64,123,181],"investigate":[65],"neural":[67,77],"tangent":[68],"kernel":[69],"(NTK)--which":[70],"reveals":[71],"gradient":[73],"descent":[74],"dynamics":[75],"networks--of":[78],"multilayer":[80],"perceptrons":[81],"(MLP)":[82],"modules":[83,168,172],"propose":[88],"coin":[90],"lightweight":[92],"through":[94],"NTK-approximating":[95],"MLP":[96,105,126,149,167,214],"fusion.":[97,215],"By":[98],"incorporating":[99],"NTK":[100,155],"into":[101,135],"process,":[104],"Fusion":[106],"not":[107],"only":[108],"preserves":[109,191],"original":[111,158],"model's":[112],"output":[113],"but":[114],"also":[115],"maintains":[116],"its":[117,176],"training":[118],"dynamics.":[119],"To":[120],"achieve":[121],"this,":[122],"reconsider":[124],"bundle":[129],"sub-MLPs":[131],"cluster":[133],"them":[134],"given":[137],"number":[138],"centroids,":[140],"which":[141],"can":[142],"then":[143],"restored":[145],"compressed":[148],"surprisingly":[151],"well":[152],"approximate":[153],"PLM.":[159],"Our":[160,216],"approach":[161],"applicable":[163],"both":[165,200],"standard":[166],"Mixture-of-Experts":[170],"(MoE)":[171],"PLMs,":[174],"demonstrating":[175],"scalability":[177],"versatility.":[179],"Additionally,":[180],"provide":[182],"theoretical":[183],"derivations":[184],"demonstrate":[186],"how":[187],"proposed":[189],"NTK.":[193],"Extensive":[194],"experiments":[195],"understanding":[203],"generation":[205],"tasks":[206],"are":[207],"provided":[208],"verify":[210],"effectiveness":[212],"code":[217],"available":[219],"at":[220],"https://github.com/weitianxin/MLP_Fusion.":[221]},"counts_by_year":[{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":1}],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2023-07-20T00:00:00"}