{"id":"https://openalex.org/W4394867600","doi":"https://doi.org/10.48550/arxiv.2404.08985","title":"T-REX: Mixture-of-Rank-One-Experts with Semantic-aware Intuition for Multi-task Large Language Model Finetuning","display_name":"T-REX: Mixture-of-Rank-One-Experts with Semantic-aware Intuition for Multi-task Large Language Model Finetuning","publication_year":2024,"publication_date":"2024-04-13","ids":{"openalex":"https://openalex.org/W4394867600","doi":"https://doi.org/10.48550/arxiv.2404.08985"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2404.08985","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2404.08985","pdf_url":"https://arxiv.org/pdf/2404.08985","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,"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"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2404.08985","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101803269","display_name":"Rongyu Zhang","orcid":"https://orcid.org/0000-0002-9174-1765"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Rongyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101671391","display_name":"Yijiang Liu","orcid":"https://orcid.org/0000-0001-5914-1607"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yijiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076154259","display_name":"Huanrui Yang","orcid":"https://orcid.org/0000-0002-3384-4512"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Huanrui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Zheng, Shenli","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Shenli","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Wang, Dan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Dan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113530023","display_name":"Yuan Du","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104294805","display_name":"Li Du","orcid":"https://orcid.org/0009-0003-6284-4612"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5013030532","display_name":"Shanghang Zhang","orcid":"https://orcid.org/0000-0003-4047-3526"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Shanghang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9972000122070312,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9972000122070312,"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/T10812","display_name":"Human Pose and Action Recognition","score":0.9728999733924866,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.9333999752998352,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7350074052810669},{"id":"https://openalex.org/keywords/intuition","display_name":"Intuition","score":0.6601826548576355},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.48927661776542664},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.47736263275146484},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.43396157026290894},{"id":"https://openalex.org/keywords/data-science","display_name":"Data science","score":0.3701487183570862},{"id":"https://openalex.org/keywords/cognitive-science","display_name":"Cognitive science","score":0.1873226761817932},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.1102398931980133}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7350074052810669},{"id":"https://openalex.org/C132010649","wikidata":"https://www.wikidata.org/wiki/Q189222","display_name":"Intuition","level":2,"score":0.6601826548576355},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48927661776542664},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.47736263275146484},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.43396157026290894},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.3701487183570862},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.1873226761817932},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.1102398931980133},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2404.08985","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2404.08985","pdf_url":"https://arxiv.org/pdf/2404.08985","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,"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.2404.08985","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2404.08985","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2404.08985","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2404.08985","pdf_url":"https://arxiv.org/pdf/2404.08985","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,"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"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.5}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4394867600.pdf"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2961085424","https://openalex.org/W4306674287","https://openalex.org/W3046775127","https://openalex.org/W3107602296","https://openalex.org/W3170094116","https://openalex.org/W4386462264","https://openalex.org/W4364306694","https://openalex.org/W4312192474","https://openalex.org/W4283697347","https://openalex.org/W4210805261"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"encounter":[4],"significant":[5],"adaptation":[6],"challenges":[7],"in":[8],"diverse":[9,146],"multitask":[10],"finetuning.":[11],"Mixture-of-experts":[12],"(MoE)":[13],"provides":[14],"a":[15,19,56,80,129],"promising":[16],"solution":[17],"with":[18,91,99,149,161],"dynamic":[20],"architecture,":[21],"enabling":[22,122],"effective":[23],"task":[24],"decoupling.":[25],"However,":[26],"scaling":[27],"up":[28,155],"the":[29,63,86,109,112],"number":[30],"of":[31,65,89,116],"MoE":[32],"experts":[33,68,78,90,127],"incurs":[34],"substantial":[35],"parameter":[36,93],"and":[37,40,134,143],"computational":[38],"overheads":[39],"suffers":[41],"from":[42],"limited":[43],"performance":[44],"gain":[45],"due":[46],"to":[47,69,83,108,156],"naive":[48],"routing":[49],"mechanisms.":[50],"In":[51,102],"this":[52],"paper,":[53],"we":[54],"design":[55],"novel":[57],"framework,":[58],"mix\\underline{\\textbf{T}}ure\\underline{\\textbf{-}}of-\\underline{\\textbf{R}}ank-on\\underline{\\textbf{E}}-e\\underline{\\textbf{X}}perts":[59],"(\\texttt{T-REX}),":[60],"which":[61],"leverages":[62],"combination":[64],"ultra-low":[66],"rank":[67],"construct":[70],"LoRA":[71],"weights":[72],"on":[73],"pretrained":[74],"LLMs.":[75],"The":[76],"rank-1":[77],"enable":[79],"mix-and-match":[81],"mechanism":[82],"quadratically":[84],"expand":[85],"vector":[87],"subspace":[88],"linear":[92],"overheads,":[94],"achieving":[95],"approximate":[96],"error":[97],"reduction":[98],"optimal":[100],"efficiency.":[101],"addition,":[103],"T-REX":[104,139,153],"offers":[105],"implicit":[106],"guidance":[107],"router,":[110],"leveraging":[111],"inherent":[113],"semantic":[114],"clustering":[115],"training":[117],"embeddings":[118],"as":[119],"prior":[120],"knowledge,":[121],"optimized":[123],"feature":[124],"allocation":[125],"across":[126,145,167],"for":[128],"smoother":[130],"convergence.":[131],"Extensive":[132],"theoretical":[133],"empirical":[135],"results":[136],"demonstrate":[137],"that":[138],"achieves":[140,154],"superior":[141],"efficiency":[142],"generalizability":[144],"tasks.":[147],"Compared":[148],"other":[150],"LoRA-based":[151],"methods,":[152],"1.78\\%":[157],"mean":[158],"accuracy":[159],"improvement":[160],"around":[162],"30\\%-40\\%":[163],"less":[164],"trainable":[165],"parameters":[166],"14":[168],"public":[169],"datasets.":[170],"\\href{https://github.com/RoyZry98/T-REX-Pytorch}{Code}":[171],"is":[172],"available.":[173]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2024-04-17T00:00:00"}