{"id":"https://openalex.org/W4392490307","doi":"https://doi.org/10.48550/arxiv.2403.00986","title":"Merging Text Transformer Models from Different Initializations","display_name":"Merging Text Transformer Models from Different Initializations","publication_year":2024,"publication_date":"2024-03-01","ids":{"openalex":"https://openalex.org/W4392490307","doi":"https://doi.org/10.48550/arxiv.2403.00986"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2403.00986","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.00986","pdf_url":"https://arxiv.org/pdf/2403.00986","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":null,"license_id":null,"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/2403.00986","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5016147947","display_name":"Neha Verma","orcid":"https://orcid.org/0000-0002-9675-8427"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Verma, Neha","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5005764737","display_name":"Maha Elbayad","orcid":"https://orcid.org/0000-0002-8389-231X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Elbayad, Maha","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/T10181","display_name":"Natural Language Processing Techniques","score":0.7770000100135803,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.7770000100135803,"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/transformer","display_name":"Transformer","score":0.6096444129943848},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.43976661562919617},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.41148841381073},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.37428000569343567},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.1429932415485382},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.08266580104827881},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.059805065393447876}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6096444129943848},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.43976661562919617},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41148841381073},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.37428000569343567},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.1429932415485382},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.08266580104827881},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.059805065393447876}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2403.00986","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.00986","pdf_url":"https://arxiv.org/pdf/2403.00986","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2403.00986","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2403.00986","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2403.00986","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.00986","pdf_url":"https://arxiv.org/pdf/2403.00986","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4392490307.pdf","grobid_xml":"https://content.openalex.org/works/W4392490307.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W2358668433","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W2382290278","https://openalex.org/W2478288626","https://openalex.org/W4391913857","https://openalex.org/W2350741829","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Recent":[0],"work":[1,24,163],"on":[2,128,135,164],"permutation-based":[3],"model":[4,60,96,123],"merging":[5,61,107,165],"has":[6,25],"shown":[7],"impressive":[8],"low-":[9],"or":[10,133],"zero-barrier":[11],"mode":[12],"connectivity":[13],"between":[14,67,119],"models":[15,109,126,148],"from":[16],"completely":[17],"different":[18],"initializations.":[19],"However,":[20],"this":[21,43],"line":[22],"of":[23,76,146],"not":[26],"yet":[27],"extended":[28],"to":[29,49,63,94,122],"the":[30,38,47,65,71,77,101,144],"Transformer":[31,52,168],"architecture,":[32,78],"despite":[33],"its":[34,80],"dominant":[35],"popularity":[36],"in":[37,42,70,92],"language":[39,137],"domain.":[40],"Therefore,":[41],"work,":[44],"we":[45,113],"investigate":[46,64],"extent":[48],"which":[50],"separate":[51],"minima":[53,69,120,145],"learn":[54],"similar":[55],"features,":[56],"and":[57,85,152,157],"propose":[58],"a":[59,129,136,159],"technique":[62],"relationship":[66],"these":[68,108,147],"loss":[72,117],"landscape.":[73],"The":[74],"specifics":[75],"like":[79],"residual":[81],"connections,":[82],"multi-headed":[83],"attention,":[84],"discrete,":[86],"sequential":[87],"input,":[88],"require":[89],"specific":[90],"interventions":[91],"order":[93],"compute":[95],"permutations":[97],"that":[98,143],"remain":[99],"within":[100],"same":[102],"functional":[103],"equivalence":[104],"class.":[105],"In":[106],"with":[110],"our":[111],"method,":[112],"consistently":[114],"find":[115],"lower":[116],"barriers":[118],"compared":[121],"averaging,":[124],"across":[125],"trained":[127,167],"masked-language":[130],"modeling":[131],"task":[132],"fine-tuned":[134],"understanding":[138],"benchmark.":[139],"Our":[140],"results":[141],"show":[142],"are":[149],"less":[150],"sharp":[151],"isolated":[153],"than":[154],"previously":[155],"understood,":[156],"provide":[158],"basis":[160],"for":[161],"future":[162],"separately":[166],"models.":[169]},"counts_by_year":[],"updated_date":"2026-08-26T07:47:46.906454","created_date":"2025-10-10T00:00:00"}