{"id":"https://openalex.org/W4403324162","doi":"https://doi.org/10.18653/v1/2026.acl-long.78","title":"SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe","display_name":"SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W4403324162","doi":"https://doi.org/10.18653/v1/2026.acl-long.78"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.acl-long.78","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.78","pdf_url":null,"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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.18653/v1/2026.acl-long.78","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102597479","display_name":"Yuxin Xiao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuxin Xiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056212977","display_name":"Shujian Zhang","orcid":"https://orcid.org/0000-0003-3216-722X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shujian Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109022963","display_name":"Marzyeh Ghassemi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Marzyeh Ghassemi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5080937886","display_name":"Wenxuan Zhou","orcid":"https://orcid.org/0000-0003-1199-885X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenxuan Zhou","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.00242047,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1705","last_page":"1718"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9940999746322632,"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":0.9940999746322632,"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/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.9861000180244446,"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/T10028","display_name":"Topic Modeling","score":0.9837999939918518,"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/recipe","display_name":"Recipe","score":0.9824339151382446},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5874931812286377},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5024082660675049},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.2761572003364563},{"id":"https://openalex.org/keywords/chemistry","display_name":"Chemistry","score":0.13175523281097412}],"concepts":[{"id":"https://openalex.org/C2778671685","wikidata":"https://www.wikidata.org/wiki/Q219239","display_name":"Recipe","level":2,"score":0.9824339151382446},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5874931812286377},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5024082660675049},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2761572003364563},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.13175523281097412},{"id":"https://openalex.org/C31903555","wikidata":"https://www.wikidata.org/wiki/Q1637030","display_name":"Food science","level":1,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.78","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.78","pdf_url":null,"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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2410.05248","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2410.05248","pdf_url":"https://arxiv.org/pdf/2410.05248","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.2410.05248","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2410.05248","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":"doi:10.18653/v1/2026.acl-long.78","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.78","pdf_url":null,"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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W3161239248","https://openalex.org/W258429745","https://openalex.org/W2740181000","https://openalex.org/W1584543623","https://openalex.org/W2561508161","https://openalex.org/W3123969856","https://openalex.org/W2098178683","https://openalex.org/W2740680361"],"abstract_inverted_index":{"To":[0],"acquire":[1],"instruction-following":[2,155],"capabilities,":[3],"large":[4],"language":[5],"models":[6],"(LLMs)":[7],"undergo":[8],"instruction":[9,25,63,95],"tuning,":[10],"where":[11],"they":[12],"are":[13],"trained":[14],"on":[15,29,67,111,142],"instruction-response":[16],"pairs":[17],"using":[18],"next-token":[19],"prediction":[20],"(NTP).":[21],"Efforts":[22],"to":[23,101,108,118,129,139,186,191],"improve":[24],"tuning":[26,64],"often":[27],"focus":[28],"higher-quality":[30],"supervised":[31],"fine-tuning":[32],"(SFT)":[33],"datasets,":[34],"typically":[35],"requiring":[36],"data":[37,98,105,183],"filtering":[38],"with":[39,86,121,160,182],"proprietary":[40],"LLMs":[41,73],"or":[42],"human":[43],"annotation.":[44],"In":[45],"this":[46,112],"paper,":[47],"we":[48],"take":[49],"a":[50,56,136],"different":[51,87],"approach":[52],"by":[53],"proposing":[54],"SFTMix,":[55],"novel":[57],"Mixup-based":[58,137],"recipe":[59],"that":[60,72,84],"elevates":[61],"LLM":[62,164],"without":[65],"relying":[66],"well-curated":[68],"datasets.":[69],"We":[70,82,125,147],"observe":[71],"exhibit":[74],"uneven":[75],"confidence":[76,88,123,132],"across":[77,163,176],"the":[78,131,149],"semantic":[79],"representation":[80],"space.":[81],"argue":[83],"examples":[85,120],"levels":[89],"should":[90],"play":[91],"distinct":[92],"roles":[93],"in":[94,153],"tuning:":[96],"Confident":[97],"is":[99,106],"prone":[100],"overfitting,":[102],"while":[103],"unconfident":[104],"harder":[107],"generalize.":[109],"Based":[110],"insight,":[113],"SFTMix":[114,152],"leverages":[115],"training":[116],"dynamics":[117],"identify":[119],"varying":[122,170],"levels.":[124],"then":[126],"interpolate":[127],"them":[128],"bridge":[130],"gap":[133],"and":[134,156,166,172,189],"apply":[135],"regularization":[138],"support":[140],"learning":[141],"these":[143],"additional,":[144],"interpolated":[145],"examples.":[146],"demonstrate":[148],"effectiveness":[150],"of":[151,169],"both":[154],"healthcare-specific":[157],"SFT":[158,167],"tasks,":[159],"consistent":[161],"improvements":[162],"families":[165],"datasets":[168],"sizes":[171],"qualities.":[173],"Extensive":[174],"analyses":[175],"six":[177],"directions":[178],"highlight":[179],"SFTMix's":[180],"compatibility":[181],"selection,":[184],"adaptability":[185],"compute-constrained":[187],"scenarios,":[188],"scalability":[190],"broader":[192],"applications.":[193]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}