{"id":"https://openalex.org/W4403883293","doi":"https://doi.org/10.48550/arxiv.2410.01109","title":"Mixing It Up: The Cocktail Effect of Multi-Task Fine-Tuning on LLM Performance -- A Case Study in Finance","display_name":"Mixing It Up: The Cocktail Effect of Multi-Task Fine-Tuning on LLM Performance -- A Case Study in Finance","publication_year":2024,"publication_date":"2024-10-01","ids":{"openalex":"https://openalex.org/W4403883293","doi":"https://doi.org/10.48550/arxiv.2410.01109"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2410.01109","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.01109","pdf_url":"https://arxiv.org/pdf/2410.01109","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,"listed_in":[],"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-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2410.01109","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114452489","display_name":"Meni Brief","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Brief, Meni","raw_affiliation_strings":["Microsoft Industry AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Industry AI","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ovadia, Oded","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ovadia, Oded","raw_affiliation_strings":["Microsoft Industry AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Industry AI","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094151881","display_name":"Gil Shenderovitz","orcid":"https://orcid.org/0000-0003-2974-3548","observed_orcids":["https://orcid.org/0000-0003-2974-3548"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shenderovitz, Gil","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114452490","display_name":"Noga Ben Yoash","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yoash, Noga Ben","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5073818861","display_name":"Rachel Lemberg","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Lemberg, Rachel","raw_affiliation_strings":["Microsoft Industry AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Industry AI","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5083805783","display_name":"Eitam Sheetrit","orcid":"https://orcid.org/0000-0002-0307-131X","observed_orcids":["https://orcid.org/0000-0002-0307-131X"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sheetrit, Eitam","raw_affiliation_strings":["Microsoft Industry AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Industry AI","institution_ids":["https://openalex.org/I1290206253"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":5,"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/T10028","display_name":"Topic Modeling","score":0.7832000255584717,"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":0.7832000255584717,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.20630000531673431,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.0019000000320374966,"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/multi-task-fine-tuning","display_name":"multi-task fine-tuning","score":0.9800000190734863},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9259999990463257},{"id":"https://openalex.org/keywords/domain-specific-llms","display_name":"domain-specific LLMs","score":0.8569999933242798},{"id":"https://openalex.org/keywords/phi-3-mini","display_name":"Phi-3-mini","score":0.7289999723434448},{"id":"https://openalex.org/keywords/gpt-4o","display_name":"GPT-4o","score":0.7039999961853027},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.597000002861023},{"id":"https://openalex.org/keywords/numerical-reasoning","display_name":"numerical reasoning","score":0.7279999852180481},{"id":"https://openalex.org/keywords/financial-reasoning","display_name":"financial reasoning","score":0.4659999907016754},{"id":"https://openalex.org/keywords/domain-knowledge-transfer","display_name":"domain knowledge transfer","score":0.47699999809265137}],"concepts":[{"id":"https://openalex.org/C138777275","wikidata":"https://www.wikidata.org/wiki/Q6884054","display_name":"Mixing (physics)","level":2,"score":0.7170397043228149},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6673581600189209},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.3658192455768585},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3296898603439331},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.22629687190055847},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.184941828250885},{"id":"https://openalex.org/C187736073","wikidata":"https://www.wikidata.org/wiki/Q2920921","display_name":"Management","level":1,"score":0.07686570286750793},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2410.01109","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.01109","pdf_url":"https://arxiv.org/pdf/2410.01109","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,"listed_in":[],"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-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2410.01109","doi":"https://doi.org/10.48550/arxiv.2410.01109","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2410.01109","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,"listed_in":[],"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":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2410.01109","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.01109","pdf_url":"https://arxiv.org/pdf/2410.01109","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,"listed_in":[],"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-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4403883293.pdf","grobid_xml":"https://content.openalex.org/works/W4403883293.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4411447356","https://openalex.org/W4390722655","https://openalex.org/W4404356213","https://openalex.org/W4416035445","https://openalex.org/W4405003679","https://openalex.org/W6968080514","https://openalex.org/W4415762603","https://openalex.org/W4414973989","https://openalex.org/W4406491431","https://openalex.org/W7220836760"],"abstract_inverted_index":{"The":[0],"application":[1],"of":[2,40,78,135,143,150,164],"large":[3],"language":[4],"models":[5,72],"(LLMs)":[6],"in":[7,24,51,169,191,199,202],"domain-specific":[8,52],"contexts,":[9],"including":[10],"finance,":[11],"has":[12],"expanded":[13],"rapidly.":[14],"Domain-specific":[15],"LLMs":[16,42,127],"are":[17,73],"typically":[18],"evaluated":[19],"based":[20],"on":[21,56,75,109],"their":[22],"performance":[23,157],"various":[25],"downstream":[26,185],"tasks":[27,80,186],"relevant":[28],"to":[29,98,175,188],"the":[30,57,63,104,133,141,162],"domain.":[31],"In":[32],"this":[33,89],"work,":[34],"we":[35,48,139,179],"present":[36],"a":[37,76,92,115,148],"detailed":[38],"analysis":[39],"fine-tuning":[41,54,183],"for":[43,184],"such":[44,95],"tasks.":[45,177],"Somewhat":[46],"counterintuitively,":[47],"find":[49],"that":[50,153,172,181],"cases,":[53],"exclusively":[55],"target":[58],"task":[59,192],"is":[60],"not":[61,196],"always":[62],"most":[64],"effective":[65],"strategy.":[66],"Instead,":[67],"multi-task":[68,136],"finetuning":[69],"-":[70,81],"where":[71],"trained":[74],"cocktail":[77],"related":[79],"can":[82],"significantly":[83],"enhance":[84],"performance.":[85],"We":[86,159],"demonstrate":[87],"how":[88],"approach":[90],"enables":[91],"small":[93],"model,":[94],"as":[96,128,147],"Phi-3-Mini,":[97],"achieve":[99],"state-of-the-art":[100],"results,":[101],"even":[102],"surpassing":[103],"much":[105],"larger":[106],"GPT-4-o":[107],"model":[108],"financial":[110,176],"benchmarks.":[111],"Our":[112],"study":[113],"involves":[114],"large-scale":[116],"experiment,":[117],"conducting":[118],"over":[119],"200":[120],"training":[121],"experiments":[122],"using":[123],"several":[124],"widely":[125],"adopted":[126],"baselines,":[129],"and":[130],"empirically":[131],"confirms":[132],"benefits":[134],"fine-tuning.":[137],"Additionally,":[138],"explore":[140],"use":[142],"general":[144],"instruction":[145],"data":[146],"form":[149],"regularization,":[151],"suggesting":[152],"it":[154,194],"helps":[155],"minimize":[156],"degradation.":[158],"also":[160],"investigate":[161],"inclusion":[163],"mathematical":[165],"data,":[166],"finding":[167],"improvements":[168,190],"numerical":[170],"reasoning":[171,208],"transfer":[173],"effectively":[174],"Finally,":[178],"note":[180],"while":[182],"leads":[187],"targeted":[189],"performance,":[193],"does":[195],"necessarily":[197],"result":[198],"broader":[200],"gains":[201],"domain":[203,207],"knowledge":[204],"or":[205],"complex":[206],"abilities.":[209]},"counts_by_year":[{"year":2026,"cited_by_count":3},{"year":2025,"cited_by_count":2}],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2024-10-30T00:00:00"}