{"id":"https://openalex.org/W4412944565","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1297","title":"Mosaic-IT: Cost-Free Compositional Data Synthesis for Instruction Tuning","display_name":"Mosaic-IT: Cost-Free Compositional Data Synthesis for Instruction Tuning","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412944565","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1297"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.1297","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1297","pdf_url":"https://aclanthology.org/2025.findings-acl.1297.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.1297.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100662583","display_name":"Ming Li","orcid":"https://orcid.org/0000-0002-4706-5173"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ming Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101366341","display_name":"P.\u2010L. Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pei Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079510014","display_name":"Chenguang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chenguang Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101661802","display_name":"Hongyu Zhao","orcid":"https://orcid.org/0000-0003-4744-0153"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hongyu Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100549939","display_name":"Yijun Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yijun Liang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114168273","display_name":"Yupeng Hou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"YuPeng Hou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Fuxiao Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fuxiao Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5039076312","display_name":"Tianyi Zhou","orcid":"https://orcid.org/0000-0001-5348-0632"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianyi 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":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"25287","last_page":"25318"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T14470","display_name":"Advanced Data Processing Techniques","score":0.9318000078201294,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T14470","display_name":"Advanced Data Processing Techniques","score":0.9318000078201294,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12157","display_name":"Geochemistry and Geologic Mapping","score":0.901199996471405,"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/mosaic","display_name":"Mosaic","score":0.8206706047058105},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6469444036483765}],"concepts":[{"id":"https://openalex.org/C110739175","wikidata":"https://www.wikidata.org/wiki/Q133067","display_name":"Mosaic","level":2,"score":0.8206706047058105},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6469444036483765},{"id":"https://openalex.org/C166957645","wikidata":"https://www.wikidata.org/wiki/Q23498","display_name":"Archaeology","level":1,"score":0.0},{"id":"https://openalex.org/C95457728","wikidata":"https://www.wikidata.org/wiki/Q309","display_name":"History","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.1297","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1297","pdf_url":"https://aclanthology.org/2025.findings-acl.1297.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.1297","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1297","pdf_url":"https://aclanthology.org/2025.findings-acl.1297.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412944565.pdf","grobid_xml":"https://content.openalex.org/works/W4412944565.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2493552207","https://openalex.org/W2397874377","https://openalex.org/W4249026204","https://openalex.org/W2027673787","https://openalex.org/W4404777802","https://openalex.org/W4250801634","https://openalex.org/W2327982655"],"abstract_inverted_index":{"Finetuning":[0],"large":[1],"language":[2],"models":[3,25],"with":[4,94,133],"a":[5,55,109],"variety":[6],"of":[7,115],"instruction-response":[8],"pairs":[9],"has":[10],"enhanced":[11],"their":[12],"capability":[13],"to":[14,29,74,89,98],"understand":[15],"and":[16,31,35,43,66,85,103,112,125],"follow":[17],"instructions.Current":[18],"instruction":[19,71,81,135],"tuning":[20,72],"primarily":[21],"relies":[22],"on":[23],"teacher":[24],"or":[26],"human":[27],"intervention":[28],"generate":[30],"refine":[32],"the":[33,76,87,91],"instructions":[34],"responses":[36,93],"for":[37],"training,":[38],"which":[39,117],"are":[40],"costly,":[41],"non-sustainable,":[42],"may":[44],"lack":[45],"diversity.In":[46],"this":[47],"paper,":[48],"we":[49],"introduce":[50],"Mosaic":[51],"Instruction":[52],"Tuning":[53],"(Mosaic-IT),":[54],"human/modelfree":[56],"compositional":[57],"data":[58,73,82],"synthesis":[59],"method":[60],"that":[61],"can":[62],"efficiently":[63],"create":[64],"rich":[65],"diverse":[67],"augmentations":[68],"from":[69],"existing":[70],"enhance":[75],"LLMs.Mosaic-IT":[77],"randomly":[78],"concatenates":[79],"multiple":[80],"into":[83],"one":[84],"trains":[86],"model":[88],"produce":[90],"corresponding":[92],"predefined":[95],"higher-level":[96],"meta-instructions":[97],"strengthen":[99],"its":[100],"multi-step":[101],"instruction-following":[102],"formatfollowing":[104],"skills.Our":[105],"extensive":[106],"evaluations":[107],"demonstrate":[108],"superior":[110],"performance":[111,120],"training":[113,130],"efficiency":[114],"Mosaic-IT,":[116],"achieves":[118],"consistent":[119],"improvements":[121],"over":[122],"various":[123],"benchmarks":[124],"an":[126],"80%":[127],"reduction":[128],"in":[129],"costs":[131],"compared":[132],"original":[134],"tuning.":[136]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}