{"id":"https://openalex.org/W4412944775","doi":"https://doi.org/10.18653/v1/2025.acl-long.1095","title":"LESA: Learnable LLM Layer Scaling-Up","display_name":"LESA: Learnable LLM Layer Scaling-Up","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412944775","doi":"https://doi.org/10.18653/v1/2025.acl-long.1095"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.1095","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1095","pdf_url":"https://aclanthology.org/2025.acl-long.1095.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":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.1095.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5030820902","display_name":"Yifei Yang","orcid":"https://orcid.org/0000-0002-7481-1979"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yifei Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104247080","display_name":"Zouying Cao","orcid":"https://orcid.org/0000-0003-2200-5630"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zouying Cao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100697875","display_name":"Xiaofei Ma","orcid":"https://orcid.org/0000-0002-0641-586X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinbei Ma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027621155","display_name":"Yao Yao","orcid":"https://orcid.org/0000-0002-6749-1268"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yao Yao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030660667","display_name":"Zhi Chen","orcid":"https://orcid.org/0000-0002-8404-4937"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhi Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029082837","display_name":"Libo Qin","orcid":"https://orcid.org/0000-0002-3619-675X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Libo Qin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100457332","display_name":"Hai Zhao","orcid":"https://orcid.org/0000-0002-3392-2584"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hai Zhao","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":"22463","last_page":"22476"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13523","display_name":"Mathematics, Computing, and Information Processing","score":0.4724000096321106,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T13523","display_name":"Mathematics, Computing, and Information Processing","score":0.4724000096321106,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T10320","display_name":"Neural Networks and Applications","score":0.44530001282691956,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.4368000030517578,"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/scaling","display_name":"Scaling","score":0.5961179137229919},{"id":"https://openalex.org/keywords/layer","display_name":"Layer (electronics)","score":0.573631763458252},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5098906755447388},{"id":"https://openalex.org/keywords/materials-science","display_name":"Materials science","score":0.2794385552406311},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.1677757203578949},{"id":"https://openalex.org/keywords/composite-material","display_name":"Composite material","score":0.13775593042373657},{"id":"https://openalex.org/keywords/geometry","display_name":"Geometry","score":0.05982792377471924}],"concepts":[{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.5961179137229919},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.573631763458252},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5098906755447388},{"id":"https://openalex.org/C192562407","wikidata":"https://www.wikidata.org/wiki/Q228736","display_name":"Materials science","level":0,"score":0.2794385552406311},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.1677757203578949},{"id":"https://openalex.org/C159985019","wikidata":"https://www.wikidata.org/wiki/Q181790","display_name":"Composite material","level":1,"score":0.13775593042373657},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.05982792377471924}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.1095","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1095","pdf_url":"https://aclanthology.org/2025.acl-long.1095.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":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.1095","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1095","pdf_url":"https://aclanthology.org/2025.acl-long.1095.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":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412944775.pdf","grobid_xml":"https://content.openalex.org/works/W4412944775.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/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Training":[0],"Large":[1],"Language":[2],"Models":[3],"(LLMs)":[4],"from":[5,65],"scratch":[6],"requires":[7],"immense":[8],"computational":[9,118],"resources,":[10],"making":[11],"it":[12],"prohibitively":[13],"expensive.Model":[14],"scaling-up":[15,33],"offers":[16],"a":[17,56,87],"promising":[18],"solution":[19],"by":[20],"leveraging":[21],"the":[22,92,117],"parameters":[23,64,82,93],"of":[24],"smaller":[25],"models":[26],"to":[27,90],"create":[28],"larger":[29],"ones.However,":[30],"existing":[31,108],"depth":[32,61],"methods":[34],"rely":[35],"on":[36],"empirical":[37],"heuristic":[38],"rules":[39],"for":[40,60],"layer":[41,67],"duplication,":[42],"which":[43],"result":[44],"in":[45],"poorer":[46],"initialization":[47,100],"and":[48,68,101,131],"slower":[49],"convergence":[50],"during":[51,120],"continual":[52,121],"pre-training.We":[53],"propose":[54],"LESA,":[55],"novel":[57],"learnable":[58],"method":[59],"scaling-up.By":[62],"concatenating":[63],"each":[66],"applying":[69],"Singular":[70],"Value":[71],"Decomposition,":[72],"we":[73],"uncover":[74],"latent":[75],"patterns":[76],"between":[77,95],"layers,":[78,97],"suggesting":[79],"that":[80,105],"inter-layer":[81],"can":[83],"be":[84],"learned.LESA":[85],"uses":[86],"neural":[88],"network":[89],"predict":[91],"inserted":[94],"adjacent":[96],"enabling":[98],"better":[99],"faster":[102],"training.Experiments":[103],"show":[104],"LESA":[106],"outperforms":[107],"baselines,":[109],"achieving":[110],"superior":[111],"performance":[112],"with":[113],"less":[114],"than":[115],"half":[116],"cost":[119],"pre-training.Extensive":[122],"analyses":[123],"demonstrate":[124],"its":[125],"effectiveness":[126],"across":[127],"different":[128],"model":[129],"sizes":[130],"tasks.":[132],"1":[133]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}