{"id":"https://openalex.org/W7140318501","doi":"https://doi.org/10.1109/access.2026.3677560","title":"SHARP: Structured Hierarchical Attention Rank Projection for Efficient Language Model Distillation","display_name":"SHARP: Structured Hierarchical Attention Rank Projection for Efficient Language Model Distillation","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140318501","doi":"https://doi.org/10.1109/access.2026.3677560"},"language":"en","primary_location":{"id":"doi:10.1109/access.2026.3677560","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2026.3677560","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1109/access.2026.3677560","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5058572236","display_name":"Jieui Kang","orcid":"https://orcid.org/0009-0000-7691-0930"},"institutions":[{"id":"https://openalex.org/I138925566","display_name":"Ewha Womans University","ror":"https://ror.org/053fp5c05","country_code":"KR","type":"education","lineage":["https://openalex.org/I138925566"]}],"countries":["KR"],"is_corresponding":true,"raw_author_name":"Jieui Kang","raw_affiliation_strings":["Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0009-0000-7691-0930","affiliations":[{"raw_affiliation_string":"Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea","institution_ids":["https://openalex.org/I138925566"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Eunjeong Yoo","orcid":"https://orcid.org/0009-0006-2022-5460"},"institutions":[{"id":"https://openalex.org/I138925566","display_name":"Ewha Womans University","ror":"https://ror.org/053fp5c05","country_code":"KR","type":"education","lineage":["https://openalex.org/I138925566"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Eunjeong Yoo","raw_affiliation_strings":["Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0009-0006-2022-5460","affiliations":[{"raw_affiliation_string":"Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea","institution_ids":["https://openalex.org/I138925566"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104299689","display_name":"Soeun Choi","orcid":null},"institutions":[{"id":"https://openalex.org/I138925566","display_name":"Ewha Womans University","ror":"https://ror.org/053fp5c05","country_code":"KR","type":"education","lineage":["https://openalex.org/I138925566"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Soeun Choi","raw_affiliation_strings":["Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0009-0009-2458-3048","affiliations":[{"raw_affiliation_string":"Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea","institution_ids":["https://openalex.org/I138925566"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123746621","display_name":"Yeonhui Kim","orcid":null},"institutions":[{"id":"https://openalex.org/I138925566","display_name":"Ewha Womans University","ror":"https://ror.org/053fp5c05","country_code":"KR","type":"education","lineage":["https://openalex.org/I138925566"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Yeonhee Kim","raw_affiliation_strings":["Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0000-0003-1882-5230","affiliations":[{"raw_affiliation_string":"Artificial Intelligence Convergence, Ewha Womans University, Seoul, South Korea","institution_ids":["https://openalex.org/I138925566"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5013362846","display_name":"Jaehyeong Sim","orcid":"https://orcid.org/0000-0001-8722-8486"},"institutions":[{"id":"https://openalex.org/I138925566","display_name":"Ewha Womans University","ror":"https://ror.org/053fp5c05","country_code":"KR","type":"education","lineage":["https://openalex.org/I138925566"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jaehyeong Sim","raw_affiliation_strings":["Department of Computer Science and Engineering, Ewha Womans University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0000-0001-8722-8486","affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, Ewha Womans University, Seoul, South Korea","institution_ids":["https://openalex.org/I138925566"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5058572236"],"corresponding_institution_ids":["https://openalex.org/I138925566"],"apc_list":{"value":1850,"currency":"USD","value_usd":1850},"apc_paid":{"value":1850,"currency":"USD","value_usd":1850},"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.58305454,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"14","issue":null,"first_page":"57983","last_page":"57996"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.3822999894618988,"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":0.3822999894618988,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.12800000607967377,"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":0.0949999988079071,"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/projection","display_name":"Projection (relational algebra)","score":0.6261000037193298},{"id":"https://openalex.org/keywords/rank","display_name":"Rank (graph theory)","score":0.5504999756813049},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.44119998812675476},{"id":"https://openalex.org/keywords/data-modeling","display_name":"Data modeling","score":0.4032999873161316},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.36329999566078186},{"id":"https://openalex.org/keywords/hierarchical-database-model","display_name":"Hierarchical database model","score":0.3594000041484833}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8001999855041504},{"id":"https://openalex.org/C57493831","wikidata":"https://www.wikidata.org/wiki/Q3134666","display_name":"Projection (relational algebra)","level":2,"score":0.6261000037193298},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5837000012397766},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.5504999756813049},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.44119998812675476},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.4032999873161316},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.36329999566078186},{"id":"https://openalex.org/C144986985","wikidata":"https://www.wikidata.org/wiki/Q871236","display_name":"Hierarchical database model","level":2,"score":0.3594000041484833},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3240000009536743},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3228999972343445},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.32260000705718994},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.30709999799728394},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.290800005197525},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2815999984741211},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2802000045776367},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.26089999079704285},{"id":"https://openalex.org/C2993807640","wikidata":"https://www.wikidata.org/wiki/Q103709453","display_name":"Attention network","level":2,"score":0.25189998745918274},{"id":"https://openalex.org/C3018263672","wikidata":"https://www.wikidata.org/wiki/Q1296251","display_name":"Efficient algorithm","level":2,"score":0.2500999867916107}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/access.2026.3677560","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2026.3677560","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},{"id":"pmh:oai:doaj.org/article:bf397f1f7fe5480098cf0846eea92dc4","is_oa":true,"landing_page_url":"https://doaj.org/article/bf397f1f7fe5480098cf0846eea92dc4","pdf_url":null,"source":{"id":"https://openalex.org/S4306401280","display_name":"DOAJ (DOAJ: Directory of Open Access Journals)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-sa","license_id":"https://openalex.org/licenses/cc-by-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"IEEE Access, Vol 14, Pp 57983-57996 (2026)","raw_type":"article"}],"best_oa_location":{"id":"doi:10.1109/access.2026.3677560","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2026.3677560","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.41362395882606506}],"awards":[],"funders":[{"id":"https://openalex.org/F4320335839","display_name":"National Institute of Information and Communications Technology","ror":"https://ror.org/016bgq349"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Knowledge":[0],"distillation":[1,89,163],"has":[2],"emerged":[3],"as":[4,29],"a":[5,53,87,105,155],"crucial":[6],"technique":[7],"for":[8,59,185,217],"compressing":[9],"large":[10,222],"language":[11,135,140,223],"models":[12,146],"into":[13,95,104],"more":[14],"deployable":[15],"versions.":[16],"While":[17],"existing":[18,162],"approaches":[19],"focus":[20],"on":[21,132,173],"transferring":[22],"knowledge":[23,219],"at":[24],"length-based":[25],"linguistic":[26],"granularities":[27,51],"such":[28],"tokens,":[30],"phrases,":[31],"and":[32,98,101,138],"sequences,":[33],"they":[34,57],"often":[35],"fail":[36],"to":[37,66,110,150,154,177,200],"capture":[38],"the":[39,60,178,186],"intrinsic":[40],"hierarchical":[41],"attention":[42,93],"mechanisms":[43],"of":[44,169,198],"modern":[45],"transformer":[46],"architectures.":[47],"Furthermore,":[48],"when":[49],"multiple":[50],"share":[52],"single":[54],"representation":[55,73],"space,":[56],"compete":[58],"same":[61],"capacity,":[62,121],"causing":[63],"complementary":[64],"information":[65],"be":[67],"lost.":[68],"We":[69],"term":[70],"this":[71,77],"phenomenon":[72],"collapse.":[74],"To":[75],"address":[76],"limitation,":[78],"we":[79],"propose":[80],"SHARP":[81,159,193],"(Structured":[82],"Hierarchical":[83],"Attention":[84],"Rank":[85],"Projection),":[86],"novel":[88],"framework":[90],"that":[91,209],"decomposes":[92],"patterns":[94],"token-level,":[96],"head-level,":[97],"layer-level":[99],"representations":[100],"projects":[102],"each":[103,116],"dedicated":[106],"orthogonal":[107,210],"subspace":[108],"prior":[109],"distillation.":[111],"This":[112],"mathematical":[113],"decomposition":[114],"provides":[115],"granularity":[117],"with":[118,144,181],"independent":[119],"representational":[120],"preventing":[122],"interference":[123],"while":[124],"enabling":[125],"parallel":[126],"optimization":[127],"pathways.":[128],"Through":[129],"extensive":[130],"experiments":[131],"both":[133],"natural":[134,139],"generation":[136],"(NLG)":[137],"understanding":[141],"(NLU)":[142],"tasks":[143,175],"teacher":[145,188],"ranging":[147],"from":[148],"350M":[149],"6.7B":[151,187],"parameters":[152],"distilled":[153],"125M":[156],"parameter":[157],"student,":[158],"consistently":[160],"outperforms":[161],"methods.":[164],"It":[165],"achieves":[166],"an":[167,214],"average":[168],"5.2%":[170],"perplexity":[171],"reduction":[172],"NLG":[174],"relative":[176],"strongest":[179],"baseline,":[180],"gains":[182],"reaching":[183],"7.8%":[184],"configuration.":[189],"On":[190],"NLU":[191],"tasks,":[192],"shows":[194],"consistent":[195],"accuracy":[196],"improvements":[197],"1.5":[199],"3.2%":[201],"across":[202],"all":[203],"model":[204,224],"scales.":[205],"These":[206],"results":[207],"confirm":[208],"capacity":[211],"allocation":[212],"is":[213],"effective":[215],"principle":[216],"multi-granularity":[218],"transfer":[220],"in":[221],"compression.":[225]},"counts_by_year":[],"updated_date":"2026-05-06T08:25:59.206177","created_date":"2026-03-26T00:00:00"}