{"id":"https://openalex.org/W7155998814","doi":"https://doi.org/10.48550/arxiv.2604.22583","title":"Adaptive Head Budgeting for Efficient Multi-Head Attention","display_name":"Adaptive Head Budgeting for Efficient Multi-Head Attention","publication_year":2026,"publication_date":"2026-04-24","ids":{"openalex":"https://openalex.org/W7155998814","doi":"https://doi.org/10.48550/arxiv.2604.22583"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.22583","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.22583","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.22583","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5134714571","display_name":"Bilal Faye","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Faye, Bilal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042941545","display_name":"Abdoulaye Mbaye","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mbaye, Abdoulaye","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047336564","display_name":"Hanane Azzag","orcid":"https://orcid.org/0000-0001-6876-0688"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Azzag, Hanane","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5023422312","display_name":"Mustapha Lebbah","orcid":"https://orcid.org/0000-0001-7245-6371"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lebbah, Mustapha","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5134714571"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.23440000414848328,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.23440000414848328,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.1834000051021576,"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"}},{"id":"https://openalex.org/T14347","display_name":"Big Data and Digital Economy","score":0.06400000303983688,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/transformer","display_name":"Transformer","score":0.5953999757766724},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5523999929428101},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.4462999999523163},{"id":"https://openalex.org/keywords/flops","display_name":"FLOPS","score":0.4004000127315521},{"id":"https://openalex.org/keywords/architecture","display_name":"Architecture","score":0.4002000093460083},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.3763999938964844}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7508999705314636},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5953999757766724},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5523999929428101},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49470001459121704},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4462999999523163},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.42289999127388},{"id":"https://openalex.org/C3826847","wikidata":"https://www.wikidata.org/wiki/Q188768","display_name":"FLOPS","level":2,"score":0.4004000127315521},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.4002000093460083},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3763999938964844},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.3314000070095062},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.31610000133514404},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.31520000100135803},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.30869999527931213},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.28700000047683716}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.22583","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.22583","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.22583","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.22583","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multi-head":[0],"attention":[1,10,51],"enables":[2],"Transformers":[3],"to":[4,68,122],"capture":[5],"diverse":[6],"representations,":[7],"but":[8],"all":[9],"heads":[11,52],"are":[12],"typically":[13],"activated":[14],"for":[15],"every":[16],"input,":[17],"regardless":[18],"of":[19,108],"task":[20],"complexity.":[21],"For":[22],"coarse-grained":[23],"tasks":[24,93],"such":[25],"as":[26,118],"text":[27,91],"classification,":[28],"where":[29],"relevant":[30],"information":[31],"is":[32],"often":[33],"global,":[34],"this":[35],"fixed":[36],"allocation":[37,117],"can":[38],"introduce":[39,80],"unnecessary":[40],"computation.":[41],"We":[42],"propose":[43],"BudgetFormer,":[44],"a":[45,54,61,65,81],"Transformer":[46,124],"architecture":[47],"that":[48,84,95],"dynamically":[49],"allocates":[50],"on":[53,90],"per-input":[55],"basis.":[56],"The":[57],"model":[58],"learns":[59],"both":[60],"head":[62,77,116],"budget":[63],"and":[64,87,99,126],"relevance":[66],"distribution":[67],"select":[69],"the":[70,106],"most":[71],"informative":[72],"heads.":[73],"To":[74],"support":[75],"effective":[76,120],"selection,":[78],"we":[79],"training":[82],"strategy":[83],"balances":[85],"exploration":[86],"exploitation.":[88],"Experiments":[89],"classification":[92],"show":[94],"BudgetFormer":[96],"reduces":[97],"FLOPs":[98],"memory":[100],"usage":[101],"while":[102],"matching":[103],"or":[104],"surpassing":[105],"performance":[107],"standard":[109],"multi-head":[110],"attention.":[111],"These":[112],"results":[113],"highlight":[114],"adaptive":[115],"an":[119],"approach":[121],"improving":[123],"efficiency":[125],"performance.":[127]},"counts_by_year":[],"updated_date":"2026-06-05T09:01:59.212387","created_date":"2026-04-28T00:00:00"}