{"id":"https://openalex.org/W4415797274","doi":"https://doi.org/10.52202/079017-2149","title":"Knowledge Composition using Task Vectors with Learned Anisotropic Scaling","display_name":"Knowledge Composition using Task Vectors with Learned Anisotropic Scaling","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W4415797274","doi":"https://doi.org/10.52202/079017-2149"},"language":"en","primary_location":{"id":"doi:10.52202/079017-2149","doi":"https://doi.org/10.52202/079017-2149","is_oa":false,"landing_page_url":"https://doi.org/10.52202/079017-2149","pdf_url":null,"source":{"id":"https://openalex.org/S4363606243","display_name":"neural information processing systems","issn_l":"1049-5258","issn":["1049-5258"],"is_oa":false,"is_in_doaj":false,"is_core":true,"listed_in":["cwts-core","norway-2"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Advances in Neural Information Processing Systems 37","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2407.02880","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5061019888","display_name":"Frederic Z. Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Frederic Z. Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039237147","display_name":"Paul S. Albert","orcid":"https://orcid.org/0000-0003-1658-1068","observed_orcids":["https://orcid.org/0000-0003-1658-1068"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Paul Albert","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5000064645","display_name":"Cristian Rodr\u00edguez-Opazo","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cristian Rodriguez-Opazo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5028024287","display_name":"Anton van den Hengel","orcid":"https://orcid.org/0000-0003-3027-8364","observed_orcids":["https://orcid.org/0000-0003-3027-8364"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anton Van Den Hengel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5058061726","display_name":"Ehsan Abbasnejad","orcid":"https://orcid.org/0000-0002-7126-317X","observed_orcids":["https://orcid.org/0000-0002-7126-317X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ehsan Abbasnejad","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":1.3433,"has_fulltext":true,"cited_by_count":6,"citation_normalized_percentile":{"value":0.84446143,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"67319","last_page":"67354"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.4672999978065491,"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":0.4672999978065491,"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":0.0812000036239624,"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/T10320","display_name":"Neural Networks and Applications","score":0.07199999690055847,"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/knowledge-composition","display_name":"knowledge composition","score":0.9879999756813049},{"id":"https://openalex.org/keywords/task-vectors","display_name":"task vectors","score":0.9589999914169312},{"id":"https://openalex.org/keywords/anisotropic-scaling","display_name":"anisotropic scaling","score":0.9139999747276306},{"id":"https://openalex.org/keywords/knowledge-distillation","display_name":"knowledge distillation","score":0.4970000088214874}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.553600013256073},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5157999992370605},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.505299985408783},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.4212999939918518},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.31189998984336853},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3005000054836273},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2922999858856201},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.2890999913215637},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2865000069141388},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26489999890327454},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.2624000012874603},{"id":"https://openalex.org/C161301231","wikidata":"https://www.wikidata.org/wiki/Q3478658","display_name":"Knowledge representation and reasoning","level":2,"score":0.26100000739097595},{"id":"https://openalex.org/C115925183","wikidata":"https://www.wikidata.org/wiki/Q1412694","display_name":"Knowledge-based systems","level":2,"score":0.2597000002861023}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.52202/079017-2149","doi":"https://doi.org/10.52202/079017-2149","is_oa":false,"landing_page_url":"https://doi.org/10.52202/079017-2149","pdf_url":null,"source":{"id":"https://openalex.org/S4363606243","display_name":"neural information processing systems","issn_l":"1049-5258","issn":["1049-5258"],"is_oa":false,"is_in_doaj":false,"is_core":true,"listed_in":["cwts-core","norway-2"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Advances in Neural Information Processing Systems 37","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2407.02880","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2407.02880","pdf_url":"https://arxiv.org/pdf/2407.02880","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.2407.02880","doi":"https://doi.org/10.48550/arxiv.2407.02880","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2407.02880","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:2407.02880","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2407.02880","pdf_url":"https://arxiv.org/pdf/2407.02880","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":[{"id":"https://openalex.org/G8241165388","display_name":"Learning to Reason in Reinforcement Learning","funder_award_id":"DP240103278","funder_id":"https://openalex.org/F4320334704","funder_display_name":"Australian Research Council"}],"funders":[{"id":"https://openalex.org/F4320315885","display_name":"Australian Government","ror":"https://ror.org/0314h5y94"},{"id":"https://openalex.org/F4320334704","display_name":"Australian Research Council","ror":"https://ror.org/05mmh0f86"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415797274.pdf","grobid_xml":"https://content.openalex.org/works/W4415797274.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7134846041","https://openalex.org/W4417300295","https://openalex.org/W7162646997","https://openalex.org/W4409693013","https://openalex.org/W7161784540","https://openalex.org/W7219155291","https://openalex.org/W4402670870","https://openalex.org/W4399597988","https://openalex.org/W7129176360","https://openalex.org/W7131651195"],"abstract_inverted_index":{"Pre-trained":[0],"models":[1],"produce":[2],"strong":[3],"generic":[4],"representations":[5,52],"that":[6,40,103,123,188],"can":[7,46,86,242],"be":[8,47,87,197],"adapted":[9],"via":[10],"fine-tuning.":[11,32],"The":[12,33],"learned":[13,110,153,190],"weight":[14],"difference":[15],"relative":[16],"to":[17,49,67,89,196,220,225,240],"the":[18,27,117,129,142,151,157,166,230,244,249,257],"pre-trained":[19,134],"model,":[20],"known":[21],"as":[22,261],"a":[23,138,262],"task":[24,36,63,73,118,172,194,206,237],"vector,":[25],"characterises":[26],"direction":[28],"and":[29,65,81,93,176,216,247,269],"stride":[30],"of":[31,35,62,72,133,147,162,168,251,259],"significance":[34],"vectors":[37,64,195,238],"is":[38,217],"such":[39,84,124],"simple":[41],"arithmetic":[42],"operations":[43],"on":[44,59,159],"them":[45],"used":[48,88],"combine":[50],"diverse":[51],"from":[53],"different":[54,109,236],"domains.":[55],"This":[56],"paper":[57],"builds":[58],"these":[60],"properties":[61],"aims":[66],"answer":[68],"(1)":[69,189],"whether":[70],"components":[71],"vectors,":[74],"particularly":[75,265],"parameter":[76,106,148,233],"blocks,":[77],"exhibit":[78],"similar":[79],"characteristics,":[80],"(2)":[82,205],"how":[83],"blocks":[85,107,149,234],"enhance":[90],"knowledge":[91,252],"composition":[92,146,208],"transfer.":[94,253],"To":[95],"this":[96],"end,":[97],"we":[98,186,255],"introduce":[99],"aTLAS,":[100],"an":[101],"algorithm":[102],"linearly":[104],"combines":[105],"with":[108,136,179,210,266],"coefficients,":[111],"resulting":[112],"in":[113,171,203],"anisotropic":[114,191],"scaling":[115,192],"at":[116],"vector":[119,207],"level.":[120],"We":[121,164],"show":[122,187,256],"linear":[125],"combinations":[126],"explicitly":[127],"exploit":[128],"low":[130],"intrinsic":[131],"dimensionality":[132],"models,":[135],"only":[137],"few":[139],"coefficients":[140],"being":[141],"learnable":[143],"parameters.":[144],"Furthermore,":[145],"leverages":[150],"already":[152],"representations,":[154],"thereby":[155],"reducing":[156],"dependency":[158],"large":[160],"amounts":[161],"data.":[163],"demonstrate":[165,270],"effectiveness":[167],"our":[169],"method":[170],"arithmetic,":[173],"few-shot":[174],"recognition":[175],"test-time":[177],"adaptation,":[178],"supervised":[180],"or":[181,212],"unsupervised":[182],"objectives.":[183],"In":[184],"particular,":[185],"allows":[193],"more":[198],"disentangled,":[199],"causing":[200],"less":[201,218,267],"interference":[202],"composition;":[204],"excels":[209],"scarce":[211],"no":[213],"labeled":[214],"data":[215],"prone":[219],"domain":[221],"shift,":[222],"thus":[223],"leading":[224],"better":[226],"generalisability;":[227],"(3)":[228],"mixing":[229],"most":[231],"informative":[232],"across":[235],"prior":[239],"training":[241],"reduce":[243],"memory":[245],"footprint":[246],"improve":[248],"flexibility":[250],"Moreover,":[254],"potential":[258],"aTLAS":[260],"PEFT":[263],"method,":[264],"data,":[268],"its":[271],"scalibility.":[272]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":2}],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2025-11-03T00:00:00"}