{"id":"https://openalex.org/W4410061435","doi":"https://doi.org/10.1145/3676151.3719379","title":"Optimization Strategies for Enhancing Resource Efficiency in Transformers & Large Language Models","display_name":"Optimization Strategies for Enhancing Resource Efficiency in Transformers & Large Language Models","publication_year":2025,"publication_date":"2025-05-03","ids":{"openalex":"https://openalex.org/W4410061435","doi":"https://doi.org/10.1145/3676151.3719379"},"language":"en","primary_location":{"id":"doi:10.1145/3676151.3719379","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3676151.3719379","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 16th ACM/SPEC International Conference on Performance Engineering","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5031145084","display_name":"Tom Wallace","orcid":"https://orcid.org/0000-0001-8936-3163"},"institutions":[{"id":"https://openalex.org/I186803428","display_name":"Brock University","ror":"https://ror.org/056am2717","country_code":"CA","type":"education","lineage":["https://openalex.org/I186803428"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Tom Wallace","raw_affiliation_strings":["Brock University, St. Catharines, Ontario, Canada"],"raw_orcid":"https://orcid.org/0000-0001-8936-3163","affiliations":[{"raw_affiliation_string":"Brock University, St. Catharines, Ontario, Canada","institution_ids":["https://openalex.org/I186803428"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005525160","display_name":"Beatrice Ombuki-Berman","orcid":"https://orcid.org/0000-0002-7296-9510"},"institutions":[{"id":"https://openalex.org/I186803428","display_name":"Brock University","ror":"https://ror.org/056am2717","country_code":"CA","type":"education","lineage":["https://openalex.org/I186803428"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Beatrice Ombuki-Berman","raw_affiliation_strings":["Brock University, St. Catharines, Ontario, Canada"],"raw_orcid":"https://orcid.org/0000-0002-7296-9510","affiliations":[{"raw_affiliation_string":"Brock University, St. Catharines, Ontario, Canada","institution_ids":["https://openalex.org/I186803428"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5019537475","display_name":"Naser Ezzati\u2010Jivan","orcid":"https://orcid.org/0000-0003-1435-6297"},"institutions":[{"id":"https://openalex.org/I186803428","display_name":"Brock University","ror":"https://ror.org/056am2717","country_code":"CA","type":"education","lineage":["https://openalex.org/I186803428"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Naser Ezzati-Jivan","raw_affiliation_strings":["Brock University, St. Catharines, Ontario, Canada"],"raw_orcid":"https://orcid.org/0000-0003-1435-6297","affiliations":[{"raw_affiliation_string":"Brock University, St. Catharines, Ontario, Canada","institution_ids":["https://openalex.org/I186803428"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I186803428"],"apc_list":null,"apc_paid":null,"fwci":3.346,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.91091701,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":95,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"105","last_page":"112"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9995999932289124,"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":0.9995999932289124,"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":0.9944999814033508,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.9761000275611877,"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/computer-science","display_name":"Computer science","score":0.6372421979904175},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.61204993724823},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.18727043271064758},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.16159626841545105},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.13830751180648804}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6372421979904175},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.61204993724823},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.18727043271064758},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.16159626841545105},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.13830751180648804}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3676151.3719379","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3676151.3719379","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 16th ACM/SPEC International Conference on Performance Engineering","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/8","score":0.6600000262260437,"display_name":"Decent work and economic growth"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":4,"referenced_works":["https://openalex.org/W2946609015","https://openalex.org/W2946794439","https://openalex.org/W4396833699","https://openalex.org/W6893007900"],"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":{"Advancements":[0],"in":[1],"Natural":[2],"Language":[3],"Processing":[4],"are":[5],"heavily":[6],"reliant":[7],"on":[8,35,112],"Transformer":[9],"architectures,":[10],"whose":[11],"improvements":[12],"come":[13],"at":[14],"substantial":[15],"resource":[16],"costs":[17],"due":[18],"to":[19],"ever-growing":[20],"model":[21],"sizes.":[22],"This":[23],"study":[24],"explores":[25],"optimization":[26,79],"techniques,":[27],"including":[28],"quantization,":[29],"knowledge":[30],"distillation,":[31],"and":[32,37,64,74,106],"pruning,":[33,66],"focusing":[34],"energy":[36,50,117],"computational":[38],"efficiency":[39],"while":[40],"retaining":[41],"performance.":[42],"Among":[43],"standalone":[44],"methods,":[45,97],"4-Bit":[46],"quantization":[47],"significantly":[48],"reduces":[49],"use":[51],"with":[52],"minimal":[53],"accuracy":[54,75],"loss.":[55],"Hybrid":[56],"approaches,":[57],"like":[58],"NVIDIA's":[59],"Minitron":[60],"approach":[61],"combining":[62],"KD":[63],"structured":[65],"further":[67],"demonstrate":[68],"promising":[69],"trade-offs":[70],"between":[71],"size":[72],"reduction":[73],"retention.":[76],"A":[77],"novel":[78],"framework":[80,86],"is":[81],"introduced,":[82],"offering":[83],"a":[84,110],"flexible":[85],"for":[87,102],"comparing":[88],"various":[89],"methods.":[90],"Through":[91],"the":[92,113],"investigation":[93],"of":[94,116],"these":[95],"compression":[96],"we":[98],"provide":[99],"valuable":[100],"insights":[101],"developing":[103],"more":[104],"sustainable":[105],"efficient":[107],"LLMs,":[108],"shining":[109],"light":[111],"often-ignored":[114],"concern":[115],"efficiency.":[118]},"counts_by_year":[{"year":2025,"cited_by_count":2}],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}