{"id":"https://openalex.org/W4414019041","doi":"https://doi.org/10.1145/3766068","title":"MaGrIP: Magnitude and Gradient-Informed Pruning for Task-Agnostic Large Language Models","display_name":"MaGrIP: Magnitude and Gradient-Informed Pruning for Task-Agnostic Large Language Models","publication_year":2025,"publication_date":"2025-09-05","ids":{"openalex":"https://openalex.org/W4414019041","doi":"https://doi.org/10.1145/3766068"},"language":"en","primary_location":{"id":"doi:10.1145/3766068","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3766068","pdf_url":null,"source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1145/3766068","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5061729841","display_name":"Uttej Kallakuri","orcid":"https://orcid.org/0009-0004-1150-3903"},"institutions":[{"id":"https://openalex.org/I145311948","display_name":"Johns Hopkins University","ror":"https://ror.org/00za53h95","country_code":"US","type":"education","lineage":["https://openalex.org/I145311948"]},{"id":"https://openalex.org/I2799853436","display_name":"Johns Hopkins Medicine","ror":"https://ror.org/037zgn354","country_code":"US","type":"healthcare","lineage":["https://openalex.org/I2799853436"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Uttej Kallakuri","raw_affiliation_strings":["Electrical and Computer Engineering, Johns Hopkins University","Electrical and Computer Engineering, Johns Hopkins University, Baltimore, United States"],"raw_orcid":"https://orcid.org/0009-0004-1150-3903","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, Johns Hopkins University","institution_ids":["https://openalex.org/I145311948"]},{"raw_affiliation_string":"Electrical and Computer Engineering, Johns Hopkins University, Baltimore, United States","institution_ids":["https://openalex.org/I145311948","https://openalex.org/I2799853436"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023630054","display_name":"Edward Humes","orcid":"https://orcid.org/0009-0002-3945-0116"},"institutions":[{"id":"https://openalex.org/I79272384","display_name":"University of Maryland, Baltimore County","ror":"https://ror.org/02qskvh78","country_code":"US","type":"education","lineage":["https://openalex.org/I79272384"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Edward Humes","raw_affiliation_strings":["Computer Science and Electrical Engineering, University of Maryland Baltimore County","Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, United States"],"raw_orcid":"https://orcid.org/0009-0002-3945-0116","affiliations":[{"raw_affiliation_string":"Computer Science and Electrical Engineering, University of Maryland Baltimore County","institution_ids":["https://openalex.org/I79272384"]},{"raw_affiliation_string":"Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, United States","institution_ids":["https://openalex.org/I79272384"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082854741","display_name":"Hasib-Al Rashid","orcid":"https://orcid.org/0000-0002-9983-6929"},"institutions":[{"id":"https://openalex.org/I79272384","display_name":"University of Maryland, Baltimore County","ror":"https://ror.org/02qskvh78","country_code":"US","type":"education","lineage":["https://openalex.org/I79272384"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hasib-Al Rashid","raw_affiliation_strings":["Computer Science and Electrical Engineering, University of Maryland Baltimore County","Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, United States"],"raw_orcid":"https://orcid.org/0000-0002-9983-6929","affiliations":[{"raw_affiliation_string":"Computer Science and Electrical Engineering, University of Maryland Baltimore County","institution_ids":["https://openalex.org/I79272384"]},{"raw_affiliation_string":"Computer Science and Electrical Engineering, University of Maryland Baltimore County, Baltimore, United States","institution_ids":["https://openalex.org/I79272384"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5084010501","display_name":"Tinoosh Mohsenin","orcid":"https://orcid.org/0000-0001-5551-2124"},"institutions":[{"id":"https://openalex.org/I145311948","display_name":"Johns Hopkins University","ror":"https://ror.org/00za53h95","country_code":"US","type":"education","lineage":["https://openalex.org/I145311948"]},{"id":"https://openalex.org/I79272384","display_name":"University of Maryland, Baltimore County","ror":"https://ror.org/02qskvh78","country_code":"US","type":"education","lineage":["https://openalex.org/I79272384"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tinoosh Mohsenin","raw_affiliation_strings":["Johns Hopkins University","University of Maryland Baltimore County","Johns Hopkins University, Baltimore, United States","University of Maryland Baltimore County, Baltimore, United States"],"raw_orcid":"https://orcid.org/0000-0001-5551-2124","affiliations":[{"raw_affiliation_string":"Johns Hopkins University","institution_ids":["https://openalex.org/I145311948"]},{"raw_affiliation_string":"University of Maryland Baltimore County","institution_ids":["https://openalex.org/I79272384"]},{"raw_affiliation_string":"Johns Hopkins University, Baltimore, United States","institution_ids":["https://openalex.org/I145311948"]},{"raw_affiliation_string":"University of Maryland Baltimore County, Baltimore, United States","institution_ids":["https://openalex.org/I79272384"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.5047,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.90681055,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":91,"max":97},"biblio":{"volume":"25","issue":"3","first_page":"1","last_page":"32"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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":0.9998999834060669,"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":0.9991999864578247,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9768999814987183,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.6415655612945557},{"id":"https://openalex.org/keywords/magnitude","display_name":"Magnitude (astronomy)","score":0.6342995762825012},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.5951088666915894},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5914610028266907},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4275325536727905},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.41792532801628113},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.40796324610710144},{"id":"https://openalex.org/keywords/biology","display_name":"Biology","score":0.12610110640525818},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.11277356743812561},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.06321540474891663},{"id":"https://openalex.org/keywords/astrophysics","display_name":"Astrophysics","score":0.05779165029525757}],"concepts":[{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.6415655612945557},{"id":"https://openalex.org/C126691448","wikidata":"https://www.wikidata.org/wiki/Q2028919","display_name":"Magnitude (astronomy)","level":2,"score":0.6342995762825012},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5951088666915894},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5914610028266907},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4275325536727905},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41792532801628113},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.40796324610710144},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.12610110640525818},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.11277356743812561},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.06321540474891663},{"id":"https://openalex.org/C44870925","wikidata":"https://www.wikidata.org/wiki/Q37547","display_name":"Astrophysics","level":1,"score":0.05779165029525757},{"id":"https://openalex.org/C6557445","wikidata":"https://www.wikidata.org/wiki/Q173113","display_name":"Agronomy","level":1,"score":0.0},{"id":"https://openalex.org/C201995342","wikidata":"https://www.wikidata.org/wiki/Q682496","display_name":"Systems engineering","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3766068","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3766068","pdf_url":null,"source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1145/3766068","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3766068","pdf_url":null,"source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":25,"referenced_works":["https://openalex.org/W569478347","https://openalex.org/W2890894339","https://openalex.org/W2911489562","https://openalex.org/W2965862774","https://openalex.org/W2970454332","https://openalex.org/W3101066076","https://openalex.org/W3130554079","https://openalex.org/W3174657338","https://openalex.org/W3177265267","https://openalex.org/W3194676777","https://openalex.org/W3215698464","https://openalex.org/W4281651027","https://openalex.org/W4366341968","https://openalex.org/W4367848206","https://openalex.org/W4385573119","https://openalex.org/W4386546945","https://openalex.org/W4387596623","https://openalex.org/W4389576338","https://openalex.org/W4392904185","https://openalex.org/W4393145114","https://openalex.org/W4393147854","https://openalex.org/W4405056149","https://openalex.org/W4405717961","https://openalex.org/W4410823267","https://openalex.org/W4411713754"],"related_works":["https://openalex.org/W2376418092","https://openalex.org/W2072983018","https://openalex.org/W1016952678","https://openalex.org/W2257644995","https://openalex.org/W1513444457","https://openalex.org/W4311345787","https://openalex.org/W2188141918","https://openalex.org/W3185336960","https://openalex.org/W2042652790","https://openalex.org/W3098003361"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"become":[5],"foundational":[6],"tools":[7],"in":[8,32,90,262,268],"natural":[9],"language":[10],"processing,":[11],"achieving":[12],"state-of-the-art":[13,92],"performance":[14,178,269],"across":[15,179],"a":[16,53,65,165,259,265],"variety":[17],"of":[18,61,88,101,120],"tasks.":[19],"However,":[20],"their":[21],"immense":[22],"size":[23,152,166],"and":[24,39,48,59,70,131,186,208,244,264],"computational":[25,99],"requirements":[26,133],"make":[27],"them":[28],"impractical":[29],"for":[30,56,160,205,216,236],"deployment":[31],"resource-constrained":[33],"environments,":[34],"such":[35,182],"as":[36,183],"edge":[37],"devices":[38],"embedded":[40],"systems.":[41],"In":[42,118],"this":[43],"work,":[44],"we":[45],"introduce":[46],"Magnitude":[47],"Gradient-Informed":[49],"Pruning":[50],"(MaGrIP)":[51],",":[52],"novel":[54],"framework":[55],"task-agnostic":[57],"pruning":[58,124,196,227],"compression":[60,95,243],"LLMs.":[62,273],"MaGrIP":[63,89,163,174],"employs":[64],"dual-threshold":[66],"strategy":[67],"combining":[68],"magnitude-":[69],"gradient-based":[71],"saliency":[72],"measures":[73],"to":[74,110,156,171,203,214,234,271],"efficiently":[75],"prune":[76],"redundant":[77],"neurons":[78],"while":[79,134,159],"retaining":[80],"task":[81],"performance.":[82],"Our":[83],"results":[84],"demonstrate":[85],"the":[86,97,102,143,150,189,206,217,220,237,240,253],"effectiveness":[87],"compressing":[91],"models.":[93],"The":[94],"reduced":[96],"total":[98,151],"complexity":[100],"FFN":[103],"layers":[104],"from":[105,153,168],"\\(\\mathcal":[106,111],"{O}(d":[107],"\\cdot":[108,115],"h)\\)":[109,116],"{O}((d":[112],"-":[113],"q)":[114],".":[117],"terms":[119],"model":[121,129,193],"size,":[122],"our":[123,147],"approach":[125],"significantly":[126],"reduces":[127,149],"both":[128],"parameters":[130],"storage":[132],"maintaining":[135],"competitive":[136],"perplexity":[137],"scores":[138],"evaluated":[139],"on":[140,200,211,231,252],"WikiText-2.":[141],"For":[142],"Gemma":[144,161,191],"7B":[145,192],"model,":[146],"method":[148],"28":[154],"GB":[155,170],"5":[157],"GB,":[158],"2B,":[162],"achieves":[164],"reduction":[167],"8":[169],"1.5":[172],"GB.":[173],"furthermore":[175],"exhibits":[176],"robust":[177],"multiple":[180],"benchmarks,":[181],"BOOLQ,":[184],"ARC-E,":[185],"CSQA.":[187],"Specifically,":[188],"pruned":[190,221],"at":[194,225],"50%":[195,226],"achieved":[197,228,258],"59.26%":[198],"accuracy":[199,210,230],"ARC-E":[201,232],"compared":[202,213,233,270],"81.06%":[204],"baseline,":[207,238],"64.74%":[209],"BoolQ":[212],"59.98%":[215],"baseline.":[218],"Similarly,":[219],"Llama":[222],"3":[223],"8B":[224],"46.76%":[229],"77.57%":[235],"reflecting":[239],"tradeoff":[241],"between":[242],"accuracy.":[245],"LLMs":[246],"compressed":[247],"using":[248],"MaGrIP,":[249],"when":[250],"deployed":[251],"Nvidia":[254],"Jetson":[255],"Orin":[256],"Nano,":[257],"2.16\u00d7":[260],"improvement":[261,267],"throughput":[263],"2.3\u00d7":[266],"baseline":[272]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":1}],"updated_date":"2026-07-27T08:26:11.824852","created_date":"2025-10-10T00:00:00"}