{"id":"https://openalex.org/W7137818810","doi":"https://doi.org/10.48550/arxiv.2603.13418","title":"GPrune-LLM: Generalization-Aware Structured Pruning for Large Language Models","display_name":"GPrune-LLM: Generalization-Aware Structured Pruning for Large Language Models","publication_year":2026,"publication_date":"2026-03-12","ids":{"openalex":"https://openalex.org/W7137818810","doi":"https://doi.org/10.48550/arxiv.2603.13418"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.13418","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13418","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":null,"license_id":null,"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.2603.13418","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126714860","display_name":"Xiaoyun Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Liu, Xiaoyun","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129700349","display_name":"Divya Saxena","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saxena, Divya","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129696057","display_name":"Jiannong Cao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cao, Jiannong","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129718389","display_name":"Yuqing Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Yuqing","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129719382","display_name":"Yiying Dong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Yiying","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5108993989","display_name":"Penghui Ruan","orcid":"https://orcid.org/0000-0003-0238-299X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ruan, Penghui","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5126714860"],"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/T10028","display_name":"Topic Modeling","score":0.6092000007629395,"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":0.6092000007629395,"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":0.06729999929666519,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.05389999970793724,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.7882000207901001},{"id":"https://openalex.org/keywords/ranking","display_name":"Ranking (information retrieval)","score":0.7735999822616577},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.7091000080108643},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.7074000239372253},{"id":"https://openalex.org/keywords/sensitivity","display_name":"Sensitivity (control systems)","score":0.5170000195503235},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.3984000086784363},{"id":"https://openalex.org/keywords/signal","display_name":"SIGNAL (programming language)","score":0.3571000099182129},{"id":"https://openalex.org/keywords/performance-metric","display_name":"Performance metric","score":0.33180001378059387}],"concepts":[{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.7882000207901001},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.7735999822616577},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7188000082969666},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.7091000080108643},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.7074000239372253},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5196999907493591},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.5170000195503235},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.44209998846054077},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.3984000086784363},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3571000099182129},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.33180001378059387},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.32179999351501465},{"id":"https://openalex.org/C2781390188","wikidata":"https://www.wikidata.org/wiki/Q25203449","display_name":"Spike (software development)","level":2,"score":0.30869999527931213},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.2939000129699707},{"id":"https://openalex.org/C42812","wikidata":"https://www.wikidata.org/wiki/Q1082910","display_name":"Partition (number theory)","level":2,"score":0.2892000079154968},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.2847000062465668},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2825999855995178},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2786000072956085},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.2734000086784363},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.27309998869895935},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.2572999894618988}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.13418","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13418","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.13418","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13418","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":null,"license_id":null,"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":{"Structured":[0],"pruning":[1,140],"is":[2,181],"widely":[3],"used":[4],"to":[5,95,158,170,185],"compress":[6],"large":[7],"language":[8],"models":[9],"(LLMs),":[10],"yet":[11],"its":[12],"effectiveness":[13],"depends":[14],"heavily":[15],"on":[16,29,68,92,119,215],"neuron":[17,24,146],"importance":[18,25,108,216],"estimation.":[19],"Most":[20],"existing":[21,76],"methods":[22],"estimate":[23],"from":[26],"activation":[27,124],"statistics":[28],"a":[30,83,137],"single":[31],"calibration":[32,36,93,120],"dataset,":[33],"which":[34],"introduces":[35],"bias":[37],"and":[38,59,173,212],"degrades":[39],"downstream":[40,199],"cross-task":[41],"generalization.":[42],"We":[43,151],"observe":[44],"that":[45,89,116,142],"neurons":[46,53,61,81,88,100,115,154],"exhibit":[47],"heterogeneous":[48],"distribution":[49,171],"sensitivity,":[50],"with":[51],"distribution-robust":[52,99],"maintaining":[54],"consistent":[55,203],"rankings":[56],"across":[57,197],"datasets":[58],"distribution-sensitive":[60,87],"showing":[62],"high":[63,210],"cross-dataset":[64],"ranking":[65,79,160],"variance.":[66],"Based":[67],"this,":[69],"we":[70,134,183,190],"identify":[71],"two":[72],"structural":[73],"limitations":[74],"in":[75,148,205],"methods.":[77],"First,":[78],"all":[80],"within":[82],"shared":[84],"space":[85],"causes":[86],"strongly":[90],"activate":[91,118],"inputs":[94],"dominate,":[96],"crowding":[97],"out":[98],"critical":[101],"for":[102,126,145],"out-of-distribution":[103],"tasks.":[104],"Second,":[105],"applying":[106],"activation-based":[107,164,179],"metrics":[109],"uniformly":[110],"can":[111],"be":[112],"unreliable.":[113],"Distribution-sensitive":[114],"infrequently":[117],"data":[121],"receive":[122],"insufficient":[123],"signal":[125],"accurate":[127],"local":[128],"ranking.":[129],"To":[130],"address":[131],"these":[132],"limitations,":[133],"propose":[135],"GPrune-LLM,":[136],"generalization-aware":[138],"structured":[139],"framework":[141],"explicitly":[143],"accounts":[144],"differences":[147],"cross-distribution":[149],"behavior.":[150],"first":[152],"partition":[153],"into":[155],"behavior-consistent":[156],"modules":[157,177],"localize":[159],"competition,":[161],"then":[162],"evaluate":[163],"metric":[165,217],"reliability":[166],"per":[167],"module":[168],"according":[169],"sensitivity":[172],"score":[174],"magnitude.":[175],"For":[176],"where":[178],"scoring":[180],"unreliable,":[182],"switch":[184],"an":[186],"activation-independent":[187],"metric.":[188],"Finally,":[189],"adaptively":[191],"learn":[192],"module-wise":[193],"sparsity.":[194],"Extensive":[195],"experiments":[196],"multiple":[198],"tasks":[200],"demonstrate":[201],"GPrune-LLM's":[202],"improvements":[204],"post-compression":[206],"generalization,":[207],"particularly":[208],"at":[209],"sparsity,":[211],"reduced":[213],"dependence":[214],"choice.":[218]},"counts_by_year":[],"updated_date":"2026-03-18T06:31:55.123368","created_date":"2026-03-18T00:00:00"}