{"id":"https://openalex.org/W7164922058","doi":"https://doi.org/10.48550/arxiv.2606.15161","title":"Beyond Layer Importance in Layer-wise Sparsity: An Inter-Layer Perturbation-Absorption Perspective","display_name":"Beyond Layer Importance in Layer-wise Sparsity: An Inter-Layer Perturbation-Absorption Perspective","publication_year":2026,"publication_date":"2026-06-13","ids":{"openalex":"https://openalex.org/W7164922058","doi":"https://doi.org/10.48550/arxiv.2606.15161"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.15161","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.15161","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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.15161","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5138698686","display_name":"Tao Jing","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jing, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138745151","display_name":"Ningxin Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Ningxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138711528","display_name":"Chen Kang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kang, Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138757770","display_name":"Dong Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Dong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138718447","display_name":"Changliang Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Changliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138698037","display_name":"Pengyuan Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Pengyuan","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":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":0.13590000569820404,"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":0.13590000569820404,"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":0.1151999980211258,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0989999994635582,"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/monotonic-function","display_name":"Monotonic function","score":0.571399986743927},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.5392000079154968},{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.5149000287055969},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.4828999936580658},{"id":"https://openalex.org/keywords/perturbation","display_name":"Perturbation (astronomy)","score":0.44749999046325684},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.40070000290870667},{"id":"https://openalex.org/keywords/decile","display_name":"Decile","score":0.3578000068664551},{"id":"https://openalex.org/keywords/perturbation-theory","display_name":"Perturbation theory (quantum mechanics)","score":0.350600004196167},{"id":"https://openalex.org/keywords/interpretability","display_name":"Interpretability","score":0.34369999170303345}],"concepts":[{"id":"https://openalex.org/C72169020","wikidata":"https://www.wikidata.org/wiki/Q194404","display_name":"Monotonic function","level":2,"score":0.571399986743927},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.5392000079154968},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.5149000287055969},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.4828999936580658},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.454800009727478},{"id":"https://openalex.org/C177918212","wikidata":"https://www.wikidata.org/wiki/Q803623","display_name":"Perturbation (astronomy)","level":2,"score":0.44749999046325684},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.4388999938964844},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.4237000048160553},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.40070000290870667},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.400299996137619},{"id":"https://openalex.org/C186060115","wikidata":"https://www.wikidata.org/wiki/Q30336093","display_name":"Biological system","level":1,"score":0.38839998841285706},{"id":"https://openalex.org/C147021879","wikidata":"https://www.wikidata.org/wiki/Q3027971","display_name":"Decile","level":2,"score":0.3578000068664551},{"id":"https://openalex.org/C174256460","wikidata":"https://www.wikidata.org/wiki/Q911364","display_name":"Perturbation theory (quantum mechanics)","level":2,"score":0.350600004196167},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.3495999872684479},{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.34369999170303345},{"id":"https://openalex.org/C56951928","wikidata":"https://www.wikidata.org/wiki/Q3539213","display_name":"Trimming","level":2,"score":0.3434999883174896},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.34040001034736633},{"id":"https://openalex.org/C2779525943","wikidata":"https://www.wikidata.org/wiki/Q1187300","display_name":"Grammaticality","level":3,"score":0.3310999870300293},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.32839998602867126},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3192000091075897},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.2989000082015991},{"id":"https://openalex.org/C125287762","wikidata":"https://www.wikidata.org/wiki/Q1758948","display_name":"Absorption (acoustics)","level":2,"score":0.289900004863739},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.28859999775886536},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.28600001335144043},{"id":"https://openalex.org/C75553542","wikidata":"https://www.wikidata.org/wiki/Q178161","display_name":"A priori and a posteriori","level":2,"score":0.275299996137619},{"id":"https://openalex.org/C129844170","wikidata":"https://www.wikidata.org/wiki/Q41299","display_name":"Quadratic equation","level":2,"score":0.2709999978542328},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.26759999990463257},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.2667999863624573},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.2572000026702881},{"id":"https://openalex.org/C137668524","wikidata":"https://www.wikidata.org/wiki/Q189813","display_name":"Censoring (clinical trials)","level":2,"score":0.2567000091075897},{"id":"https://openalex.org/C159774933","wikidata":"https://www.wikidata.org/wiki/Q902086","display_name":"Attenuation coefficient","level":2,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.15161","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.15161","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.15161","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.15161","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":"Preprint"},"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":{"The":[0],"considerable":[1],"layer-wise":[2,25],"redundancy":[3],"in":[4],"large":[5],"language":[6],"models":[7],"(LLMs)":[8],"has":[9],"established":[10],"non-uniform":[11],"sparsity":[12],"allocation":[13,26,30],"across":[14,110,133,193],"layers":[15,81,93,100],"as":[16,36,143],"the":[17,49,57,76,116,129,137,152],"standard":[18],"pruning":[19,148],"approach":[20],"for":[21],"efficient":[22],"compression.":[23],"Existing":[24],"methods":[27],"that":[28,177],"estimate":[29],"strategy":[31],"from":[32,44],"local":[33,45],"signals":[34],"such":[35],"activation":[37],"outliers":[38],"or":[39],"weight":[40],"spectra":[41],"mainly":[42],"derive":[43],"layer":[46,169],"importance,":[47],"whereas":[48],"final":[50],"post-pruning":[51],"performance":[52],"is":[53,122],"also":[54],"influenced":[55],"by":[56,182,185,191],"network's":[58],"subsequent":[59],"compensatory":[60],"capacity.":[61],"In":[62,89],"this":[63,68],"paper,":[64],"we":[65,163],"directly":[66],"characterize":[67],"property":[69],"through":[70],"controlled":[71],"perturbation":[72,144],"experiments.":[73],"We":[74],"make":[75],"following":[77],"empirical":[78],"findings.":[79],"First,":[80],"exhibit":[82],"highly":[83],"heterogeneous":[84],"responses":[85],"to":[86,139,147],"pruning-scale":[87],"perturbations.":[88],"most":[90],"cases,":[91],"early":[92],"amplify":[94],"perturbations,":[95],"while":[96],"middle":[97],"and":[98,112,136,170,180,187],"late":[99],"actively":[101],"absorb":[102],"them,":[103],"with":[104],"relative":[105],"L2":[106],"drift":[107],"decreasing":[108],"monotonically":[109],"depth":[111],"direction":[113],"realigning":[114],"toward":[115],"unperturbed":[117],"hidden-state":[118],"trajectory.":[119],"Second,":[120],"absorption":[121,140,166],"a":[123],"large-perturbation":[124],"phenomenon.":[125],"Under":[126],"small":[127],"perturbations":[128],"network":[130],"exhibits":[131],"amplification":[132],"all":[134],"layers,":[135],"transition":[138],"occurs":[141],"smoothly":[142],"magnitude":[145],"grows":[146],"scale.":[149],"This":[150],"enriches":[151],"linearized":[153],"accumulation":[154],"theory":[155],"underlying":[156],"related":[157],"works.":[158],"Building":[159],"on":[160],"these":[161],"findings,":[162],"define":[164],"an":[165,174],"coefficient":[167],"per":[168],"propose":[171],"absorption-aware":[172],"correction,":[173],"orthogonal":[175],"augmentation":[176],"improves":[178],"OWL":[179],"AlphaPruning":[181],"reducing":[183],"perplexity":[184],"7.13%":[186],"boosting":[188],"zero-shot":[189],"accuracy":[190],"1.02%":[192],"multiple":[194],"model":[195],"families":[196],"at":[197],"70%":[198],"sparsity.":[199]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-17T00:00:00"}