{"id":"https://openalex.org/W4417086593","doi":"https://doi.org/10.48550/arxiv.2512.04296","title":"GRASP: GRouped Activation Shared Parameterization for Parameter-Efficient Fine-Tuning and Robust Inference of Transformers","display_name":"GRASP: GRouped Activation Shared Parameterization for Parameter-Efficient Fine-Tuning and Robust Inference of Transformers","publication_year":2025,"publication_date":"2025-12-03","ids":{"openalex":"https://openalex.org/W4417086593","doi":"https://doi.org/10.48550/arxiv.2512.04296"},"language":null,"primary_location":{"id":"pmh:oai:arXiv.org:2512.04296","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.04296","pdf_url":"https://arxiv.org/pdf/2512.04296","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2512.04296","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5031018070","display_name":"Malyaban Bal","orcid":"https://orcid.org/0000-0003-0287-2139"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Bal, Malyaban","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5032635465","display_name":"Abhronil Sengupta","orcid":"https://orcid.org/0000-0002-5545-4494"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sengupta, Abhronil","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5031018070"],"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/T10036","display_name":"Advanced Neural Network Applications","score":0.5267999768257141,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.5267999768257141,"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.07569999992847443,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.06989999860525131,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/robustness","display_name":"Robustness (evolution)","score":0.6717000007629395},{"id":"https://openalex.org/keywords/grasp","display_name":"GRASP","score":0.6658999919891357},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.6381000280380249},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6288999915122986},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.48179998993873596},{"id":"https://openalex.org/keywords/probabilistic-logic","display_name":"Probabilistic logic","score":0.4648999869823456},{"id":"https://openalex.org/keywords/gaussian-process","display_name":"Gaussian process","score":0.3917999863624573},{"id":"https://openalex.org/keywords/logarithm","display_name":"Logarithm","score":0.38199999928474426}],"concepts":[{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6717000007629395},{"id":"https://openalex.org/C171268870","wikidata":"https://www.wikidata.org/wiki/Q1486676","display_name":"GRASP","level":2,"score":0.6658999919891357},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.6381000280380249},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6288999915122986},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6123999953269958},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.48179998993873596},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.4648999869823456},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4120999872684479},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4065000116825104},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.3917999863624573},{"id":"https://openalex.org/C39927690","wikidata":"https://www.wikidata.org/wiki/Q11197","display_name":"Logarithm","level":2,"score":0.38199999928474426},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.36169999837875366},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.36070001125335693},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.3296000063419342},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.32359999418258667},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.3231000006198883},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.3174000084400177},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2930000126361847},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.265500009059906},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.26089999079704285},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C63002673","wikidata":"https://www.wikidata.org/wiki/Q2260590","display_name":"Scoring rule","level":2,"score":0.2538999915122986},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2538999915122986}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2512.04296","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.04296","pdf_url":"https://arxiv.org/pdf/2512.04296","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2512.04296","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2512.04296","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2512.04296","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.04296","pdf_url":"https://arxiv.org/pdf/2512.04296","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"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":{"Parameter-efficient":[0],"fine-tuning":[1],"(PEFT)":[2],"provides":[3],"a":[4,13,31,51,108],"scalable":[5],"alternative":[6],"to":[7,77,95,168],"full-model":[8],"adaptation":[9],"by":[10],"updating":[11],"only":[12],"small":[14],"subset":[15],"of":[16,41,66,74,153,161,175],"parameters":[17,68,166],"in":[18,117,164],"large":[19],"pre-trained":[20,97],"models.":[21],"We":[22],"introduce":[23],"GRASP":[24,147],"-":[25,30],"GRouped":[26],"Activation":[27],"Shared":[28],"Parameterization":[29],"lightweight":[32],"PEFT":[33,155],"framework":[34],"that":[35],"partitions":[36],"the":[37,64,72,75,96,151],"D-dimensional":[38],"token":[39],"representations":[40],"selected":[42],"layers":[43],"into":[44],"K":[45],"<<":[46],"D":[47],"groups":[48],"and":[49,54,120,142,170,187],"learns":[50,90],"shared":[52],"scaling":[53],"shifting":[55],"vector":[56],"for":[57,130,185],"each":[58],"group.":[59],"This":[60,103],"grouped":[61],"modulation":[62],"reduces":[63],"number":[65],"trainable":[67,165],"significantly":[69,121],"while":[70,157],"preserving":[71],"ability":[73],"model":[76],"learn":[78],"task-specific":[79],"features.":[80],"Building":[81],"on":[82,132],"this":[83],"formulation,":[84],"we":[85],"further":[86],"propose":[87],"StochGRASP,":[88],"which":[89],"Gaussian":[91],"distributions":[92],"as":[93],"perturbations":[94],"weights":[98,119],"rather":[99],"than":[100],"deterministic":[101,180],"values.":[102],"probabilistic":[104],"parameterization":[105],"along":[106],"with":[107],"noise-aware":[109],"loss":[110],"function":[111],"formulation":[112],"enables":[113],"modelling":[114],"hardware-level":[115],"variability":[116],"programmed":[118],"improves":[122],"robustness":[123],"under":[124],"non-ideal":[125],"inference":[126],"conditions-an":[127],"important":[128],"requirement":[129],"deployment":[131],"edge-based":[133],"emerging":[134],"AI":[135],"hardware.":[136],"Across":[137],"GLUE":[138],"(RoBERTa-base":[139],"&":[140],"RoBERTa-large)":[141],"E2E":[143],"NLG":[144],"(GPT-2":[145],"Medium),":[146],"matches":[148],"or":[149],"exceeds":[150],"performance":[152],"established":[154],"methods":[156],"achieving":[158],"an":[159],"order":[160],"magnitude":[162],"reduction":[163],"compared":[167],"LoRA":[169],"BitFit.":[171],"Under":[172],"varying":[173],"levels":[174],"noise,":[176],"StochGRASP":[177],"consistently":[178],"outperforms":[179],"variants,":[181],"demonstrating":[182],"its":[183],"suitability":[184],"energy-efficient":[186],"noise-prone":[188],"hardware":[189],"platforms.":[190]},"counts_by_year":[],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2025-12-06T00:00:00"}