{"id":"https://openalex.org/W4414198566","doi":"https://doi.org/10.1109/dac63849.2025.11132899","title":"Grasp: Group-based Prediction of Activation Sparsity for Fast LLM Inference","display_name":"Grasp: Group-based Prediction of Activation Sparsity for Fast LLM Inference","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4414198566","doi":"https://doi.org/10.1109/dac63849.2025.11132899"},"language":"en","primary_location":{"id":"doi:10.1109/dac63849.2025.11132899","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11132899","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 62nd ACM/IEEE Design Automation Conference (DAC)","raw_type":"proceedings-article"},"type":"article","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/A5055766787","display_name":"Jiho Shin","orcid":"https://orcid.org/0000-0002-2697-8221"},"institutions":[{"id":"https://openalex.org/I124633538","display_name":"University of Seoul","ror":"https://ror.org/05en5nh73","country_code":"KR","type":"education","lineage":["https://openalex.org/I124633538"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jiho Shin","raw_affiliation_strings":["University of Seoul,Seoul,South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Seoul,Seoul,South Korea","institution_ids":["https://openalex.org/I124633538"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026611490","display_name":"Hoeseok Yang","orcid":"https://orcid.org/0000-0002-7929-7470"},"institutions":[{"id":"https://openalex.org/I16269868","display_name":"Santa Clara University","ror":"https://ror.org/03ypqe447","country_code":"US","type":"education","lineage":["https://openalex.org/I16269868"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hoeseok Yang","raw_affiliation_strings":["Santa Clara University,Santa Clara,CA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Santa Clara University,Santa Clara,CA,USA","institution_ids":["https://openalex.org/I16269868"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029634385","display_name":"Youngmin Yi","orcid":"https://orcid.org/0000-0001-9802-2109"},"institutions":[{"id":"https://openalex.org/I148751991","display_name":"Sogang University","ror":"https://ror.org/056tn4839","country_code":"KR","type":"education","lineage":["https://openalex.org/I148751991"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Youngmin Yi","raw_affiliation_strings":["Sogang University,Seoul,South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Sogang University,Seoul,South Korea","institution_ids":["https://openalex.org/I148751991"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.11444812,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9524999856948853,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9524999856948853,"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/T10028","display_name":"Topic Modeling","score":0.945900022983551,"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/inference","display_name":"Inference","score":0.6704999804496765},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6636999845504761},{"id":"https://openalex.org/keywords/grasp","display_name":"GRASP","score":0.6349999904632568},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5841000080108643},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5098000168800354},{"id":"https://openalex.org/keywords/approximate-inference","display_name":"Approximate inference","score":0.4659000039100647},{"id":"https://openalex.org/keywords/sign","display_name":"Sign (mathematics)","score":0.39640000462532043},{"id":"https://openalex.org/keywords/computational-complexity-theory","display_name":"Computational complexity theory","score":0.3594000041484833}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7904999852180481},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6704999804496765},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6636999845504761},{"id":"https://openalex.org/C171268870","wikidata":"https://www.wikidata.org/wiki/Q1486676","display_name":"GRASP","level":2,"score":0.6349999904632568},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5841000080108643},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5121999979019165},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5098000168800354},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.4659000039100647},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.4528000056743622},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.42879998683929443},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.40139999985694885},{"id":"https://openalex.org/C139676723","wikidata":"https://www.wikidata.org/wiki/Q1193832","display_name":"Sign (mathematics)","level":2,"score":0.39640000462532043},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.3594000041484833},{"id":"https://openalex.org/C134261354","wikidata":"https://www.wikidata.org/wiki/Q938438","display_name":"Statistical inference","level":2,"score":0.3522999882698059},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3508000075817108},{"id":"https://openalex.org/C200288055","wikidata":"https://www.wikidata.org/wiki/Q2621792","display_name":"Element (criminal law)","level":2,"score":0.35040000081062317},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.34950000047683716},{"id":"https://openalex.org/C45804977","wikidata":"https://www.wikidata.org/wiki/Q7239673","display_name":"Predictive modelling","level":2,"score":0.33180001378059387},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.28850001096725464},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.287200003862381},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.2833999991416931},{"id":"https://openalex.org/C32230216","wikidata":"https://www.wikidata.org/wiki/Q7882499","display_name":"Uncertainty quantification","level":2,"score":0.2825999855995178},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2822999954223633},{"id":"https://openalex.org/C111696304","wikidata":"https://www.wikidata.org/wiki/Q2303697","display_name":"Sorting","level":2,"score":0.26570001244544983},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.262800008058548}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/dac63849.2025.11132899","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11132899","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 62nd ACM/IEEE Design Automation Conference (DAC)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":5,"referenced_works":["https://openalex.org/W2152575748","https://openalex.org/W4320930577","https://openalex.org/W4385571157","https://openalex.org/W4404401018","https://openalex.org/W4410583234"],"related_works":[],"abstract_inverted_index":{"Optimizing":[0],"LLM":[1,31],"inference":[2],"has":[3],"become":[4],"increasingly":[5],"important":[6],"as":[7],"the":[8,17,21,78,96,111,155],"demand":[9],"for":[10,26,83,103],"efficient":[11,121],"on-device":[12],"deployments":[13],"grows.":[14],"To":[15],"reduce":[16],"computational":[18,130],"overhead":[19],"in":[20,119],"MLP":[22],"components,":[23],"which":[24,148],"account":[25],"a":[27,66],"significant":[28],"portion":[29],"of":[30,113],"inference,":[32],"ReLU-fied":[33,84],"LLMs":[34],"have":[35,46],"been":[36,47],"introduced":[37],"to":[38,49,109,133,150],"maximize":[39],"activation":[40,59,125],"sparsity.":[41,60],"Several":[42],"sparsity":[43,70,126,139],"prediction":[44,71,90,127,140],"methods":[45],"developed":[48],"efficiently":[50],"skip":[51],"unnecessary":[52],"memory":[53],"accesses":[54],"and":[55,100,129,142],"computations":[56],"by":[57,92],"predicting":[58],"In":[61],"this":[62],"paper,":[63],"we":[64],"propose":[65],"novel":[67],"magnitude-based,":[68],"training-free":[69],"technique":[72],"called":[73],"Grasp":[74,136],"that":[75],"builds":[76],"on":[77],"existing":[79],"sign":[80],"bitbased":[81],"method":[82,88],"LLMs.":[85],"The":[86],"proposed":[87],"enhances":[89],"accuracy":[91,128,141],"grouping":[93],"values":[94],"considering":[95],"distribution":[97],"within":[98],"vectors":[99],"explicitly":[101],"accounting":[102],"statistical":[104],"outliers.":[105],"This":[106],"allows":[107],"us":[108],"estimate":[110],"impact":[112],"each":[114],"element":[115],"more":[116],"accurately":[117],"yet":[118],"an":[120],"way,":[122],"improving":[123],"both":[124],"efficiency.":[131],"Compared":[132],"the-state-of-the-art":[134],"technique,":[135],"achieves":[137],"higher":[138,145],"$11":[143],"\\%$":[144],"skipping":[146],"efficiency,":[147],"corresponds":[149],"$1.85":[151],"\\times$":[152],"speedup":[153],"against":[154],"dense":[156],"inference.":[157]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}